├── .env.example ├── .eslintrc.json ├── .github ├── funding.yml └── workflows │ └── test.yml ├── .gitignore ├── .husky └── pre-commit ├── .npmrc ├── .prettierignore ├── .prettierrc.cjs ├── fixtures └── bensbites.beehiiv.com │ ├── 1-billion-gpus-tiktok.md │ ├── 100-people-small-ai-2023.md │ ├── 1000-employees-let-go-googles-assistant-hardware-teams.md │ ├── 100000-tokens-single-prompt.md │ ├── 100m-fund-ai.md │ ├── 10m-fund-ai-gold-math.md │ ├── 19-insights-microsofts-future-work-report-2023.md │ ├── 2-year-old-ai-startup-gets-acquired-13b.md │ ├── 29bn-offer.md │ ├── 2x-15bn-valuations.md │ ├── 3-ways-ai-will-disrupt-business-models.md │ ├── 30min-ai-challenge.md │ ├── 3x-boost-open-ais-valuation.md │ ├── 9-ideas-profit-ai-gold-rush.md │ ├── a16z-discusses-investing-ideogram-midjourney-rival.md │ ├── a16z-investor-no-hole-ai-infrastructure-investment.md │ ├── a16z-wants-invest-consumer-ai.md │ ├── adobe-firefly-2-live-major-upgrades.md │ ├── ai-actions-zapier-bridge-ai-platform-5000-apps.md │ ├── ai-app-store.md │ ├── ai-apps-y-combinator.md │ ├── ai-art-officially-replaces-stock-images.md │ ├── ai-bookworms.md │ ├── ai-builders-starting-fear-misuse.md │ ├── ai-can-guess-3d-objects-moves.md │ ├── ai-can-help-people-perform-better-work-83ac.md │ ├── ai-changing-marketing-jobs.md │ ├── ai-chatbots-instagram.md │ ├── ai-chips-running.md │ ├── ai-chips-sealing-deal-flow-vcs.md │ ├── ai-coding-exploding.md │ ├── ai-collaboration-everyday-work-just-got-real.md │ ├── ai-cure-addiction.md │ ├── ai-deflationary.md │ ├── ai-demo-day.md │ ├── ai-email-assistant.md │ ├── ai-engineers-not-phds.md │ ├── ai-ethics-program-nyu-open-ai.md │ ├── ai-faqs-need-know.md │ ├── ai-feedback-human-feedback.md │ ├── ai-goes-tokyo-drift.md │ ├── ai-gold-rush.md │ ├── ai-helps-read-ancient-roman-scrolls-buried-volcano.md │ ├── ai-hiring-changing.md │ ├── ai-important-story-past-several-years.md │ ├── ai-improves-breast-cancer-detection.md │ ├── ai-jobs.md │ ├── ai-journalism-2023.md │ ├── ai-moatless-industry.md │ ├── ai-panic-campaigns-targeting-influence-public-policy.md │ ├── ai-partnerships-acquisitions-funding.md │ ├── ai-playing-matchmaker-valentines.md │ ├── ai-proteins.md │ ├── ai-replacing-programming.md │ ├── ai-sassy-not-sentient.md │ ├── ai-staff-google-now-betting-startup.md │ ├── ai-transparency-protocol.md │ ├── ai-video-isnt-patchy-anymore-almost-real.md │ ├── ai-website-generators.md │ ├── ai-will-change-courts-work.md │ ├── ai-will-save-world.md │ ├── ai-workplace.md │ ├── ai-wrappers-get-bad-wrap.md │ ├── ai-writing-trend.md │ ├── aigenerated-real.md │ ├── airbnb-acquires-ai-company.md │ ├── ais-impact-jobs-look-frontend-engineering.md │ ├── ais-mojo.md │ ├── ais-money-magnet-supercharged.md │ ├── aiverse.md │ ├── alexa-gets-ai-skills-characters-music-20questions.md │ ├── alexa-getting-generative-ai-boost-finally.md │ ├── always-online-models-perplexity.md │ ├── amazon-anthropic-partner-4b-deal.md │ ├── amazon-back-ai-game.md │ ├── amazon-bedrock-available-now.md │ ├── amazon-q-preview-assistant-work-business.md │ ├── amazon-rufus-change-shop-ai.md │ ├── amazon-search-bar-ai-repairs.md │ ├── amazon-will-teach-ai-free.md │ ├── amazon-will-use-ai-robots-speed-delivery.md │ ├── amazons-ai-sellers.md │ ├── amazons-copilot-called-bedrock.md │ ├── amazons-new-generative-ai-product-images.md │ ├── analysing-intercoms-ai-strategy.md │ ├── andrew-ng-debunks-fear-ai.md │ ├── anthropic-google-back-together.md │ ├── anthropic-raising.md │ ├── anthropic-releases-claude-21.md │ ├── anthropics-primer-challenges-ai-evaluation.md │ ├── anthropics-responsible-scaling-policy.md │ ├── apis-poppadoms.md │ ├── apple-increases-research-spend-ai.md │ ├── apple-removes-ai-apps-china.md │ ├── apples-ai-service.md │ ├── apples-doubling-ai-siri-messages.md │ ├── apples-entering-ai-hardware.md │ ├── arc-max-blends-ai-browser.md │ ├── arrakis-deserted.md │ ├── artificial-general-intelligence-lifetime.md │ ├── artists-can-poison-ai-models.md │ ├── atlassian-intelligence-live-jira-confluence.md │ ├── autonomous-agents-not-chatbots.md │ ├── aya-ai-languages.md │ ├── bada-bing.md │ ├── banning-chatgpt.md │ ├── bard-boy.md │ ├── bard-can-now-connect-google-apps-services.md │ ├── bard-gets-buff-logic.md │ ├── bard-learns-code.md │ ├── bard-song-googles-slow-culture.md │ ├── battle-gpu-wealth-rich-brats-vs-frugal-lads.md │ ├── battle-opensource-ai.md │ ├── bay-bridge-sf-compute-train-llms-cheap.md │ ├── bens-bites-hackathon-win-15k.md │ ├── better-aiwriting-tool.md │ ├── big-announcements-intel-innovation-2023.md │ ├── big-tech-searching-ai-profits-havent-found-yet.md │ ├── bing-ai-chrome-mobile.md │ ├── bing-chat-chrome-safari.md │ ├── brave-browser-adds-ai-chatbot.md │ ├── breakout-year-ai.md │ ├── bret-taylors-stealth-ai-startup-almost-unicorn.md │ ├── build-ai-app-without-code-5-minutes.md │ ├── build-aipowered-company.md │ ├── build-website-30-seconds-ai.md │ ├── business-relationships-ai-era.md │ ├── bytedance-using-gpt4-train-models.md │ ├── camera-makers-want-watermark-human-images-not-ai-ones.md │ ├── camera-movements-ai.md │ ├── can-ai-get-patent-latest-guidelines.md │ ├── can-ai-take-your-job-a-copywriter-tested-it.md │ ├── can-find-cork.md │ ├── canva-becomes-ai-magician-can.md │ ├── canva-launches-magic-studio-partners-runway-ml-video.md │ ├── captchas-gone.md │ ├── character-ai-now-lets-join-group-chat-ai-characters.md │ ├── character-ai-talks-google-next-round.md │ ├── chatgpt-aftermath.md │ ├── chatgpt-api-10x-cheaper.md │ ├── chatgpt-becomes-secure.md │ ├── chatgpt-enterprise-new-exclusive-bens-bites-interview.md │ ├── chatgpt-plugins-iphone-moment-ai.md │ ├── chatgpt-plugins.md │ ├── chatgpt-record-revenue-mobile.md │ ├── chatgpt-upgrades-spotify-playlists-ai-checklists.md │ ├── chatgpt-vision-voice.md │ ├── chatgpt-won-123k.md │ ├── chief-ai-officer.md │ ├── china-ai-wars.md │ ├── china-allows-copyright-ai-images.md │ ├── chinas-chatgpt-rival-coming.md │ ├── chinas-status-ai.md │ ├── chinese-black-market-ai-chips.md │ ├── claude-2-enters-arena.md │ ├── clearbits-team-uses-ai.md │ ├── cloudflare-partners-hugging-face-make-ai-models-accessible-affordable.md │ ├── code-interpreter-gpt4-api.md │ ├── comeback-gans-new-image-editing-research.md │ ├── companies-use-ai-customer-onboarding.md │ ├── concentration-ai-jobs-needs-intervention.md │ ├── configure-chatgpt-custom-instructions.md │ ├── conversational-answer-engine-huh.md │ ├── copilot-get-specific-button-keyboards.md │ ├── copilot-pro-upgrades-ai-experience-microsoft-users.md │ ├── daily-digest-1-year-sydney.md │ ├── daily-digest-10m-math-geeks.md │ ├── daily-digest-251-mistake-correction.md │ ├── daily-digest-271.md │ ├── daily-digest-272.md │ ├── daily-digest-abandoned-models.md │ ├── daily-digest-agent-attack.md │ ├── daily-digest-ai-2024.md │ ├── daily-digest-ai-characters-business.md │ ├── daily-digest-ai-got-cheaper.md │ ├── daily-digest-ai-jingles.md │ ├── daily-digest-ai-legends-fear.md │ ├── daily-digest-ai-not-enough-stop-google.md │ ├── daily-digest-ai-poison-new-unicorn.md │ ├── daily-digest-ai-safety-summit-day-1.md │ ├── daily-digest-ai-shareholders.md │ ├── daily-digest-ai-taking-courts.md │ ├── daily-digest-ai-threat-detection.md │ ├── daily-digest-back.md │ ├── daily-digest-biggest-weekend-ai.md │ ├── daily-digest-birthday-season.md │ ├── daily-digest-build-together-win-together.md │ ├── daily-digest-can-gemini-save-google.md │ ├── daily-digest-cant-stop-growing.md │ ├── daily-digest-chatgpt-plus-back.md │ ├── daily-digest-china-japan.md │ ├── daily-digest-claude-pi.md │ ├── daily-digest-crypto-journalism-ai.md │ ├── daily-digest-days-decades-happen.md │ ├── daily-digest-deal-no-deal.md │ ├── daily-digest-detecting-ai-images.md │ ├── daily-digest-draw-bard.md │ ├── daily-digest-else-got-access.md │ ├── daily-digest-everyone-said-kumbaya.md │ ├── daily-digest-false-agi-risk.md │ ├── daily-digest-first-ai-elections.md │ ├── daily-digest-free-ai-coding-2023.md │ ├── daily-digest-gemini-leaks.md │ ├── daily-digest-gen-ai-daily-life.md │ ├── daily-digest-go.md │ ├── daily-digest-going-ai.md │ ├── daily-digest-google-choose-attack-defend.md │ ├── daily-digest-google-products-get-ai-glow.md │ ├── daily-digest-google-samsung-bring-ai-phones.md │ ├── daily-digest-google-strikes-back.md │ ├── daily-digest-googles-secret-stock-ai.md │ ├── daily-digest-halfbillion-valuations.md │ ├── daily-digest-impossible-results-fake-identities.md │ ├── daily-digest-internal-ai-use.md │ ├── daily-digest-know.md │ ├── daily-digest-lawsuit-may-open-can-worms.md │ ├── daily-digest-lean-google-zucks-ai.md │ ├── daily-digest-local-ai-personal-ai.md │ ├── daily-digest-look-back-ai.md │ ├── daily-digest-lot-money-flying-around.md │ ├── daily-digest-love-air.md │ ├── daily-digest-mark-zuckerbergs-ai-update.md │ ├── daily-digest-maybe-money.md │ ├── daily-digest-meta-vs-midjourney.md │ ├── daily-digest-microsoft-made-google-sing.md │ ├── daily-digest-mindblowing-mistral.md │ ├── daily-digest-miracles-will-wait.md │ ├── daily-digest-musk-ai-born.md │ ├── daily-digest-new-codellama-better-gpt4.md │ ├── daily-digest-new-llm-town.md │ ├── daily-digest-new-open-source-model-intel.md │ ├── daily-digest-nextgen-ai-videos.md │ ├── daily-digest-nvidia-strikes.md │ ├── daily-digest-nyt-sues-openai.md │ ├── daily-digest-oh-great-google.md │ ├── daily-digest-open-source-gpts.md │ ├── daily-digest-open-source-multimodal.md │ ├── daily-digest-openai-isnt-dead-not-just-yet.md │ ├── daily-digest-prepared-whats-next-openai.md │ ├── daily-digest-quick.md │ ├── daily-digest-realistic-text-video.md │ ├── daily-digest-reality-ai-jobs.md │ ├── daily-digest-regulation-copyright.md │ ├── daily-digest-research-alpha.md │ ├── daily-digest-robots-learning-tricks.md │ ├── daily-digest-safe-open.md │ ├── daily-digest-shaky-demo-ai-wearables.md │ ├── daily-digest-siri-chatgpt-friends-now.md │ ├── daily-digest-sneaky-move-chinese-llms.md │ ├── daily-digest-social-leaders-ai.md │ ├── daily-digest-startups-models.md │ ├── daily-digest-successful-ai-product.md │ ├── daily-digest-tailwinds-nvidia.md │ ├── daily-digest-take-deep-breath.md │ ├── daily-digest-talk-anyone.md │ ├── daily-digest-tools-make-safe-ai.md │ ├── daily-digest-voice-translation.md │ ├── daily-digest-whats-possible-everything.md │ ├── daily-digest-wrong-path-ahead.md │ ├── dall-e-3-image-generation-chatgpt.md │ ├── database-ai-tools-every-use-case.md │ ├── day-1-uks-ai-safety-summit-heres-need-know.md │ ├── decline-gpt4.md │ ├── deel-uses-ai-business.md │ ├── ditch-coffee-get-ai.md │ ├── double-win-open-ai.md │ ├── dropbox-announces-universal-ai-search.md │ ├── duolingo-cuts-10-contractors-uses-ai.md │ ├── editing-image-ease.md │ ├── eleven-labs-grants-startups-need-ai-voices.md │ ├── eleven-labs-launches-ai-dubbing-feature-29-languages.md │ ├── eleven-labs-new-ai-unicorn.md │ ├── eu-goes-hard-ai.md │ ├── eureka-robots-can-pen-spinning-tricks.md │ ├── eus-ai-act-gets-one-step-ahead.md │ ├── everyone-gets-chatbot.md │ ├── everything-changed.md │ ├── exnvidia-engineer-nvidias-moat-big-tech-dominance-government-regulation.md │ ├── fashion-week-ai-wearables.md │ ├── figjam-ai-brainstorm-ai-figma.md │ ├── figma-ai-tool-magic-videos-dance-generation.md │ ├── figma-makes-chatgpt-visual.md │ ├── figmas-way-ahead-ai-without-adobe.md │ ├── filtering-harmful-content-using-ai.md │ ├── final-beatles-song-ai-official.md │ ├── final-beatles-song-now.md │ ├── finding-unique-features-inside-llms-interpretability-research-anthropic.md │ ├── fireflies-saving-lives.md │ ├── fireside-chat-sam-altman.md │ ├── first-ai-laugh-lex-fridman-interviews-richard-feynman.md │ ├── former-quartz-leader-building-new-york-times-ai-team.md │ ├── fresh-openai-new-embedding-models-cost-savings.md │ ├── frontier-model-forum-google-microsoft-open-ai-anthropic.md │ ├── frontier-model-forum.md │ ├── fuyu8b-adept-multimodal-open-source.md │ ├── gemini-pro-15-google-10m-content-window.md │ ├── gemini-pro-api-now-live.md │ ├── get-team-use-ai.md │ ├── getting-started-building-gpts.md │ ├── github-universe-copilot.md │ ├── going-openai-dev-day.md │ ├── google-add-image-generation-long-drafts-search-ai.md │ ├── google-announces-chatgpt-rival-microsoft-ready.md │ ├── google-assistant-get-llm-update.md │ ├── google-assures-companies-legal-cover-using-ai-models.md │ ├── google-bard-can-now-speak.md │ ├── google-brain-deepmind-merge-together.md │ ├── google-doesnt-trust-bard.md │ ├── google-drops-ai-think-tanks.md │ ├── google-gemini-not-coming-year.md │ ├── google-goes-ai.md │ ├── google-isnt-ready-give.md │ ├── google-launches-gemini-advanced-compete-openai.md │ ├── google-maps-start-generative-ai-pilot-local-guides.md │ ├── google-planning-breakup-broadcom-wants-build-tpus-inhouse.md │ ├── google-search-competitor.md │ ├── google-secret-ai-tool.md │ ├── google-spending-2b-anthropic.md │ ├── google-starts-bringing-ai-chrome.md │ ├── google-takes-openais-high-pay-offers-ai-talent.md │ ├── google-will-add-bard-assistant.md │ ├── googles-bard-starts-creating-images-like.md │ ├── googles-gemini-around-corner.md │ ├── googles-gemini-beats-gpt4-cost.md │ ├── googles-gonna-keep-ball-rolling.md │ ├── googles-new-ai-solves-math-olympiad-problems.md │ ├── googles-new-image-source-checker.md │ ├── government-bans-will-ai-emerge-victorious.md │ ├── gpt3-alternative.md │ ├── gpt3-google-sheets.md │ ├── gpt3-playground.md │ ├── gpt3-summariser-reads-websites.md │ ├── gpt4-around-corner.md │ ├── gpt4-details-leaked.md │ ├── gpt4-rumours.md │ ├── gpt4.md │ ├── gpt5-works-sam-altman.md │ ├── grok-grok-grok-ai-model-elon-musk.md │ ├── hardware-bets-software-wins.md │ ├── hd-image-creation.md │ ├── heres-gets-paid-ai.md │ ├── honestly-loads-big-stuff-today.md │ ├── hugging-face-google-cloud-teaming.md │ ├── humane-demoes-ai-pin-new-gen-wearable-ai.md │ ├── humane-reveals-ai-pin-paris-fashion-week.md │ ├── ibm-puts-500m-venture-fund-enterprise-ai.md │ ├── ignitetech-using-ai-change-business.md │ ├── image-generators-get-off-hook-lawsuit.md │ ├── images-just-got-crazier.md │ ├── inflection2-makes-bold-claims.md │ ├── inside-eacc-new-religion-silicon-valley.md │ ├── inside-look-agi-house-epicenter-sf-ai-hackathons.md │ ├── inside-story-building-glean.md │ ├── intel-starts-new-ai-software-company.md │ ├── interviewing-ceo-scale.md │ ├── iphone-x-chatgpt.md │ ├── jasper-cuts-internal-valuation-growth-slows.md │ ├── jumping-exploding-ai-trends.md │ ├── key-statements-openai-wef-davos.md │ ├── lamini-ditches-nvidia-favour-amd.md │ ├── leading-news-websites-now-block-ai.md │ ├── leaks-expect-open-ais-first-dev-day.md │ ├── lenovo-nvidia-announce-hybrid-ai-solutions.md │ ├── let-ai-games-begin-google-anthropic-vs-microsoft-open-ai.md │ ├── lets-think-step-step.md │ ├── linkedin-adding-ai-across.md │ ├── llm-beginners-guide.md │ ├── llm-suffer-reversal-curse.md │ ├── llms-making-robots-smarter.md │ ├── llms-vs-search-engines.md │ ├── london-office-open-ai.md │ ├── looking-back-llms.md │ ├── lyria-googles-new-ai-music-model.md │ ├── make-songs-copilot-suno.md │ ├── makebelief-rules-ai-regulation.md │ ├── marc-andreessens-take-ai-unemployment.md │ ├── medical-llms-testing-shows-good-signs.md │ ├── meet-claude.md │ ├── meeting-now-email.md │ ├── meta-claims-sota-art-generation.md │ ├── meta-launches-ai-chatbots-stickers.md │ ├── meta-losing-ai-talent.md │ ├── meta-open-source-agi.md │ ├── meta-paying-celebrities-millions-dollars-ai-chatbots.md │ ├── meta-release-ai-persona-chatbots-week.md │ ├── meta-shows-new-image-video-generation-models.md │ ├── meta-trying-label-ai-content-platforms.md │ ├── metas-ai-upgrades.md │ ├── metas-ai-warrior-llama.md │ ├── metas-computer-vision-tool.md │ ├── metas-new-model.md │ ├── metas-secret-voice-model.md │ ├── microsoft-1-google-0-100bn-mistake.md │ ├── microsoft-chip-shoulder.md │ ├── microsoft-copilot-everyday-ai-companion.md │ ├── microsoft-delivered-will-google.md │ ├── microsoft-ignites-copilot-era.md │ ├── microsoft-launching-spree.md │ ├── microsoft-makes-new-ai-team-small-models.md │ ├── microsoft-new-ai-chip-can-come-next-month.md │ ├── microsoft-partnering-semafor-boost-news-reporting-ai.md │ ├── microsoft-researchers-accidentally-exposed-38tb-private-data.md │ ├── microsoft-stepping-gear.md │ ├── microsoft-surface-ai-event-nyc-expect.md │ ├── microsoft-teams-chatgpt-7.md │ ├── microsofts-copilot-gets-better-looks-images-new-model-comes-everywhere.md │ ├── microsofts-fills-seat-openais-board.md │ ├── microsofts-next-copilots-copilot-sales-service-now-grabs.md │ ├── microsofts-plans-beyond-open-ai.md │ ├── microsofts-plans-dev-day.md │ ├── microsofts-situationship-openai.md │ ├── midjourney-can-outpaint-now.md │ ├── mistral-ai-drops-new-model-mixture-experts.md │ ├── mistral-ai-openai-competitor-rocketed-2bn-12-months.md │ ├── multimodal-next-step-ai-research-meta-thinks-yes.md │ ├── music-image.md │ ├── nadella-testifies-google-apple-deal.md │ ├── nasa-ibms-ai-model.md │ ├── near-future-ai-becoming-clearer.md │ ├── new-ai-alliance-wants-open-responsible-ai.md │ ├── new-ai-business-models.md │ ├── new-ai-image-generator-getty-images.md │ ├── new-entries-aws-bain.md │ ├── new-google-search.md │ ├── new-image-model-stable-diffusion.md │ ├── new-image-speech-capabilities.md │ ├── new-letter-ai-safety-risk.md │ ├── new-match-found-open-source-funding.md │ ├── new-search-town.md │ ├── newsletter.json │ ├── no-arr-just-akr-always-keep-raising.md │ ├── no-cloud-access-china.md │ ├── no-mention-ai-wwdc.md │ ├── no-mundane-work.md │ ├── nocode-ai-apps.md │ ├── nocode-model-builder.md │ ├── nontechnical-people-using-ai-code.md │ ├── normal-people-define-rules-ai.md │ ├── nvidia-announces-new-ai-chip-h200.md │ ├── nvidia-building-chat-pdf-apps-im-not-joking.md │ ├── nvidia-cooking-sushai-japan.md │ ├── nvidia-latest-launch-spree.md │ ├── nvidia-lose-big-china.md │ ├── nvidia-makes-easy-run-local-llms.md │ ├── nvidia-market-cap-almost-1-trillion-dollars.md │ ├── nvidias-new-team-will-build-custom-ai-chips-meta-openai.md │ ├── old-copyrights-case-can-influence-ai.md │ ├── one-year-microsoft-copilot-new-features.md │ ├── open-ai-acquihiring-now.md │ ├── open-ai-adds-voice-chat-images-chatgpt.md │ ├── open-ai-announces-gpts-work-automatically.md │ ├── open-ai-back.md │ ├── open-ai-building-two-type-ai-agents.md │ ├── open-ai-calls-experts-red-teaming.md │ ├── open-ai-fights-back-court.md │ ├── open-ai-fires-sam-altman-weekend-drama-summarized.md │ ├── open-ai-gives-employees-time-cash-shares.md │ ├── open-ai-introduces-data-partnerships.md │ ├── open-ai-isnt-ready-play-spot-ai-image.md │ ├── open-ai-makes-licensing-deal-alex-springer.md │ ├── open-ai-partners-arizona-state-university.md │ ├── open-ai-planning-features-developers.md │ ├── open-ai-releases-dalle-3.md │ ├── open-ai-seeking-a-90-billion-valuation.md │ ├── open-ai-share-sale-will-continue.md │ ├── open-ai-shuts-ai-detector.md │ ├── open-ai-vs-google-game.md │ ├── open-ai-wants-chatgpt-classroom.md │ ├── open-ai-wants-google-researchers-even-pay-millions.md │ ├── open-ais-new-values-agi-aligned.md │ ├── open-ais-preparedness-framework.md │ ├── open-ais-tricky-path-corporate-sales.md │ ├── open-source-ai.md │ ├── open-source-launch-day.md │ ├── openai-adds-election-guardrails-chatgpt.md │ ├── openai-adds-metadata-ai-images-chatgpt.md │ ├── openai-common-sense-media-partner-kidsfriendly-ai.md │ ├── openai-exploring-making-ai-chips.md │ ├── openai-hustles-beat-google-launch-multimodal-llm.md │ ├── openai-kills-chat-pdf-apps.md │ ├── openai-love-journalism.md │ ├── openai-signed-secret-letter-buy-brainlike-ai-chips.md │ ├── openai-stops-stateactors-abusing-ai.md │ ├── openai-update.md │ ├── openai-wants-pay-publishers-not-much.md │ ├── openai-will-deal-first-ai-elections.md │ ├── openais-3d-dalle-ai-modern-marketing-gpt4-10x-better-chatgpt.md │ ├── openais-chatgpt.md │ ├── openais-new-experiment-crowdsourcing-ai-policy.md │ ├── openais-newest-model.md │ ├── openais-tellall-interview-robots-flips-ai-nhs.md │ ├── optout-google-ai-ranking-google-search.md │ ├── palm-2-big-bard.md │ ├── pause-ai-immediately.md │ ├── perplexity-launches-low-latency-api-open-source-llms.md │ ├── personal-search-engine.md │ ├── phind-claims-beats-gpt4.md │ ├── pi-new-ai-chatbot.md │ ├── pika-labs-bringing-heat-ai-videos.md │ ├── pixels-possibilities-ai-vision.md │ ├── prefer-flan.md │ ├── pretty-unbelievable.md │ ├── promise-challenges-crypto-ai-applications-vitalik.md │ ├── protect-keys.md │ ├── pulitzer-winner-sues-open-ai.md │ ├── purple-llama-meta-evals-models-open-source-safety.md │ ├── quora-raises-new-funding-ai-chatbot-platform-poe.md │ ├── real-people-fake-stories.md │ ├── real-winner-ai.md │ ├── recordspeed-ai-images.md │ ├── reddit-wants-block-search-crawlers.md │ ├── regulating-ai-world.md │ ├── regulations-way.md │ ├── replit-ai-everyone.md │ ├── responsible-innovation-lab-one-ai-protocol.md │ ├── reverse-prompt-engineering.md │ ├── rewind-pendant-new-ai-wearable-town.md │ ├── rise-ai-sidekicks.md │ ├── roblox-assistant-make-game-dev-easier.md │ ├── robot-code.md │ ├── robots-recycle.md │ ├── rogue-ai-concern-cope.md │ ├── rogue-ai-drone-didnt-exist.md │ ├── safety-red-teaming-compliance.md │ ├── sam-altman-raising-money-building-semiconductor-factories.md │ ├── sam-altman-wants-make-hardware-ai-now.md │ ├── sam-altmans-letter-firerehire-openai.md │ ├── sam-altmans-testimony-ai-regulation.md │ ├── samsung-s24-talks-big-ai.md │ ├── samsung-unveils-llm-samsung-gauss.md │ ├── sandwich-video-uses-ai-business.md │ ├── satya-nadella-talks-open-ai-sam-altman.md │ ├── satya-nadellas-annual-letter-shareholders.md │ ├── saucy-ai-models.md │ ├── scale-ais-new-research-lab.md │ ├── scale-collaborates-nvidia-nemo-steerlm.md │ ├── science.md │ ├── scrambling-attempts-openai-board.md │ ├── sdxl-turbo-generates-images-youre-typing.md │ ├── seamless-speech-speech-translation-meta.md │ ├── secret-talks-among-us-companies-chinese-experts-ai-safety.md │ ├── sequoia-act-two-generative-ai.md │ ├── sequoia-implications-gpu-overinvestment.md │ ├── seven-figure-ai-salaries-not-far.md │ ├── shaky-economics-gpu-reselling-startups.md │ ├── shopify-others-hop-ai-bandwagon.md │ ├── singers-choose-make-money-deepfakes.md │ ├── slack-launches-native-ai-features-enterprises.md │ ├── sleeper-llms-bypass-current-safety-alignment-techniques.md │ ├── snap-releases-lens-studio-50-chatgpt-integration.md │ ├── solve-ai-discrimination.md │ ├── sota-speech-text-translation.md │ ├── sparrow.md │ ├── spotify-translates-podcasts-new-languages.md │ ├── stability-ai-release-llm.md │ ├── stability-ai-starts-20month-subscription-plan.md │ ├── stability-ais-new-model-stablelm-zephyr-3b-can-run-smartphones.md │ ├── stable-diffusion-20.md │ ├── stable-diffusion-music.md │ ├── stable-diffusion-phone.md │ ├── stack-overflow-fights-back.md │ ├── startup-profile-autoblocks.md │ ├── startup-profle-metal.md │ ├── state-ai-2023-report-air-street-capital.md │ ├── state-ai-report-summarised.md │ ├── stopping-innovation-companies-trying-get-ahead-ai.md │ ├── story-replicate-350m-company-powering-milliondollar-ai-companies.md │ ├── superalignment-superintelligence.md │ ├── tab-whats-new-ai-necklace.md │ ├── taking-week-off-exploding-ai-trends.md │ ├── techniques-improve-claudes-accuracy.md │ ├── teslas-making-bunker-supercomputer-dojo.md │ ├── textto3d.md │ ├── texttomusic-textto4d.md │ ├── texttoproduct.md │ ├── tiktok-text.md │ ├── tiktoks-new-chatbot.md │ ├── top-ai-models-offer-little-transparency-even-open-source-ones.md │ ├── toss-coin-bard-advanced.md │ ├── uk-government-announces-ai-advisory-service.md │ ├── uk-government-gets-hands-top-ai-models.md │ ├── unicorn-growth-really-distributed.md │ ├── unified-microsoft-copilot-across-windows.md │ ├── universal-music-sues-anthropic.md │ ├── unseen-startups-insane-funding.md │ ├── update-msft-openai.md │ ├── updates-open-ai-dev-day.md │ ├── us-govt-goes-ai-school.md │ ├── us-warns-eu-ai-act.md │ ├── use-ai-get-better-email.md │ ├── vector-embeddings-also-store-money.md │ ├── vision-language-model.md │ ├── voice-actors-can-now-earn-money-elevenlabs.md │ ├── volkswagen-planning-add-chatgpt-cars.md │ ├── wake-call-many-jobs.md │ ├── wave-chatgpt-apps.md │ ├── webbrowsing-copilot.md │ ├── whats-deal-bidens-executive-order-ai.md │ ├── whats-going-open-ai.md │ ├── whats-new-chatgpt-memory-temporary-chats.md │ ├── whats-next-open-ai.md │ ├── white-house-force-cloud-companies-disclose-ai-customers.md │ ├── white-house-hackathon.md │ ├── white-house-mega-deal-ai-companies.md │ ├── whos-hiring-ai-week.md │ ├── will-open-source-win.md │ ├── windows-os-now-aiintegrated.md │ ├── windows-releases-big.md │ ├── wordle-prompts.md │ ├── write-app-idea-get-app.md │ ├── wrote.md │ ├── xbox-inworld-partner-make-games-feel-natural.md │ ├── youtube-goes-ai-music.md │ ├── youtube-summaries.md │ ├── youtube-wants-use-ai-creating-videos.md │ ├── youtubes-ceo-hints-rollout-aipowered-features.md │ └── zapconnect-2023-building-ai-easier-zapier.md ├── license ├── next.config.mjs ├── package.json ├── pnpm-lock.yaml ├── postcss.config.cjs ├── public ├── favicon.ico ├── icon.png ├── logo-light.png ├── social.jpg └── social.png ├── readme.md ├── src ├── bin │ ├── export-newsletter-links-csv.ts │ ├── process-newsletter-links.ts │ ├── resolve-beehiiv-newsletter.ts │ ├── scratch.ts │ ├── upsert-newsletter-links-meilisearch.ts │ └── upsert-newsletter-links.ts ├── components │ ├── ActiveLink │ │ └── ActiveLink.tsx │ ├── Button │ │ ├── Button.tsx │ │ └── styles.module.css │ ├── Footer │ │ ├── Footer.tsx │ │ └── styles.module.css │ ├── Header │ │ ├── Header.tsx │ │ ├── Logo.tsx │ │ └── styles.module.css │ ├── HeroButton │ │ ├── HeroButton.tsx │ │ └── styles.module.css │ ├── Layout │ │ ├── Layout.tsx │ │ ├── providers.tsx │ │ └── styles.module.css │ ├── LoadingSpinner │ │ ├── LoadingSpinner.tsx │ │ └── styles.module.css │ ├── Markdown │ │ ├── Markdown.tsx │ │ └── styles.module.css │ ├── PageHead │ │ └── PageHead.tsx │ ├── SearchOptions │ │ ├── SearchOptions.tsx │ │ └── styles.module.css │ ├── SearchResult │ │ ├── SearchResult.tsx │ │ └── styles.module.css │ ├── SearchResults │ │ ├── SearchResults.tsx │ │ └── styles.module.css │ └── SearchResultsList │ │ ├── SearchResultsList.tsx │ │ └── styles.module.css ├── icons │ ├── Clear.tsx │ ├── Discord.tsx │ ├── External.tsx │ ├── GitHub.tsx │ ├── Globe.tsx │ ├── Search.tsx │ ├── Twitter.tsx │ └── index.ts ├── lib │ ├── bootstrap.ts │ ├── config.ts │ ├── hooks │ │ └── search.ts │ ├── markdown-to-html.ts │ └── readme.md ├── pages │ ├── _app.tsx │ ├── _document.tsx │ ├── about │ │ ├── index.tsx │ │ └── styles.module.css │ ├── api │ │ └── search.ts │ ├── index.module.css │ └── index.tsx ├── server │ ├── api.ts │ ├── beehiiv.ts │ ├── config.ts │ ├── embedding.ts │ ├── got.ts │ ├── markdown.ts │ ├── readme.md │ ├── resolve-twitter-data.ts │ ├── services │ │ ├── iframely.ts │ │ ├── meilisearch.ts │ │ ├── openai.ts │ │ ├── pinecone.ts │ │ ├── storage.ts │ │ └── twitter.ts │ ├── thumbnails.ts │ ├── twitter.test.ts │ ├── twitter.ts │ ├── types.ts │ ├── unfurl-tweet.test.ts │ ├── unfurl-tweet.ts │ ├── url-utils.ts │ └── utils.ts └── types │ └── index.ts ├── styles └── globals.css ├── tailwind.config.cjs └── tsconfig.json /.env.example: -------------------------------------------------------------------------------- 1 | # ------------------------------------------------------------------------------ 2 | # This is an example .env file. 3 | # 4 | # All of these environment vars must be defined either in your environment or in 5 | # a local .env file in order to run the demo for this project. 6 | # ------------------------------------------------------------------------------ 7 | 8 | # ----------------------------------------------------------------------------- 9 | # OpenAI 10 | # ----------------------------------------------------------------------------- 11 | 12 | OPENAI_API_KEY= 13 | 14 | # ----------------------------------------------------------------------------- 15 | # Pinecone Vector Search 16 | # ----------------------------------------------------------------------------- 17 | 18 | PINECONE_API_KEY= 19 | PINECONE_BASE_URL= 20 | PINECONE_NAMESPACE= 21 | -------------------------------------------------------------------------------- /.eslintrc.json: -------------------------------------------------------------------------------- 1 | { 2 | "extends": "next/core-web-vitals" 3 | } 4 | -------------------------------------------------------------------------------- /.github/funding.yml: -------------------------------------------------------------------------------- 1 | github: [transitive-bullshit] 2 | -------------------------------------------------------------------------------- /.github/workflows/test.yml: -------------------------------------------------------------------------------- 1 | name: CI 2 | 3 | on: [push, pull_request] 4 | 5 | jobs: 6 | test: 7 | name: Test Node.js ${{ matrix.node-version }} 8 | runs-on: ubuntu-latest 9 | strategy: 10 | fail-fast: false 11 | matrix: 12 | node-version: 13 | - 19 14 | - 18 15 | - 16 16 | 17 | steps: 18 | - name: Checkout 19 | uses: actions/checkout@v3 20 | 21 | - name: Install Node.js 22 | uses: actions/setup-node@v3 23 | with: 24 | node-version: ${{ matrix.node-version }} 25 | 26 | - name: Install pnpm 27 | uses: pnpm/action-setup@v2 28 | id: pnpm-install 29 | with: 30 | version: 8 31 | run_install: false 32 | 33 | - name: Get pnpm store directory 34 | id: pnpm-cache 35 | shell: bash 36 | run: | 37 | echo "STORE_PATH=$(pnpm store path)" >> $GITHUB_OUTPUT 38 | 39 | - uses: actions/cache@v3 40 | name: Setup pnpm cache 41 | with: 42 | path: ${{ steps.pnpm-cache.outputs.STORE_PATH }} 43 | key: ${{ runner.os }}-pnpm-store-${{ hashFiles('**/pnpm-lock.yaml') }} 44 | restore-keys: | 45 | ${{ runner.os }}-pnpm-store- 46 | 47 | - name: Install dependencies 48 | run: pnpm install --frozen-lockfile 49 | 50 | - name: Run test 51 | run: pnpm run test 52 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | # See https://help.github.com/articles/ignoring-files/ for more about ignoring files. 2 | 3 | *.swp 4 | .idea 5 | 6 | # dependencies 7 | /node_modules 8 | /.pnp 9 | .pnp.js 10 | 11 | # testing 12 | /coverage 13 | 14 | # next.js 15 | /.next/ 16 | /out/ 17 | 18 | # production 19 | /build 20 | 21 | # misc 22 | .DS_Store 23 | *.pem 24 | 25 | # debug 26 | npm-debug.log* 27 | yarn-debug.log* 28 | yarn-error.log* 29 | .pnpm-debug.log* 30 | 31 | # local env files 32 | .env*.local 33 | 34 | # vercel 35 | .vercel 36 | 37 | # typescript 38 | *.tsbuildinfo 39 | next-env.d.ts 40 | 41 | # local env files 42 | .env 43 | .env.local 44 | .env.build 45 | .env.development.local 46 | .env.test.local 47 | .env.production.local 48 | 49 | # data dumps 50 | out/ 51 | 52 | fixtures/*/links.csv 53 | fixtures/*/links.json 54 | fixtures/*/twitter.json 55 | -------------------------------------------------------------------------------- /.husky/pre-commit: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env sh 2 | . "$(dirname -- "$0")/_/husky.sh" 3 | 4 | npm run pre-commit 5 | -------------------------------------------------------------------------------- /.npmrc: -------------------------------------------------------------------------------- 1 | enable-pre-post-scripts=true 2 | -------------------------------------------------------------------------------- /.prettierignore: -------------------------------------------------------------------------------- 1 | .snapshots/ 2 | build/ 3 | dist/ 4 | node_modules/ 5 | .next/ 6 | .vercel/ 7 | third-party/ 8 | out/ -------------------------------------------------------------------------------- /.prettierrc.cjs: -------------------------------------------------------------------------------- 1 | module.exports = { 2 | plugins: [require('@trivago/prettier-plugin-sort-imports')], 3 | singleQuote: true, 4 | jsxSingleQuote: true, 5 | semi: false, 6 | useTabs: false, 7 | tabWidth: 2, 8 | bracketSpacing: true, 9 | bracketSameLine: false, 10 | arrowParens: 'always', 11 | trailingComma: 'none', 12 | importOrder: ['^node:.*', '', '^@/', '^[./]'], 13 | importOrderSeparation: true, 14 | importOrderSortSpecifiers: true, 15 | importOrderGroupNamespaceSpecifiers: true 16 | } 17 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/10m-fund-ai-gold-math.md: -------------------------------------------------------------------------------- 1 | # $10M fund for AI to a gold in math 2 | 3 | [The AI-MO Prize is a new $10 million challenge fund](https://aimoprize.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=10m-fund-for-ai-to-a-gold-in-math) launched by XTX Markets to accelerate the development of AI models that can reason mathematically and solve problems at a level comparable to high-performing humans. 4 | 5 | ## What’s going on here? 6 | 7 | The $10M AIMO fund puts up a challenge to get AI to do math (and get a gold medal). 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/d6a52bb0-22a7-4d6e-b1a9-3bbc119c7534/image.png?t=1701172894) 10 | 11 | ## What does this mean? 12 | 13 | AIMO fund will award progress prizes for models that achieve milestones, plus a $5 million grand prize for the first publicly shared model that wins a gold medal equivalent in an approved AI math competition. 14 | 15 | The first approved competitions, with progress presentations at the 2024 International Math Olympiad (IMO), will help compare different AI problem-solving strategies in a way accessible to the general public. 16 | 17 | As per Fields medalist and IMO winner Terence Tao, people would be eager to know if AI can match the world’s brightest young mathematical minds. You bet we would. 18 | 19 | ## Why should I care? 20 | 21 | Reasoning and planning are two major drawbacks of models like GPTs (autoregressive LLMs). IMO problems require both at a great level to solve them (there’s an average of 3 hours of time to solve a single question). 22 | 23 | The AI-MO Prize connects AI progress to this benchmark of mathematical excellence. IMO gold medalists become leaders in science, engineering and technology, so an AI matching that achievement could profoundly impact these fields. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/3x-boost-open-ais-valuation.md: -------------------------------------------------------------------------------- 1 | # 3X boost for Open AI's valuation 2 | 3 | Investment firm Thrive Capital is leading a deal to buy shares from OpenAI employees. The deal would boost [OpenAI's valuation to at least $80 billion](https://www.theinformation.com/articles/thrive-capital-to-lead-purchase-of-openai-employee-shares-at-80-billion-plus-valuation?utm_source=bensbites\&utm_medium=referral\&utm_campaign=3x-boost-for-open-ai-s-valuation), more than triple its valuation just 6 months ago. 4 | 5 | ## What’s going on here? 6 | 7 | Thrive Capital is leading the share buyback for OpenAI’s latest round. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/3082709a-1799-436f-99b2-63a7874b3714/image.png) 10 | 11 | ## What does this mean? 12 | 13 | After making $28 million in revenue last year, OpenAI’s annualized revenue now tops $1.3 billion. Still, this new investment in OpenAI values the startup at over 60 times its annualized revenue. Thrive Capital also invested in OpenAI at its $27 billion valuation in April and now the latest deal will boost OpenAI's valuation over 3 times in just 6 months. 14 | 15 | This enormous new valuation suggests investors expect OpenAI will continue raising huge new funding rounds. Rumours are that Sam Altman wants to raise $100 billion to develop advanced AI. While some investors are wary of overpaying, others are eager to get a piece of the red-hot AI startup. 16 | 17 | ## Why should I care? 18 | 19 | OpenAI's exploding valuation reveals just how much promise investors see in artificial intelligence. ChatGPT's viral success proves consumer appetite for AI is real. I wonder how long OpenAI is considered a “startup”. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-actions-zapier-bridge-ai-platform-5000-apps.md: -------------------------------------------------------------------------------- 1 | # AI Actions by Zapier - The bridge between your AI platform and 5000+ apps. 2 | 3 | [Zapier introduces new AI Actions feature](https://actions.zapier.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-actions-by-zapier-the-bridge-between-your-ai-platform-and-5000-apps), which allows AI platforms to use Zapier's existing integrations and actions. AI Actions enables AI systems to execute workflows and automations through Zapier's platform without needing to build custom integrations. 4 | 5 | ## What's going on here? 6 | 7 | Zapier launches AI Actions - A tool for builders to run any Zapier action from AI platforms. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/50563b83-2f50-40bf-ab7a-cc3629e284a1/image.png) 10 | 11 | ## What does this mean? 12 | 13 | AI Actions gives AI platforms access to Zapier's catalog of 20,000+ actions like sending Slack messages, creating HubSpot leads, or drafting Gmail replies. This removes the typical hurdles of third-party authentication and API integration. Now, an AI system's natural language command can act as the trigger, and Zapier's actions handle the execution. 14 | 15 | AI Actions supports major AI platforms like ChatGPT and GPT models out of the box. It can also be used natively within Zapier to enable automations there. Finally, developers can build custom private integrations using the AI Actions REST API and API keys. 16 | 17 | ## Why should I care? 18 | 19 | With AI Actions, you can quickly automate workflows leveraging existing Zapier actions. No longer needing to build custom APIs saves significant development time and resources. 20 | 21 | And since you can use natural language, the automations can be done by the non-technical team members as well, freeing up the engineers for technically hard problems. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-changing-marketing-jobs.md: -------------------------------------------------------------------------------- 1 | # AI is changing marketing jobs, or is it? 2 | 3 | [AI is changing marketing jobs, or is it?](https://www.wsj.com/articles/an-anticipated-wave-of-ai-specialist-jobs-has-yet-to-arrive-01a8169c?mod=followamazon\&utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-is-changing-marketing-jobs-or-is-it)There's been lots of hype about how AI will revolutionize marketing, but so far big brands haven't created many new roles dedicated to overseeing this transformation. 4 | 5 | ## What's going on here? 6 | 7 | Despite predictions that AI will fundamentally alter marketing, major consumer brands have largely not hired people for AI-specific leadership positions or changed their organizational structures. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f6d810e0-fab3-4a71-aad4-05121fdbd99f/image.png?t=1703246743) 10 | 11 | via WSJ 12 | 13 | ## What does this mean? 14 | 15 | Companies seem to be moving slowly and deliberately, rather than rushing to appoint "AI evangelists". They want to avoid past mistakes like overhyping technology like NFTs. Legal and cultural challenges around using AI in a responsible, on-brand way have also slowed adoption. 16 | 17 | The outlier (or visionary) is Coco-Cola here. It has promoted two executives to the newly created roles of global head of generative AI and global head of marketing AI. 18 | 19 | ## Why should I care? 20 | 21 | Data from Indeed says that marketing jobs with AI requirements were 8% lower in November than last year same time. And from Linkedin, the data shows that total views for marketing job listings that mentioned AI grew 14% faster than views for marketing jobs that didn’t. 22 | 23 | So, another way to look at it is that AI will become a default part of these jobs and skill sets without needing to be a special role. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-chips-sealing-deal-flow-vcs.md: -------------------------------------------------------------------------------- 1 | # AI chips are sealing the deal flow for VCs 2 | 3 | [Venture capital firms are leveraging access to hardware](https://www.theinformation.com/articles/vcs-dangle-ai-chips-to-woo-founders?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-chips-are-sealing-the-deal-flow-for-vcs) and their relationships with major cloud providers to help AI startups gain access to GPUs needed for developing AI models. 4 | 5 | ## What's going on here? 6 | 7 | Venture firms are using their connections at Amazon, Microsoft, Google, and others to get AI startups preferential access to limited GPU chips. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/984b1d15-84d5-44e0-86dc-c3c2e9d24309/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Firms like Greylock, Madrona, and Bessemer have been instrumental in helping their portfolio companies like Adept, Inflection, and Laredo Labs land deals for thousands of Nvidia GPUs from cloud providers. This allows startups developing AI models to bypass waits and secure the specialized hardware they need. 14 | 15 | ## Why should I care? 16 | 17 | The VC relationships are providing a competitive advantage to their portfolio companies in the race to build cutting-edge AI. Given the current supply constraints on chips, startups not backed by top-tier VC firms can face debilitating delays in getting access. But with VC assistance, they can get prioritized and accelerate development. 18 | 19 | It shows how influential VCs can shortcut logjams using their connections. For any startup needing significant compute power, allying with a top VC may determine whether they sink or swim right now. 20 | 21 | *Related read:* *[Lamini ditches Nvidia in favour of AMD chips.](https://bensbites.beehiiv.com/p/lamini-ditches-nvidia-favour-amd)* 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-faqs-need-know.md: -------------------------------------------------------------------------------- 1 | # AI FAQs you need to know 2 | 3 | Ethan Mollick, a professor at Wharton, has [answered many AI-related questions](https://www.oneusefulthing.org/p/what-people-ask-me-most-also-some?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-faqs-you-need-to-know) people ask him. We’re borrowing the wits of our AI master Ethan for a quick summary of these FAQs. 4 | 5 | ## Do AI writing detectors work? 6 | 7 | No, they don’t. They fail most of the time, and this failure hurts people with English as a second language more. 8 | 9 | ## Images are easy, right? 10 | 11 | Not quite so! Twisted pictures and creepy features vanished a year ago, and now hands, eyes, skin tones etc. can also fool the best eyes out there. 12 | 13 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/7d4db010-13f5-4fb4-adac-aba152942e7e/image.png) 14 | 15 | ## How to use AI in my work? 16 | 17 | Try it. The magic box is good at some things, and bad at others. Find out how it performs on different parts of your work. 18 | 19 | ## What about my data privacy? 20 | 21 | Disable the training on your data for ChatGPT and others. If you're a big wig, use the enterprise options for better privacy control. 22 | 23 | ## Will I get sued with a copyright claim? 24 | 25 | The laws are still unclear. Though companies like Microsoft and Adobe promise to take the legal headaches if you use their tools. 26 | 27 | ## What’s the future? 28 | 29 | It’s hard to tell. Current tools are constantly changing because of human feedback and other factors. The latest addition of voice and image with LLMs is the next thing to look out for. 30 | 31 | ps: go read the full post by sensei for more nuance. 32 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-hiring-changing.md: -------------------------------------------------------------------------------- 1 | # AI hiring is changing 2 | 3 | [There’s a shift in AI hiring](https://www.signalfire.com/blog/ai-vy-league?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-hiring-is-changing)—from Ivy League degrees to real-world experience at top AI startups. Top AI talent is now found in startups like OpenAI, Cohere etc., not just Harvard, Stanford PhDs or ex-Googlers. 4 | 5 | ## What’s going on here? 6 | 7 | SignalFire's data shows a new source of AI talent emerging, dubbed AI-vy league. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/fffd938f-9c43-498c-9a0e-acd8774f2e4d/image.png?t=1707128216) 10 | 11 | via SignalFire 12 | 13 | ## What does this mean? 14 | 15 | The report says that since 2015 there’s been a drop in the importance of traditional degrees for AI roles. The same trend follows for people who went to top schools like Ivy-leagues or big tech companies like Meta or Google. 16 | 17 | But then what type of talent is being hired, and where do they come from? Turns out the focus is on practical AI skills and achievements, not just academic credentials. And this top talent is concentrated in the “AI-vy League" i.e. 9 AI startups. These are: 18 | 19 | Open AI, Anthropic, Inflection AI, Mosaic ML (acquired by DataBricks), Cohere and AI21 laps, Stability AI, Midjourney, and Hugging Face. (in no order) 20 | 21 | Though the de-credentialization is in part due to necessity—the tech world needs so much AI talent that demand may outstrip supply from top universities. 22 | 23 | ## Why should I care? 24 | 25 | Big tech companies are now targeting these startups for recruitment, indicating a shift in what's valued in AI talent. This trend democratizes the field, making it more about ability and less about educational background. Useful signals if you’re hiring or want a job in the AI space. 26 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-improves-breast-cancer-detection.md: -------------------------------------------------------------------------------- 1 | # AI improves breast cancer detection 2 | 3 | New research shows[AI can improve breast cancer detection rates](https://www.ft.com/content/d3b462a3-8085-43d7-9e37-a78564907d2c?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-improves-breast-cancer-detection) compared to human radiologists alone. 4 | 5 | ## What's going on here? 6 | 7 | AI analysis caught 13% more breast cancers than radiologists in a study of 25,000 mammograms, proving it's an effective safety net. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/66dcebc0-4778-4b6c-a599-4de546658b65/image.png?t=1700216855) 10 | 11 | ## What does that mean? 12 | 13 | The AI tools called Mia spotted cancers radiologists missed, demonstrating it complements human expertise. The research let radiologists use the tool in three different ways, resulting in 5%, 10% and 13% more detections. The additional flags identified invasive cancers that are likely to spread, i.e. it has life-saving potential. 14 | 15 | AI addition also made the diagnosis process 45% faster. Mia is being used at 16 hospitals in the UK and is being rolled out in the US. 16 | 17 | ## Why should I care? 18 | 19 | Missed breast cancers cause avoidable death and suffering. AI offers a solution to improve detection worldwide. You should welcome new tech if it catches disease sooner when treatment is more successful. 20 | 21 | With faster diagnosis, doctors can also cater to a larger population. Using AI as an additional tool with the radiologists in the loop can give patients reassurance. Still, more work can be done to show how AI reaches decisions before fully trusting it. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ai-will-change-courts-work.md: -------------------------------------------------------------------------------- 1 | # AI will change how courts work. 2 | 3 | [AI is set to change legal work](https://www.bloomberg.com/news/articles/2023-12-31/ai-and-the-supreme-court-justice-roberts-says-it-will-transform-judges-work?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ai-will-change-how-courts-work), but human judges will still be needed for the nuance of court cases. Well, that’s what Chief Justice of the US Supreme Court John Roberts thinks. 4 | 5 | ## What’s going on here? 6 | 7 | The US Chief Justice talked about how artificial intelligence will impact the US court system. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5aabc918-9387-4cbb-8011-468b5bed83f9/image.png?t=1704279755) 10 | 11 | ## What does that mean? 12 | 13 | In his year-end report, Roberts said AI is gonna majorly change how judges do their jobs and understand the role of AI in the cases they get. This comes amid the hype about AI that's already starting to change legal practices. 14 | 15 | Roberts thinks AI can't fully replace important people in court because humans are needed to interpret body language and make good guesses. People still trust human judges more than machines. But he predicts human judges will be around "for a while" longer. 16 | 17 | ## Why should I care? 18 | 19 | The upshots are that AI will become vital for legal research and increase access to information, but risks invading privacy and making law less human. As examples like AI-generated text show, the technology often messes up legal briefs by citing fake cases or getting facts wrong. 20 | 21 | So while AI has the potential to help the legal system, its limitations mean human judges are still essential. The transition period will be important as the courts decide how to adopt AI tools (or not). 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/airbnb-acquires-ai-company.md: -------------------------------------------------------------------------------- 1 | # AirBnb acquires an AI company 2 | 3 | [Airbnb has acquired an AI startup called Gameplanner AI](https://news.airbnb.com/airbnb-has-acquired-gameplanner-ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=airbnb-acquires-an-ai-company) for just under $200 million in its first acquisition as a public company. 4 | 5 | ## What’s going on here? 6 | 7 | Airbnb's acquisition of Gameplanner AI signals the company's increased focus on leveraging AI to enhance the user experience. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/82df0e7a-364f-423b-b467-0c36ca086baf/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Gameplanner AI was founded in 2020 by Adam Cheyer, a Siri co-founder, and has been operating in stealth mode. The deal valued the startup at close to $200 million. The acquisition will accelerate some of Airbnb's AI projects, likely around using AI as a personalized travel concierge that learns users' preferences. 14 | 15 | This acquisition aligns with CEO Brian Chesky's vision to incorporate AI that acts as a virtual assistant enhancing Airbnb experiences. The deal is Airbnb's first acquisition since 2019 and may generate interest on Wall Street regarding potential shifts in the company's M\&A strategy. 16 | 17 | ## Why should I care? 18 | 19 | Generative AI is creeping into (for good) every big company. Chesky is extremely bullish on the future of the tech, saying AI could better match customers with their perfect rentals. AirBnB recently launched an AI-powered photo tour tool that helps hosts sort their listing images. Expect more such tools to come to the platform soon. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/alexa-gets-ai-skills-characters-music-20questions.md: -------------------------------------------------------------------------------- 1 | # Alexa gets AI skills: characters, music and 20-questions. 2 | 3 | [Amazon announced new generative AI capabilities for Alexa](https://techcrunch.com/2024/01/09/amazons-alexa-gets-new-generative-ai-powered-experiences/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=alexa-gets-ai-skills-characters-music-and-20-questions) at CES 2023. Three developers - Character.AI, Splash, and Volley - introduced Alexa skills powered by large language models. 4 | 5 | ## What’s going on here? 6 | 7 | Amazon is partnering with AI startups to make their capabilities accessible by Alexa. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/42ebfb14-3c2f-427e-99f3-675de288cd71/image.png?t=1704969476) 10 | 11 | ## What does that mean? 12 | 13 | The integration for these AI tools via Alexa let you use their primary feature by chatting with Alexa. For example, Character.AI's skill lets you chat with different fictional personas like a librarian or personal trainer. 14 | 15 | Similarly, Splash's skill generates custom songs when you sing or rap lyrics. Volley’s integration creates a modern "20 Questions" game, with Alexa asking intelligent questions based on your answers. 16 | 17 | ## Why should I care? 18 | 19 | Amazon’s been putting LLMs in Alexa. They recently added an LLM in the backend for Alexa to make it sound less robotic. With this new release, they are taking the approach of partnership. 20 | 21 | There are many LLM in your hand startups that are getting attention these days, but for home assistants and bigger entertainment devices, Google and Amazon are still the major players. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/alexa-getting-generative-ai-boost-finally.md: -------------------------------------------------------------------------------- 1 | # Alexa is getting generative AI boost, finally. 2 | 3 | Amazon announced new [generative AI capabilities for Alexa](https://techcrunch.com/2023/09/20/amazon-brings-generative-ai-to-alexa?utm_source=bensbites\&utm_medium=referral\&utm_campaign=alexa-is-getting-generative-ai-boost-finally) that will power more conversational and personalized experiences. The new generative AI model will allow Alexa to have **more dynamic conversations**that respond to cues like **body language and vocal tone.** 4 | 5 | ## What's going on here? 6 | 7 | Amazon is bringing more advanced AI to Alexa to enable more natural conversations. 8 | 9 | ## What does this mean? 10 | 11 | Alexa will be able to tap into APIs to control smart home devices based on abstract descriptions. It will also exhibit a more robust personality informed by user preferences and history. For example, it may recommend recipes based on past grocery purchases. The AI can also dynamically adjust its tone to express emotions like excitement and laughter. This should lead to richer, more free-flowing conversations compared to Alexa's more rigid responses previously. 12 | 13 | ## Why should I care? 14 | 15 | This update brings Alexa closer to Amazon’s vision of an intelligent assistant you can chat with conversationally. The adaptability and expanded memory will make interactions feel more natural. This could be a major step toward mainstream adoption of voice assistants beyond simple commands. Of course, the actual experience remains to be seen, but it may finally live up to the promise of an AI assistant that knows you and can hold contextual conversations. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/always-online-models-perplexity.md: -------------------------------------------------------------------------------- 1 | # Always online models from Perplexity 2 | 3 | [Perplexity has launched two new AI models](https://blog.perplexity.ai/blog/introducing-pplx-online-llms?utm_source=bensbites\&utm_medium=referral\&utm_campaign=always-online-models-from-perplexity) - pplx-7b-online and pplx-70b-online. These publicly available models will provide helpful, accurate and up-to-date responses, addressing key limitations around freshness and hallucinations. 4 | 5 | ## What’s going on here? 6 | 7 | Perplexity’s new models access timely information from the internet to give useful, factual responses - things offline models struggle with. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/3c83fffc-1b36-4de3-b37b-33a186728fa6/image.png?t=1701341033) 10 | 11 | ## What does this mean? 12 | 13 | The models build on existing open-source models like Mistral-7B and Llama-70B base models. Then these models are layered (fine-tuning) with current web information to answer questions needing recent data better than offline models. The goal fo these models is to give helpful, non-hallucinated responses. 14 | 15 | Perplexity evaluated the models for freshness, accuracy and usefulness and they think the models are well-suited for real-world uses like answering consumer questions where up-to-date info is key. As part of the launch, Perplexity has opened up access to the models through its public API and Perplexity Labs playground. 16 | 17 | ## Why should I care? 18 | 19 | Consumer products like ChatGPT are adding search to the products. Search results reduce hallucination and people care about real-time info. 20 | 21 | But figuring out the search when building with most models is tiresome. PPLX’s new online models API come with their SEO expertise built in for you. No need for that extra search API. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/amazon-bedrock-available-now.md: -------------------------------------------------------------------------------- 1 | # Amazon Bedrock is available now. 2 | 3 | [AWS has announced new AI products](https://www.businesswire.com/news/home/20230928577730/en/AWS-Announces-Powerful-New-Offerings-to-Accelerate-Generative-AI-Innovation?utm_source=bensbites\&utm_medium=referral\&utm_campaign=amazon-bedrock-is-available-now), expanding its AI catalogue to help companies adopt generative AI. **Amazon Bedrock, its fully managed service** for building on top of foundation models is now **generally available.** Multiple AWS products are getting additional AI-enabled features. 4 | 5 | ## What's going on here? 6 | 7 | AWS launched Amazon Bedrock, a fully managed AI service giving access to top foundation models, plus other tools to build customized AI apps. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/e4bc16b8-2fcb-48db-8615-bac70fa432b8/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Amazon Bedrock lets companies easily experiment with leading AI models like Anthropic's Claude and Meta’s Llama 2 through a single API. It simplifies development so companies can focus on creating differentiated AI solutions. Bedrock also helps companies customize models with their own data privately and securely. Amazon Titan Embeddings model is a new addition to their suite of home grown models. Amazon CodeWhisperer is getting an enterprise tier with code auto-complete using companies' private code. Amazon Quicksight is evolving into a generative BI platform with natural language data queries. 14 | 15 | ## Why should I care? 16 | 17 | Bedrock’s goal is to save Amazon’s position as the top cloud provider. With it Amazon’s trying to cement AWS as a one-stop AI shop. With their own coding and embedding models, they might have an edge over Microsoft which is offering OpenAI models on Azure and paying a hefty fee. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/amazon-q-preview-assistant-work-business.md: -------------------------------------------------------------------------------- 1 | # Amazon Q is in preview - The assistant for work and business. 2 | 3 | [Amazon has launched Amazon Q,](https://aws.amazon.com/about-aws/whats-new/2023/11/aws-amazon-q-preview/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=amazon-q-is-in-preview-the-assistant-for-work-and-business) an AI assistant for business use cases like customer service, knowledge sharing, and app development. 4 | 5 | ## What's going on here? 6 | 7 | Amazon Q will help employees by answering questions, summarizing reports, writing content, and automating tasks using a company's connected data and documents. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/193eb85a-83e6-4fb8-9dff-c98a6feabbe8/image.png?t=1701264316) 10 | 11 | ## What does this mean? 12 | 13 | It has 40+ built-in enterprise software connectors and only accesses files each user can view. It cites internal sources used to generate each response. 14 | 15 | There are three main use cases: 16 | 17 | - First, it's an enterprise knowledge assistant. Employees can ask questions and get authoritative answers based on internal data. 18 | 19 | - Second, it assists developers by providing expert guidance on building, deploying and managing AWS applications. 20 | 21 | - Third, it's integrated into Amazon Connect, helping customer service agents resolve tickets faster using real-time recommendations. 22 | 23 | ## Why should I care? 24 | 25 | This matters because having an AI customized to a company's data helps employees work smarter. It enables faster answers, insights and decisions. Over time, this can increase productivity, improve customer service, empower workers, reduce costs, and streamline operations. 26 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/amazon-will-use-ai-robots-speed-delivery.md: -------------------------------------------------------------------------------- 1 | # Amazon will use more AI and robots to speed up delivery. 2 | 3 | [Amazon is rolling out a major overhaul of its fulfilment centres](https://www.wsj.com/tech/amazon-introducing-warehouse-overhaul-with-robotics-to-speed-deliveries-40e3e65?utm_source=bensbites\&utm_medium=referral\&utm_campaign=amazon-will-use-more-ai-and-robots-to-speed-up-delivery) using new AI and robotics to speed up deliveries. Amazon says that the new technology is designed to work alongside human employees to reduce injuries. Nice, but what about lunch and pee time mate? 👉👈 4 | 5 | ## What's going on here? 6 | 7 | Amazon's new warehouse system will significantly cut delivery times and make inventory tracking much faster. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/b1a2cb55-d9a7-40f3-a930-333ded2c6cca/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Amazon's revamp introduces robots and AI into its warehouses to reduce the time it takes to process orders. The centrepiece is a robotic arm called Sparrow and a new sortation system named Sequoia. Together, these will slash the time to fulfil orders by up to 25% while identifying inventory 75% faster. Amazon’s plans for Sequoia include the “same-day delivery sites” it’s working on. 14 | 15 | ## Why should I care? 16 | 17 | The claim from Amazon is that automation is not about eliminating jobs but rather mundane tasks. The goal is to integrate robots seamlessly into workflows. Though, I believe the pattern of people who can’t adapt to these newer workflows (which, I agree, is easier said than done) will lose their jobs. 18 | 19 | Rivals like Walmart are also turning warehouse jobs into robot management roles. Amazon would also start to test a bipedal robot named Digit in its operations. We’re not ready for LLMs, now imagine AI + robotics. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/anthropics-responsible-scaling-policy.md: -------------------------------------------------------------------------------- 1 | # Anthropic's Responsible Scaling Policy 2 | 3 | Anthropic published its [Responsible Scaling Policy](https://www.anthropic.com/index/anthropics-responsible-scaling-policy?utm_source=bensbites\&utm_medium=referral\&utm_campaign=anthropic-s-responsible-scaling-policy) to manage risks as AI becomes more capable. It includes a series of technical and organizational protocols they are adopting to help us manage the risks of developing increasingly capable AI systems. 4 | 5 | ## What's going on here? 6 | 7 | Anthropic introduced a framework called AI Safety Levels to target catastrophic risks from misuse as AI advances. 8 | 9 | ## What does this mean? 10 | 11 | The policy defines safety requirements at higher AI Safety Levels. Current models like Claude are Level 2, requiring security measures. Future Level 3 models may show dangerous capabilities, so Anthropic won't deploy them until safety is proven through red team testing. This balances safety and beneficial progress. 12 | 13 | ## Why should I care? 14 | 15 | This policy matters because advanced AI holds risks if misused. Anthropic's framework provides a model for other companies and policymakers looking to develop AI safely. While not perfect, it explains the risk of AI models without going all doom and gloom. We need more companies thoughtfully addressing safety. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/apples-entering-ai-hardware.md: -------------------------------------------------------------------------------- 1 | # Apple's entering into AI with hardware. 2 | 3 | Apple has released new Macs powered by M3 chips, showing they are[embracing AI through custom hardware.](https://www.youtube.com/watch?v=ctkW3V0Mh-k\&utm_source=bensbites\&utm_medium=referral\&utm_campaign=apple-s-entering-into-ai-with-hardware) The M3 Max chip targets AI/ML developers with its powerful GPU stack. 4 | 5 | ## What’s going on here? 6 | 7 | Apple’s news M3 chips have new GPUs optimized for AI, up to 2.5x faster than M1. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5c9d797f-92b1-499d-a36e-fddd88adb861/image.png) 10 | 11 | ## What does that mean? 12 | 13 | The new M3 chips have a new GPU architecture with a feature Apple calls dynamic caching. Dynamic caching allocates GPU memory adaptively in real-time, increasing the utilization of the GPU. The new neural engine is 60% faster, speeding up ML. 14 | 15 | The M3 Max has a 16-core CPU and 40-core GPU supporting 128GB of unified memory. Apple claims that’ll allow developers to work with models with billions of parameters. 16 | 17 | Apple is keeping pace with AI-focused offerings from rivals like Qualcomm. Qualcomm made a similar claim this month with their Snapdragon X Elite of being able to run a 13B model on-device. 18 | 19 | ## Why should I care? 20 | 21 | If you use graphics/ML apps, the new Macs will bring meaningful speed improvements. If you’re a developer, Apple's platform is now more compelling for generative AI projects needing local compute. 22 | 23 | And there’s the obvious answer to why should you care: because it’s APPLE. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/arc-max-blends-ai-browser.md: -------------------------------------------------------------------------------- 1 | # Arc Max blends AI in your browser. 2 | 3 | [Max, an AI feature within the Arc browser](https://arc.net/max?utm_source=bensbites\&utm_medium=referral\&utm_campaign=arc-max-blends-ai-in-your-browser) adds various productivity and navigational features. For context: the Arc browser is themed around customizing the browser to your needs. 4 | 5 | # What's going on here? 6 | 7 | Max helps you navigate the web faster with AI features within the Arc browser. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/70a2503a-aec0-41bb-ac26-93b216c4e151/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Max provides several features within the Arc browser like answering questions using information on the current page, previewing summaries of links without opening them, getting answers from ChatGPT, renaming tabs and downloads, and organizing files. Max is available to existing customers for [free](https://twitter.com/heyecs/status/1709238608148795840?utm_source=bensbites\&utm_medium=referral\&utm_campaign=arc-max-blends-ai-in-your-browser) now. 14 | 15 | ## Why should I care? 16 | 17 | These AI-powered features can save you time by reducing clicks, renaming and organizing your files and tabs more intelligently, and providing relevant information directly within your browser. The Max features act like a personal assistant that makes you more productive online. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/arrakis-deserted.md: -------------------------------------------------------------------------------- 1 | # Arrakis is deserted 2 | 3 | [OpenAI recently scrapped a new AI model called Arrakis](https://www.theinformation.com/articles/openai-dropped-work-on-new-arrakis-ai-model-in-rare-setback?utm_source=bensbites\&utm_medium=referral\&utm_campaign=arrakis-is-deserted) after it failed to meet expectations. Arrakis was meant to allow OpenAI to run its AI systems more efficiently and cheaply. 4 | 5 | ## What's going on here? 6 | 7 | OpenAI halted work on a model codenamed Arrakis this spring after realizing the model underperformed and did not deliver the expected cost savings. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/84b55950-29fd-4777-8de1-32c0f9578140/image.png) 10 | 11 | ## What does this mean? 12 | 13 | While not immediately impacting OpenAI's business, it may slow future progress as engineers shift focus. Microsoft, an OpenAI partner, was also hoping to use Arrakis to lower costs, so this delays their integration plans. Open AI researchers have since pivoted to making GPT-4 faster with hopes of a model termed GPT-4 Turbo (which could’ve been Arrakis’ final name, if continued). 14 | 15 | Arrakis was meant to use a technique called sparsity. Google’s Jeff Dean has also referred to sparsity as an important trend looking ahead. 16 | 17 | ## Why should I care? 18 | 19 | The Arrakis news highlights the complexity of pushing AI forward. There are no guarantees, even for the most capable companies. The delay caused by the Arrakis project could mean that other companies like Google and Anthropic might catch up to Open AI soon. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/artists-can-poison-ai-models.md: -------------------------------------------------------------------------------- 1 | # Artists can poison AI models 2 | 3 | A new tool called [Nightshade allows artists to add invisible changes](https://www.technologyreview.com/2023/10/23/1082189/data-poisoning-artists-fight-generative-ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=artists-can-poison-ai-models) to the pixels in their art before posting it online. This "poisons" the images so they damage AI models if scraped for training data without permission. 4 | 5 | ## What's going on here? 6 | 7 | Artists can now make their work AI-hostile to prevent AI companies from using it without consent. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/84da050d-b94e-4de3-afec-54ed736b9e4f/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Nightshade exploits vulnerabilities in AI image models trained on vast scraped data. It alters uploaded art (without changing the visuals) to make models learn incorrect associations. For example, poisoned dog images could get models to output cats instead. This forces companies to find and remove every corrupted sample, which is challenging. 14 | 15 | Poison spreads too - altering "fantasy art" also affects related concepts like "dragons". Larger models need more poisoned data for major damage. The researchers behind Nightshade claim it tips the power balance back towards artists. 16 | 17 | ## Why should I care? 18 | 19 | From one POV, Nightshade might pressure AI companies to respect artist rights and pay royalties. Artists can now confidently share work publicly again. 20 | 21 | But somewhere at the back of my mind, this also feels like adding spikes to the road, so cars can’t drive. Bad data samples can decrease the model performance significantly. 22 | 23 | But the implications are clear - data models can no longer rely on good old scraping to get data. If they do, they are shooting themselves in the foot. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/aya-ai-languages.md: -------------------------------------------------------------------------------- 1 | # Aya - AI for all languages. 2 | 3 | [Cohere For AI just dropped a major multilingual AI bomb.](https://txt.cohere.com/aya/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=aya-ai-for-all-languages) LLMs are hot right now. But they often leave out a huge chunk of the world due to language barriers. Gotta think beyond English. Well, Cohere just did with [Aya](https://cohere.com/research/aya?utm_source=bensbites\&utm_medium=referral\&utm_campaign=aya-ai-for-all-languages). 4 | 5 | ## What's going on here? 6 | 7 | Aya is a multilingual LLM that supports a whopping 101 languages. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ea480973-ec48-473b-9aa3-ba2168381c73/image.png?t=1707906753) 10 | 11 | ## What does this mean? 12 | 13 | Most AI models get stuck on English and a few others, leaving a huge chunk of the globe out of the loop. Cohere for AI partnered with 3,000 researchers across the globe to create Aya. 14 | 15 | Aya was trained on the largest multilingual dataset (513M prompts!) and outperforms other open-source multilingual models in tricky tasks like understanding, summarizing, and translating various languages. It also covers 50+ previously underserved languages; hello Somali and Uzbek! 16 | 17 | The project is open-sourcing both the Aya model, as well as the multilingual instruction fine-tuned dataset. 18 | 19 | ## Why should I care? 20 | 21 | More multilingual AI powers up nuanced conversations across cultures, and build tools that work for everyone, not just some of us. Especially when most of us are working with global teams now. Okay, forget teams. You can finally explain your tech job to family (no promises though). 22 | 23 | And, the Aya dataset is a treasure for multilingual research. You can contribute to the Aya project or use it to do some stuff on your own. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/bard-song-googles-slow-culture.md: -------------------------------------------------------------------------------- 1 | # A Bard song for Google's slow culture 2 | 3 | Google is urgently working on its AI chatbot Bard to compete with ChatGPT. This is [shaking up Google's historically slow culture.](https://www.theinformation.com/articles/googles-wartime-urgency-to-chase-chatgpt-shakes-up-culture?utm_source=bensbites\&utm_medium=referral\&utm_campaign=a-bard-song-for-google-s-slow-culture) 4 | 5 | ## What's going on here? 6 | 7 | Google has adopted a "wartime" mentality to expedite Bard, its ChatGPT competitor. Teams that work on Bard are moving much faster than other Google teams. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c0502c97-0127-4362-b319-b8aa65ec37fc/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The Bard team shortcuts legal reviews and executes projects in days or weeks rather than months. This startles some employees but energizes others seeking to make their mark on cutting-edge AI. Bard's leader Sissie Hsiao conveys urgency, asking about work-life balance and suggesting those concerned should transfer off her team. 14 | 15 | Hsiao runs Bard like a lean startup using "pods" - small groups working in parallel on different features. She and other execs closely monitor pods in regular reviews. Hsiao is a heavyweight at Google known for monetizing mobile apps. So she likely aims to eventually bring profit from Bard. 16 | 17 | ## Why should I care? 18 | 19 | Google is laser-focused on not falling behind in AI. This intense push suggests chatbots like Bard are viewed as existentially important to Google's future. The company is uncharacteristically willing to ruffle feathers to build cutting-edge AI quickly. 20 | 21 | Google is marshalling immense resources behind Bard. With its best talent on the case, Bard could evolve from joke to juggernaut practically overnight. And let’s not forget it’s still free. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/bay-bridge-sf-compute-train-llms-cheap.md: -------------------------------------------------------------------------------- 1 | # Bay Bridge by SF Compute - Train your LLMs for cheap. 2 | 3 | The San Francisco Compute Company is selling the [cheapest access ever to beastly H100 training clusters.](https://sfcompute.com/blog/worlds-cheapest-supercomputer?utm_source=bensbites\&utm_medium=referral\&utm_campaign=bay-bridge-by-sf-compute-train-your-llms-for-cheap) We're talking bleeding-edge setups with 1024 interconnected H100 GPUs and massive InfiniBand networking. 4 | 5 | ## What's going on here? 6 | 7 | SF Compute is renting AI startups time on an H100 cluster by the month or even just weeks. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f678bcda-8f2e-4ed6-a1d4-8d746ab0a609/image.png) 10 | 11 | ## What does this mean? 12 | 13 | San Francisco Compute is offering short-term bursts on thousands of interconnected H100s with InfiniBand. They have one cluster online now, called Angel Island. The next one, Bay Bridge, comes online in December. 14 | 15 | Other providers force you to buy a whole year upfront, which costs like $25 million for 1024 H100s. SF Compute is planning to sell the same beast rig for just the months you need, like $2.5 million for one month. It's more per hour than what most folks will do, which doesn't always make it a great choice for inference. But for training very large models, it's cheaper since you can say goodbye to multi-year lock-ins. 16 | 17 | ## Why should I care? 18 | 19 | If you're an AI shop wanting to play with the biggest models, you need access to hardware like this. But until now you had to pre-pay tens of millions a year to play. SF Compute also guarantees bare metal grade performance, ML infra support and refund if things go wrong. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/big-announcements-intel-innovation-2023.md: -------------------------------------------------------------------------------- 1 | # Big Announcements At Intel Innovation 2023 2 | 3 | Intel made several big announcements at their [2023 Innovation event](https://www.crn.com/news/components-peripherals/6-big-announcements-at-intel-innovation-2023-from-288-core-cpu-to-ai-supercomputer?utm_source=bensbites\&utm_medium=referral\&utm_campaign=big-announcements-at-intel-innovation-2023) related to AI and computing, including **new CPUs launching in December**, a plan for a 288-core Xeon chip, and a deal to build an **AI supercomputer for Stability AI.** They're aiming to challenge Nvidia in AI with their CPUs, GPUs, and dedicated AI accelerators. 4 | 5 | ## What's going on here? 6 | 7 | While Nvidia has dominated the AI chip space recently with its GPUs, Intel is determined to become a major player. Their strategy includes integrating AI acceleration into CPUs to complement discrete offerings like their Gaudi chips. 8 | 9 | ## What does this mean? 10 | 11 | - Intel’s Core Ultra (Meteor Lake) is launching on December 14th, featuring integrated AI acceleration with a neural processing unit. 12 | 13 | - 5th-gen Xeon Scalable server processors will also launch on Dec. 14. Future generation will include a 288-core CPU (in 2024). 14 | 15 | - Intel will build a large AI supercomputer with Xeon and Gaudi2 processors for Stability AI. 16 | 17 | - Intel Developer Cloud platform is now available, giving access to the latest Intel chips. Project Strata, a commercial software platform for building edge infrastructure will come in 2024. 18 | 19 | ## Why should I care? 20 | 21 | More competition from Intel versus just Nvidia should drive innovation faster. Intel is also offering developers a platform to access and build AI solutions. While Nvidia may still lead, Intel's major investments in AI-focused chips and infrastructure signal more choices for developers. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/brave-browser-adds-ai-chatbot.md: -------------------------------------------------------------------------------- 1 | # Brave browser adds AI chatbot 2 | 3 | The privacy-focused web browser [Brave has launched Leo, a new AI chatbot](https://brave.com/leo-release/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=brave-browser-adds-ai-chatbot) integrated directly into the browser. The Leo chatbot competes with Bing's new ChatGPT integration and Google's Bard AI. 4 | 5 | ## What's going on here? 6 | 7 | Brave is offering an AI chatbot Leo, that aligns with its commitment to user privacy. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/710c7040-87cb-43d3-a9f7-35f25dc10b45/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Leo provides standard AI chatbot features like answering questions and summarizing webpages. However, conversations with Leo are not recorded or used to train AI models. Users also don't need to create an account to use Leo. This contrasts with Bing Chat and Google Bard which connect conversations and search histories to user accounts. 14 | 15 | The base Leo chatbot uses Meta's Llama 2 AI model and is free for all Brave desktop users. An upgraded Leo Premium subscription offers faster response times and the ability to choose Anthropic's Claude AI model instead. Additional premium features include priority access during peak times, higher rate limits, and early access to new capabilities. 16 | 17 | ## Why should I care? 18 | 19 | Brave is promising an AI assistant that doesn't compromise user privacy, unlike the tech giants' offerings. With Leo, you can access AI chatbot features without sharing personal data or linking conversations to your identity. 20 | 21 | More browsers are adding AI assistants to their native experience. Opera has already done it a while back, now Brave is joining in with Leo. Although Bard exists, Chrome still doesn’t have a native AI chatbot. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/bret-taylors-stealth-ai-startup-almost-unicorn.md: -------------------------------------------------------------------------------- 1 | # Bret Taylor’s stealth AI startup is almost an unicorn. 2 | 3 | [Bret Taylor’s AI startup is about to become an unicorn.](https://www.bloomberg.com/news/articles/2024-01-26/bret-taylor-s-ai-startup-to-get-funding-at-near-1-billion-value?utm_source=bensbites\&utm_medium=referral\&utm_campaign=bret-taylor-s-stealth-ai-startup-is-almost-an-unicorn) The ex-CEO at Salesforce and an OpenAI board member, is making moves with his AI firm, Sierra. 4 | 5 | ## What’s going on here? 6 | 7 | Sierra is in talks for new funding valuing it at $1B. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/e2fb5904-33b3-4d68-9a2e-39ad2ea2f637/image.png?t=1706272109) 10 | 11 | via Bloomberg 12 | 13 | ## What does this mean? 14 | 15 | Taylor founded this AI startup with Clay Bavor, a former Google hotshot. Now, it is on the brink of hitting unicorn status with a cool nearly $1 billion valuation, thanks to Sequoia Capital leading a hefty $85 million investment round. 16 | 17 | Benchmark led an early investment in the company last year and Sierra already has some significant traction. With big names like Taylor and Bavor at the helm and Sequoia Capital opening its wallet, Sierra's one to watch. Taylor has already said that he’s with OpenAI for a short while—until the board situation sorts out. He’s got plans. 18 | 19 | ## Why should I care? 20 | 21 | Taylor’s got a knack for leading and innovating—riding Salesforce's ups and Salesforce's downs, and now trying to stabilise OpenAI’s board. Such people can’t sit on the sidelines when AI is changing industries. Together with Bavor, Taylor is bringing that magic to Sierra. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/can-ai-get-patent-latest-guidelines.md: -------------------------------------------------------------------------------- 1 | # Can AI get a patent? \[Latest Guidelines Feb 2024] 2 | 3 | **Can AI get a patent?** That's the question buzzing around tech. The US Patent Office (USPTO) just dropped some [guidelines](https://public-inspection.federalregister.gov/2024-02623.pdf?utm_source=bensbites\&utm_medium=referral\&utm_campaign=can-ai-get-a-patent-latest-guidelines-feb-2024) so we all know where things stand. 4 | 5 | ## What’s going on here? 6 | 7 | The USPTO says only humans can be inventors and get patents. AI might be crazy smart, but it ain't legally a person. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/82fb5dad-9c3f-4324-858f-ded8ed415bfd/image.png?t=1707829164) 10 | 11 | ## What does this mean? 12 | 13 | Basically, it means that humans get the legal credit, with AI being like a super-advanced tool. Even if AI helps out, you need a human inventor on the patent to make it official. 14 | 15 | You can't just tell the AI "make something cool" and claim the patent. You gotta actually contribute to the invention's development. 16 | 17 | The USPTO also says that technology can lead to new laws and it will update standards to match changes if Congress modifies laws on AI. 18 | 19 | ## Why should I care? 20 | 21 | This makes a big difference for all of us building with AI. If you're a founder working on AI-generated inventions, make sure you can prove your role in the process or there's no patent protection. For investors, the current stakes are clear—strong IP depends on the human driving the AI, not just the AI itself. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/character-ai-now-lets-join-group-chat-ai-characters.md: -------------------------------------------------------------------------------- 1 | # Character AI now lets you join a group chat with AI characters 2 | 3 | [CharacterAI announces Character Group Chats](https://blog.character.ai/new-feature-announcement-character-group-chat/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=character-ai-now-lets-you-join-a-group-chat-with-ai-characters) where you can interact with multiple AI characters and humans in the same room. CharacterAI allows users to chat with personality chatbots like Elon Musk to fictional characters like Super Mario. For context, it has the highest DAU/MAU ratio for all consumer AI apps. 4 | 5 | ## What’s going on here? 6 | 7 | CharacterAI launched Group Chats - A new way for users to interact with multiple AI characters together in group chats. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/d749c0bd-74c3-4f05-ab32-3ed1f2a824ef/image.png) 10 | 11 | ## What does this mean? 12 | 13 | CharacterAI+ members can create Group Chats and invite other humans and characters into the group. The current limit is 5 humans and 5 characters per Group Chat. The feature is mobile only for now. The team at CharacterAI feels this is going one step closer to sci-fi interaction with humans. 14 | 15 | ## Why should I care? 16 | 17 | Group Chats enables new possibilities for collaborative storytelling, gameplaying, studying, and more. In comparison to solo chat, you can create language learning groups, a fictional world setting for your novel and a lot more. 18 | 19 | The AI simulation demos are usually discarded as we’re not there yet. But these group chats could be the realistic version that is already here. And there’s little scepticism on adoption based on how engaged CharacteAI users are. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/character-ai-talks-google-next-round.md: -------------------------------------------------------------------------------- 1 | # Character AI in talks with Google for its next round. 2 | 3 | [**Character AI is looking to raise again.**](https://www.reuters.com/technology/google-talks-invest-ai-startup-characterai-sources-2023-11-10/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=character-ai-in-talks-with-google-for-its-next-round) It allows people to create virtual characters and chat with them, which is highly popular among younger audiences. The tea is that Google wants to get a piece for themselves. 4 | 5 | ## What's going on here? 6 | 7 | Google wants to invest many hundred millions into Character.AI 8 | 9 | ## What does this mean? 10 | 11 | The current round is supposed to value Character AI at $5B. It last raised $150M in March 2023 at a $1B valuation with a16z leading the round. 12 | 13 | Character AI was founded by former Google employees Noam Shazeer and Daniel De Freitas. They are already using Google Cloud and TPUs for training their models and this will get Google more skin in the AI characters/personas game. 14 | 15 | ## Why should I care? 16 | 17 | Characters are an AI use case that business AI tools gloss over. Meta is going heavy in that direction with their celebrity-inspired AI avatars and in some form, OpenAI’s newly launched GPTs could also capture some part of that market. Character AI is the leader right now and Google might be trying to use it in their favour. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/china-allows-copyright-ai-images.md: -------------------------------------------------------------------------------- 1 | # China allows copyright for AI images. 2 | 3 | [The Beijing Internet Court said an AI-generated picture qualified as original artwork](https://www.scmp.com/tech/tech-trends/article/3243570/beijing-courts-ruling-ai-generated-content-can-be-covered-copyright-eschews-us-stand-far-reaching?utm_source=bensbites\&utm_medium=referral\&utm_campaign=china-allows-copyright-for-ai-images) for copyright protection thanks to the human creator's input in designing it. To compare, the US rulings have been implying that AI creations cannot be copyrighted. 4 | 5 | ## What’s going on here? 6 | 7 | The Chinese court ruled that AI-generated images are subject to copyright law. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/d1ee1967-ce3c-4170-8822-39ca7b4dee48/image.png?t=1701771783) 10 | 11 | ## What does that mean? 12 | 13 | A blogger shared an AI image created by someone else and got sued. The Court decided that the AI image creator has the copyright for the image because there was an intellectual investment in writing prompts, specifying parameters and processes to match aesthetic choice. 14 | 15 | The blogger is fined 500-Yuan (about $70), though the decision can be appealed. Also, the court says that AI copyright should be discussed case-by-case. 16 | 17 | ## Why should I care? 18 | 19 | A US court disallowed giving users copyright to AI-generated material earlier this year. While many people are focused on the backwards-looking debate of the copyright drama: “Hey, they used my data to train”, I think the more important question is forward-looking: to understand who gets credit for the new images. 20 | 21 | India earlier said that AI is not a person, so it can not get copyright but users might based on differences in cases. This decision from China seems similar. There’s certainly a difference in how different countries are looking at AI. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/cloudflare-partners-hugging-face-make-ai-models-accessible-affordable.md: -------------------------------------------------------------------------------- 1 | # Cloudflare partners with Hugging Face to make AI models more accessible and affordable 2 | 3 | **What's going on here?** 4 | 5 | [Cloudflare is launching Workers AI](https://blog.cloudflare.com/partnering-with-hugging-face-deploying-ai-easier-affordable/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=cloudflare-partners-with-hugging-face-to-make-ai-models-more-accessible-and-affordable)to make AI inference easy and accessible for all developers. 6 | 7 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/07c581a1-c0ec-40dc-a211-ccde50b25935/image3-26.png) 8 | 9 | **What does this mean?** 10 | 11 | Workers AI allows developers to run popular AI models with just a few lines of code, without needing to manage infrastructure. It is powered by Cloudflare's global network of GPUs and provides curated pre-trained models like text generation, translation, image classification and more. 12 | 13 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/17dfa08e-1a3f-48fa-873d-9039dc319017/image1-24.png) 14 | 15 | **Why should I care?** 16 | 17 | This makes it incredibly simple for any developer to start building with AI, even with zero machine-learning expertise. Now you can deploy an AI-powered app in minutes. For those already using other services, Workers AI is platform-agnostic so you can still access it via API. Cloudflare is also prioritizing developer experience, privacy and global reach. The models are private by default, learning only from your prompts. Cloudflare plans to expand to 100 data centres this year for low latency worldwide. It's currently in open beta so developers can try it out and provide feedback to improve the product before launch. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/concentration-ai-jobs-needs-intervention.md: -------------------------------------------------------------------------------- 1 | # Concentration of AI jobs needs intervention 2 | 3 | [Geographical concentration of generative AI jobs isn’t going away on its own.](https://www.brookings.edu/articles/new-data-shows-that-without-intervention-generative-ai-jobs-will-continue-to-cluster-in-the-same-big-tech-hubs/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=concentration-of-ai-jobs-needs-intervention)The analysis comes from research by Brookings and other sources, shedding light on the narrative that AI could democratize tech geographically. 4 | 5 | ## What's going on here? 6 | 7 | Despite the hype that generative AI could democratize tech across various regions, the reality is less rosy. Most AI jobs remain concentrated in a few key metro areas. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/bd2c3914-4666-4916-ae64-5b33de865968/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The stats are pretty eye-opening. Over 60% of generative AI jobs posted in the year ending in July 2023 are clustered in just 10 metro areas. More than a fifth of those are in the Bay Area alone. This mirrors earlier findings that over half of the nation’s AI activity in areas like federal contracting and startups were also focused in a few key regions, such as the Bay Area and 13 "early adopter" metro areas. 14 | 15 | ## Why should I care? 16 | 17 | If you're gunning for a career in AI or planning a startup, this geographic bottleneck matters. It's not just about Silicon Valley hogging the limelight anymore; it's about real opportunities—or the lack thereof—in other places. But it’s not set in stone. Interventions like federal and state programs could still tilt the scales. Knowing where the jobs are (and aren’t) helps you strategize, whether you’re job hunting or policy-making. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/copilot-get-specific-button-keyboards.md: -------------------------------------------------------------------------------- 1 | # Copilot to get specific button on keyboards 2 | 3 | [Microsoft is introducing the first major change to PC keyboards in 30 years](https://www.axios.com/2024/01/04/microsoft-copilot-ai-button-windows?utm_source=bensbites\&utm_medium=referral\&utm_campaign=copilot-to-get-specific-button-on-keyboards) - a dedicated key for its AI assistant, Windows Copilot. Microsoft is confident that AI will be integral to the PC as we go ahead. 4 | 5 | ## What's going on here? 6 | 7 | Microsoft sees AI as the next big shift in how people interact with computers and is baking it into Windows PCs via new hardware like the Copilot key. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/b89fba99-dde9-463b-bf78-a5417e7c4798/image.png?t=1704452656) 10 | 11 | ## What does this mean? 12 | 13 | The Copilot key replaces a menu button and will debut on new PCs announced at CES this week. Microsoft executive VP Yusuf Mehdi expects it will become standard on most new Windows machines by the end of 2023. 14 | 15 | The new button will launch Copilot, allowing the new standard for AI—the simple language chatbox for everything. If you’ve disabled Copilot, it’ll open Windows search. Acc. to Mehdi, the goal is to reimagine the PC experience for the AI era (via Axios). 16 | 17 | ## Why should I care? 18 | 19 | The addition of the Copilot key signals Microsoft believes AI is now mainstream enough for broad adoption. Apple’s M-series chips are currently thought of as the best playground for on-device AI. Although Microsoft is also talking about adding new chips with AI-specific processing power under the hood, changing the user behaviour with specific buttons could offload the disadvantage of cloud AI (if it is one). 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/duolingo-cuts-10-contractors-uses-ai.md: -------------------------------------------------------------------------------- 1 | # Duolingo cuts 10% of contractors as it uses more AI 2 | 3 | [Duolingo is reducing its contractor workforce](https://www.bloomberg.com/news/articles/2024-01-08/duolingo-cuts-10-of-contractors-in-move-to-greater-use-of-ai?utm_source=bensbites\&utm_medium=referral\&utm_campaign=duolingo-cuts-10-of-contractors-as-it-uses-more-ai) by using AI for content. Though Duolingo’s use of AI is already public with their partnership with OpenAI for Duolingo Max. 4 | 5 | ## What's going on here? 6 | 7 | Duolingo laid off 10% of its contractors as it uses more AI for content creation. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c9173ed0-541e-4dcb-808e-1af99efbe270/image.png?t=1704803306) 10 | 11 | ## What does this mean? 12 | 13 | Last week one of the laid-off contractors posted on Reddit about this. A Duolingo spokesperson confirmed it, saying around 10% of contractors got the axe. Duolingo said no full-timers were cut in this move. 14 | 15 | There’s more AI coming. The CEO said generative AI will let Duolingo make new content much faster, like instructional video scripts. The company also uses AI for automated feedback and conversations in the app. 16 | 17 | ## Why should I care? 18 | 19 | The spokesperson said the layoffs weren't straight-up AI replacement. AI is used by Duo staff in their work. They just don’t need that many people now. It’s media speak. It's like saying you're going home—only to change clothes before hitting the club. 20 | 21 | But don’t get me wrong, AI replacing boring work rather than whole jobs is good. The sad truth is some will lose work short-term (and long-term if we don't evolve with the tech). 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/eleven-labs-new-ai-unicorn.md: -------------------------------------------------------------------------------- 1 | # Eleven Labs is the new AI unicorn. 2 | 3 | You've probably heard about ElevenLabs, that cool AI voice tech company. [They just bagged $80 million in Series B funding](https://elevenlabs.io/blog/series-b/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=eleven-labs-is-the-new-ai-unicorn), pushing their total funds to a neat $101 million. And guess what? They're now valued at over a billion dollars! 4 | 5 | ## What’s going on here? 6 | 7 | ElevenLabs is on a roll with funding and new products in the pipeline. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/b291482d-4cd6-4346-b69b-438ec016b6fd/image.png?t=1705927955) 10 | 11 | ## What does this mean? 12 | 13 | The cash influx is led by big names like Andreessen Horowitz and others. Started in 2022 by a couple of tech-savvy guys, Piotr Dabkowski and Mati Staniszewski, Elevan Labs is one of the leading AI voice companies now. From a modest team of 5, they've scaled up to 40 and are eyeing about 100 by year's end. 14 | 15 | They will be rolling out some fancy new tools like a Dubbing Studio for movie buffs, a Voice Library Marketplace for voice creators to earn some dough, and a Mobile Reader App that turns text and URLs into audio in the upcoming weeks. 16 | 17 | ## Why should I care? 18 | 19 | Voice Cloning is a big market and ElevenLabs is a name to watch. Movies, audiobooks, or just browsing the web, their tech could change how you experience media. Plus, if you're in the tech biz or a voice artist, this could open up some exciting opportunities. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/figmas-way-ahead-ai-without-adobe.md: -------------------------------------------------------------------------------- 1 | # Figma's way ahead: with AI, without Adobe. 2 | 3 | [Figma is betting its AI tools can improve meetings and collaboration](https://www.fastcompany.com/91002435/figjam-is-betting-its-ai-can-make-meetings-less-terrible-even-without-adobes-backing?utm_source=bensbites\&utm_medium=referral\&utm_campaign=figma-s-way-ahead-with-ai-without-adobe) despite the failed Adobe acquisition. Figma's new AI features like Jambot and FigJam AI aim to make meetings more engaging and productive. 4 | 5 | ## What's going on here? 6 | 7 | Figma is adding AI to its ideation tool FigJam. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ce723e18-3ad1-4dc5-b3c0-5b446c1f8cd6/image.png?t=1704192841) 10 | 11 | ## What does this mean? 12 | 13 | Adobe was to acquire Figma for $20B. Regulators came in, the deal fell off. But in the past few months, Figma has been aggressively pursuing AI to differentiate itself. It acquired an AI startup called Diagram, held internal hackathons and launched various experimental AI features. 14 | 15 | These new tools primarily got added to its product called Figjam—the ideation and collaboration tool. They can now suggest icebreakers, summarize discussions, and rewrite meeting notes. As per Figma, early user response has been very positive, with people embracing the playful creative elements. 16 | 17 | ## Why should I care? 18 | 19 | Figma is expanding beyond its core design audience with AI. Two-thirds of its users are now non-designers. The larger selling point now is improved collaboration and ideation, regardless of technical skills. It won’t be surprising to see startups expand into additional tasks around their main offering. 20 | 21 | While the community is constantly hacking on “Figma to code” stuff, we haven’t seen much from Figma itself (apart from some plugins). 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/former-quartz-leader-building-new-york-times-ai-team.md: -------------------------------------------------------------------------------- 1 | # Former Quartz leader is building the New York Times' AI team 2 | 3 | So, the [New York Times is diving into AI](https://www.nytco.com/press/zach-seward-is-the-newsrooms-editorial-director-of-a-i-initiatives/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=former-quartz-leader-is-building-the-new-york-times-ai-team), but not with a lawsuit this time. It is assembling a fresh crew to explore how generative AI can jazz up their newsroom. They're on the lookout for some tech-savvy engineers and sharp editors to join the party. 4 | 5 | ## What's going on here? 6 | 7 | The New York Times is setting up a team to get AI into its journalism. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/6d92b515-a1b7-40c9-82a0-e985eb58efbf/image.png?t=1706701992) 10 | 11 | ## What does this mean? 12 | 13 | Zach Seward (founding editor of Quartz—went on to lead the publication) is leading the charge of NYT’s AI plans. And he is now announcing his plans on Threads. Where? Yup, that mix-breed of Twitter and Instagram. 14 | 15 | Anyway, he wants to recruit a mix of machine learning wizards, software engineers, designers, and editorial minds. Zach’s also gonna build out guidelines on where the NYT will use generative AI (and where it won’t). But the Times claims that reporting will always by humans. 16 | 17 | ## Why should I care? 18 | 19 | The Times has had its ups and downs with AI, blocking crawlers and even throwing down the legal punches against AI bigwigs like OpenAI and Microsoft. It's a bit hush-hush whether they'll buddy up with an AI firm or go DIY with their tools. 20 | 21 | One thing is clear, even the Times can’t take an “anti-AI” standpoint to the whole. They can shout “OpenAI bad” all they want, but not “AI bad”. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/fresh-openai-new-embedding-models-cost-savings.md: -------------------------------------------------------------------------------- 1 | # Fresh from OpenAI: New embedding models and cost savings 2 | 3 | [Fresh from OpenAI](https://openai.com/blog/new-embedding-models-and-api-updates?utm_source=bensbites\&utm_medium=referral\&utm_campaign=fresh-from-openai-new-embedding-models-and-cost-savings): They’ve launched two spanking new embedding models, spiced up GPT-4 Turbo, and made GPT-3.5 Turbo more wallet-friendly. There’s more on moderation and API usage below. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI's latest drop is all about giving devs more bang for their buck and some nifty new toys to play with. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1609695f-e598-451b-8b13-61e2fe0caf50/image.png?t=1706273527) 10 | 11 | ## What does this mean? 12 | 13 | The new embedding models: We've got a small one (text-embedding-3-small) that's leaner in price (we're talking a 5X price drop)and a large one (text-embedding-3-large) that’s the new heavyweight champ. 14 | 15 | GPT-3.5 Turbo is about to get cheaper, slashing input costs by 50% and output by 25%. GPT-4 Turbo's new preview model promises to end those frustrating 'lazy' responses (coders unite) and smooth over language hiccups for non-English tasks. 16 | 17 | The moderation model got an upgrade, making it tougher on potentially harmful content. We are also getting API key-level metrics in the usage dashboard for better tracking and the ability to assign permissions to API keys. 18 | 19 | ## Why should I care? 20 | 21 | This update is a toolkit upgrade. And a much-needed one at that. The old embedding models were still the leader but other companies had started to catch up. RAG and chat with data apps are gonna become cool again. 22 | 23 | And, everyone hates lazy chat, so let’s see how active the updated version is. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/frontier-model-forum.md: -------------------------------------------------------------------------------- 1 | # Frontier Model Forum 2 | 3 | **[Frontier Model Forum updates](https://openai.com/blog/frontier-model-forum-updates?utm_source=bensbites\&utm_medium=referral\&utm_campaign=frontier-model-forum)** - Together with Anthropic, Google, and Microsoft, OpenAI announced the new Executive Director of the Frontier Model Forum and a new $10 million AI Safety Fund. 4 | 5 | ## What's going on here? 6 | 7 | Major AI companies are banding together to promote AI safety research and practices. 8 | 9 | ## What does this mean? 10 | 11 | The Frontier Model Forum was formed by leading AI companies to advance responsible development of powerful AI models. They have now appointed Chris Meserole, who has extensive experience in AI policy, as Executive Director. The companies, along with philanthropic donors, are also committing over $10 million to launch an AI Safety Fund that will support independent research into AI safety techniques like red teaming. The Forum has published its first technical update on red teaming definitions and case studies to establish a common baseline. They aim to share knowledge and best practices on safely building, testing and evaluating the most capable AI systems. 12 | 13 | ## Why should I care? 14 | 15 | As AI becomes more powerful, ensuring it is developed safely and ethically is crucial. This collaboration between industry leaders shows commitment to making AI trustworthy by investing in safety research and bringing in outside experts. While AI promises immense benefits, its risks need to be addressed. Initiatives like the Frontier Model Forum and AI Safety Fund are important steps toward responsible AI innovation that considers not just capabilities, but consequences. Supporting independent scrutiny and high safety standards will lead to AI that works for people. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/google-secret-ai-tool.md: -------------------------------------------------------------------------------- 1 | # Google has a secret AI tool 2 | 3 | Google’s Gemini is the next awaited model in the AI space. But according to a leak, [Gemini is not the only AI thing Google is working on](https://medium.com/@bedros-p/gemini-is-coming-to-makersuite-so-are-stubbs-32248f3924aa?utm_source=bensbites\&utm_medium=referral\&utm_campaign=google-has-a-secret-ai-tool) at the moment. There might be a new tool called Stubbs on Google Makersuite for easily building basic AI prototype apps visually with no code required. 4 | 5 | ## What's going on here? 6 | 7 | Leaks show that Google will launch an AI app generator with its multimodal LLM Gemini. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/be7a54d3-9e42-4730-85f4-f1249ed7cc37/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Stubbs could allow users to prototype AI apps with a visual builder. Publish, share and remix options are also in the leaked screenshots meaning you might be able to share prototypes and edit Stubbs shared by others. 14 | 15 | Multimodal Gemini may power Makersuite and Google’s Vertex AI, before getting into Bard. It has an answer with an image option which might be generating images (or just including web images in its answers like Bard does). Another product Jetway, which might be just a sub-model of Gemini, could also be able to output HTML content. 16 | 17 | ## Why should I care? 18 | 19 | If you ask any developer building on top of LLMs, they’ll tell you that it’s way easier to work with Open AI’s API. The leaked images of Stubbs show a straightforward platform for creating GenAI apps. This could bring much-needed developer attention to Google’s AI models. Other than that, let’s wait because all of these are just leaks. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/google-spending-2b-anthropic.md: -------------------------------------------------------------------------------- 1 | # Google spending $2B more on Anthropic. 2 | 3 | Anthropic is the do-godder Witcher of AI and everyone is tossing (coins) billions at it. Amazon tossed $1.5B some weeks back (with a promise of going up to $4B in future) and became its preferred cloud partner. Google was feeling left behind (after being Anthropic’s early supporter) and now they are also shelling more quids. [Some $2B worth of ‘em.](https://www.wsj.com/tech/ai/google-commits-2-billion-in-funding-to-ai-startup-anthropic-db4d4c50?utm_source=bensbites\&utm_medium=referral\&utm_campaign=google-spending-2b-more-on-anthropic) 4 | 5 | ## What’s going on here? 6 | 7 | Google is reportedly investing $2B in Anthropic. 8 | 9 | ## What does this mean? 10 | 11 | According to the Wall Street Journal, Google has invested $500M of these $2B upfront with the rest of the $1.5B over time. I want to know who has the bigger stake in Anthropic now: Google or Amazon. My guess is that Google invested just enough to keep that majority stake. 12 | 13 | In total, Anthropic has already got $2B in the past few weeks with $4B more on the horizon. They are sitting on a huge cash pile which, if I’m to guess, is going into the training of their next model. 14 | 15 | ## Why should I care? 16 | 17 | Google’s investment seems weird, again. Google has its flagship model Gemini coming out soon. Much of this investment feels like Google hedging their bets. 18 | 19 | Remember the pitch deck from Anthropic that leaked earlier this year? Their claim was that in the upcoming 2-3 years, the leading model providers will be so far ahead that it’ll be impossible to catch them. Obviously, OpenAI is one of those leading providers. With all this cash, Anthropic wants to be another one itself. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/google-takes-openais-high-pay-offers-ai-talent.md: -------------------------------------------------------------------------------- 1 | # Google takes on OpenAI's high pay offers for AI talent. 2 | 3 | [Google is using special stock compensation](https://www.theinformation.com/articles/googles-defense-against-openai-talent-grab-special-stock?utm_source=bensbites\&utm_medium=referral\&utm_campaign=google-takes-on-openai-s-high-pay-offers-for-ai-talent) to retain its top AI researchers as OpenAI tries to poach them with huge pay packages. 4 | 5 | ## What’s going on here? 6 | 7 | Google is guarding its AI talent from leaving to join OpenAI. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/0090d729-e989-4326-9965-c066db56040d/image.png?t=1705492355) 10 | 11 | ## What does this mean? 12 | 13 | Google is tapping a pool of fast-vesting stock grants worth up to millions per person to prevent its key AI talent at DeepMind from leaving for OpenAI. Google is pulling out all the stops to keep its top AI experts from being lured away by OpenAI's massive compensation offers. 14 | 15 | Select DeepMind employees working on Google's Gemini model to rival ChatGPT are being given large grants of Google stock that vest in just 1 year, compared to the normal 4 year vesting period. These grants are in addition to salary and bonuses and can be worth millions based on the person. 16 | 17 | ## Why should I care? 18 | 19 | The talent war between Google and OpenAI is crazy and it’s not new. Many key AI researchers at OpenAI have previously worked at Google and led breakthroughs in the AI space. The departures of key personnel could further back Google's efforts to challenge OpenAI, and hence Google is trying to leverage its stock and existing talent pool to stay competitive. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/google-will-add-bard-assistant.md: -------------------------------------------------------------------------------- 1 | # Google will add Bard to its assistant 2 | 3 | Google announced the upcoming [combination of Google Assistant and Bard.](https://blog.google/products/assistant/google-assistant-bard-generative-ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=google-will-add-bard-to-its-assistant) At the Made by Google launch event, Google teased “Assistant with Bard” - the personal assistant with generative AI. 4 | 5 | ## What's going on here? 6 | 7 | Google showcases its plans for integrating Bard into Google Assistant. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c5cf8660-de3c-4414-89cb-5c9fc0cc8718/image.png) 10 | 11 | ## What does this mean? 12 | 13 | This new assistant aims to make it easier to manage tasks like planning trips, sorting emails, making grocery lists, and sending texts - as a human assistant would. The Assistant as usual will access the web and your apps to bring all the context, over which Bard will reason and generate answers/commands which the Assistant can then carry out. Assistant with Bard will be available as an opt-in after testing. 14 | 15 | ## Why should I care? 16 | 17 | The current lineup of features coming to Bard has been pushing it into a personal assistant direction. Still, accessing Bard via a website on your smartphone is not the ideal UX. Assistant with Bard is the natural progression for Google—the one that can finally create a seamless assistant experience. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/gpt5-works-sam-altman.md: -------------------------------------------------------------------------------- 1 | # GPT-5 is in under works - Sam Altman 2 | 3 | [Sam Altman did an interview with the Financial Times](https://www.ft.com/content/dd9ba2f6-f509-42f0-8e97-4271c7b84ded?utm_source=bensbites\&utm_medium=referral\&utm_campaign=gpt-5-is-in-under-works-sam-altman) and namedropped many things—GPT 5 being one of them. Sam talked about their true product, revenue growth, relationship with Microsoft and more. 4 | 5 | ## What's going on here? 6 | 7 | Sam Altman talks GPT5, superintelligence and future funding. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/b2976204-c3df-44e3-8758-a1475b95d262/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Altman says the partnership with Microsoft is working well, and he expects Microsoft to keep investing in OpenAI's pursuit of AGI. Revenue growth has been good this year, but costs to train AI models remain high, so OpenAI continues to operate at a loss. To build out its enterprise business, OpenAI has hired executives like COO Brad Lightcap. 14 | 15 | OpenAI’s true product is "superintelligence". Tools like GPTs to create more autonomous agents capable of complex tasks are a path to it. Altman confirmed that OpenAI is working on GPT-5, though didn’t give any timeline hints. Apart from compute, GPT-5 will need more data to be better than the previous models and OpenAI’s data programs may be a step in that direction. 16 | 17 | Altman also commented on the high demand for Nvidia’s chips being a factor in work this year but seems to think that the problem will be solved soon. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/hugging-face-google-cloud-teaming.md: -------------------------------------------------------------------------------- 1 | # Hugging Face and Google Cloud are teaming up 2 | 3 | [Hugging Face and Google Cloud are teaming up.](https://huggingface.co/blog/gcp-partnership?utm_source=bensbites\&utm_medium=referral\&utm_campaign=hugging-face-and-google-cloud-are-teaming-up) This is all about pushing the boundaries of open AI, making it a playground for us to innovate with the latest and greatest in AI models and tech. 4 | 5 | ## What’s going on here? 6 | 7 | In a nutshell, Hugging Face and Google are joining forces to supercharge AI development with open-source models and cutting-edge cloud technology. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/0f1c888f-5533-428e-a945-78a1b25ce645/image.png?t=1706270269) 10 | 11 | ## What does this mean? 12 | 13 | This collaboration is a game-changer. Google brings its heavyweight AI research and cloud hardware to the table, while Hugging Face adds its vast repository of transformer models and open-source libraries into the mix. For us, this means more tools in our toolkit: easy access to top-tier AI models, seamless integration with Google Cloud services, and the horsepower of Google's hardware, from TPUs to the latest GPUs. 14 | 15 | ## Why should I care? 16 | 17 | Imagine having the freedom to experiment with over a million AI models and datasets, and then deploying your creations with Google's infrastructure. Whether you're crafting the next big AI-driven service or fine-tuning models for your startup, this collaboration means you'll have the best of both worlds at your fingertips. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/humane-reveals-ai-pin-paris-fashion-week.md: -------------------------------------------------------------------------------- 1 | # Humane reveals Ai Pin at Paris Fashion Week 2 | 3 | **Humane**, a tech startup founded by former Apple designers Imran Chaudhri and Bethany Bongiorno, has revealed [its first product - the Ai Pin](https://hu.ma.ne/media/humanexcoperni?utm_source=bensbites\&utm_medium=referral\&utm_campaign=humane-reveals-ai-pin-at-paris-fashion-week) - at Paris Fashion Week. The wearable device was previewed in collaboration with fashion brand Coperni. 4 | 5 | ## What's going on here? 6 | 7 | Humane gave a sneak peek of its inaugural product, an AI-powered wearable called the Ai Pin, by having it worn by supermodel Naomi Campbell during the Coperni fashion show. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/8acaf198-4111-403d-a17c-ccdc77a89ed0/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Ai Pin is Humane's vision for AI-driven consumer tech that integrates seamlessly into daily life. The screenless, standalone wearable has sensors for natural interactions and doesn't need a smartphone. It uses optical recognition, laser projection, and Qualcomm's Snapdragon platform to enable AI capabilities. The full features will be unveiled on November 9th. 14 | 15 | ## Why should I care? 16 | 17 | This provides a first look at a new category of AI wearables for the post-smartphone era. With privacy built-in and a body-centric form, Humane aims to pioneer more personal, trust-based relationships between humans and AI. The preview highlights that the company is moving forward with shipping its first product. We may be entering the post-smartphone era of personal devices. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/ibm-puts-500m-venture-fund-enterprise-ai.md: -------------------------------------------------------------------------------- 1 | # IBM puts $500M into its venture fund for enterprise AI. 2 | 3 | [IBM has announced a $500 million investment fund](https://www.axios.com/2023/11/07/ibm-enterprise-ai-venture-fund?utm_source=bensbites\&utm_medium=referral\&utm_campaign=ibm-puts-500m-into-its-venture-fund-for-enterprise-ai) to back AI startups focused on enterprise and business customers. IBM wants to tap into innovative technologies and business models outside the company with this fund. 4 | 5 | ## What's going on here? 6 | 7 | IBM is jumping into the AI investing race to transform into a more startup-friendly company. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/84f6e13c-ea0e-47df-9451-7c6cb6321f67/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The $500M fund is an open broadcast to entrepreneurs that IBM is open for business and wants to work with new AI companies, just like other big tech companies. 14 | 15 | Specifically, IBM plans to invest across all stages without fixed targets. It has already backed prominent AI startups Hugging Face and HiddenLayer. The focus is on startups with AI solutions tailored to specific industries or business functions. IBM wants to avoid funding direct competitors to its portfolio companies. 16 | 17 | Rob Thomas, Chief Commercial Officer at IBM also feels that this wave may be led by companies (and markets) outside the US. 18 | 19 | ## Why should I care? 20 | 21 | IBM's want to reinvent itself into a more startup-oriented company compared to its former self. IBM likely hopes these investments will produce new technologies and revenue streams. 22 | 23 | This investment fund signals an impending shakeup in enterprise AI. IBM's openness to startups may accelerate AI adoption in more regulated sectors like healthcare, finance and manufacturing. 24 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/inflection2-makes-bold-claims.md: -------------------------------------------------------------------------------- 1 | # Inflection-2 makes bold claims 2 | 3 | Inflection, the AI startup by Reid Hoffman and Mustafa Suleyman has released their [new LLM Inflection-2](https://inflection.ai/inflection-2?utm_source=bensbites\&utm_medium=referral\&utm_campaign=inflection-2-makes-bold-claims). They claim it’s the world’s second-best LLM after GPT-4 based on benchmarks. 4 | 5 | ## What’s going on here? 6 | 7 | Inflection AI releases its second LLM Inflection 2. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f92d7f6e-fa72-46bf-97ca-8abfd46e289c/image.png?t=1700744459) 10 | 11 | ## What does that mean? 12 | 13 | Inflection-2 will soon power its consumer product PI (short for personal intelligence). Inflection-2 was trained on H100 GPUs (instead of A100s), resulting in lesser cost and higher speed while serving it, despite it being way bigger than Inflection-1. 14 | 15 | They have not released details about model size, training size etc., just a bunch of selective comparisons with models. So, the claim of the second-best LLM is a bit fragile. But the performance on benchmarks is still impressive putting it in the top 3-4 LLMs alongside GPT-4 and Claude 2. 16 | 17 | ## Why should I care? 18 | 19 | Soon after formation, Inflection announced that it’ll create the largest cluster of H100 GPUs to train LLMs. And that move seems to be working out for them based on the claims about cost and speed. Though the team behind Inflection is pro-regulation and state certifications, so don’t expect the full power of Inflection-2 to be available via Pi. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/intel-starts-new-ai-software-company.md: -------------------------------------------------------------------------------- 1 | # Intel starts new AI software company. 2 | 3 | [Intel is forming a new independent company called Articul8 AI](https://www.reuters.com/technology/intel-spins-out-ai-software-firm-with-backing-digitalbridge-2024-01-03/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=intel-starts-new-ai-software-company) to commercialize its artificial intelligence software, with financial backing from DigitalBridge Group and other investors. 4 | 5 | ## What's going on here? 6 | 7 | Intel is spinning out an independent AI software company called Articul8 AI. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1551d2c3-e478-4a9b-b9dc-e2aaf04f7f13/image.png?t=1704371424) 10 | 11 | ## What does this mean? 12 | 13 | Articul8 grew out of Intel's work with Boston Consulting Group on AI systems that address customers' privacy, security, and cost concerns about using cloud-based AI services. It will focus on getting enterprise AI systems into production that companies can run safely and in a cost-sustainable way within their own data centres. 14 | 15 | Articul8 will have an independent board, with Intel remaining a shareholder. Arun Subramaniyan, former vice president and general manager in Intel’s data center and AI group will become chief executive of Articul8. This aligns with Intel spinning out other units like Mobileye and its programmable chip division. 16 | 17 | ## Why should I care? 18 | 19 | The launch of Articul8 signals growing demand for usable enterprise AI amid data privacy concerns. Intel's backing shows its commitment to making AI technology more accessible through new business models. For companies, Articul8 could enable practical AI adoption while maintaining control of data. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/leading-news-websites-now-block-ai.md: -------------------------------------------------------------------------------- 1 | # Most leading news websites now block AI 2 | 3 | We're diving into the world of[AI and news media](https://www.wired.com/story/most-news-sites-block-ai-bots-right-wing-media-welcomes-them/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=most-leading-news-websites-now-block-ai). The gist? Top news outlets, especially the mainstream ones, are blocking AI data collection bots, but right-wing media outlets are saying, "Come on in." 4 | 5 | ## What’s going on here? 6 | 7 | Nearly 90% of leading news platforms are shutting their doors on AI bots, while their right-wing counterparts are rolling out the welcome mat. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/62aa32fa-a28b-4884-95d3-5be232081ddc/image.png?t=1706185276) 10 | 11 | ## What does this mean? 12 | 13 | Okay, so major players like The New York Times and The Guardian are part of this blockade. They're trying to control how their content feeds into AI systems. Meanwhile, right-wing sites like Breitbart aren’t bothered about blocking these bots. 14 | 15 | Researchers are speculating this might be a tactical move to influence AI models, considering these models learn from the content they consume. Also, right-wing media isn't too involved in the copyright scuffle that's got places like The New York Times at odds with AI companies. 16 | 17 | ## Why should I care? 18 | 19 | Here's why this matters to you: If you're into AI, tech, or just keeping up with the digital world, this is a big deal. It's about who gets to feed information into the ever-growing AI brain. This clash between media giants and AI firms could shape the future of how information is processed and presented by AI systems. We're looking at a tug-of-war over digital influence, and it's happening right now. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/meta-release-ai-persona-chatbots-week.md: -------------------------------------------------------------------------------- 1 | # Meta to release AI persona chatbots this week 2 | 3 | Wall Street Jorunal’s exclusive report mentions that [Meta is developing new AI chatbots](https://www.wsj.com/tech/ai/meta-ai-chatbot-younger-users-dab6cb32?utm_source=bensbites\&utm_medium=referral\&utm_campaign=meta-to-release-ai-persona-chatbots-this-week) that imitate fictional characters and celebrities, hoping to drive engagement among Gen Z. Meta’s goal is to bring the teenagers and young adults, who have flocked to TikTok, back over to Meta's apps like Instagram. 4 | 5 | ## What's going on here? 6 | 7 | Meta is rushing to release AI bots as early as this week, in sync with their Meta Connect conference starting Wednesday. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/e2e15f0d-74e0-4d3c-b2c3-c7b4e7d5ca80/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Meta is taking a big gamble on AI entertainment bots to boost usage, especially among youth. Internally codenamed "Gen AI Personas," these bots like "Bob the Robot" embody sassy personalities inspired by cartoons and celebs. The goal is to give Meta's messaging platforms their own version of viral sensations like Snapchat's My AI bot. 14 | 15 | ## Why should I care? 16 | 17 | Meta's desperate to stay culturally relevant among fickle young audiences. They want to create experiences powered by large language models that keep users on their apps. But personality-driven bots have already caused issues for Meta and rivals (Snapchat), making inappropriate comments in early testing. While responses may seem more "human" with added personality, it risks unsafe content. Still, higher level of comfort with AI in younger audiences gives Meta hope. More chatting could mean more ads and data collection down the line. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/microsoft-makes-new-ai-team-small-models.md: -------------------------------------------------------------------------------- 1 | # Microsoft makes new AI team for small models 2 | 3 | [Microsoft's latest strategy in the AI game is small models.](https://www.theinformation.com/articles/microsoft-forms-team-to-make-cheaper-generative-ai?utm_source=bensbites\&utm_medium=referral\&utm_campaign=microsoft-makes-new-ai-team-for-small-models) Meet their new GenAI team that’ll focus on developing smaller, more efficient AI models. At WEF, Satya Nadella (Microsoft’s CEO) mentioned the work on small models as “controlling our own destiny.” 4 | 5 | ## What’s going on here? 6 | 7 | Microsoft is focusing on small language models (SLMs) to cut down costs and computing power. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f8223192-267b-4812-a73f-0149da4a7c5c/image.png?t=1706092502) 10 | 11 | ## What does this mean? 12 | 13 | Microsoft has been big on AI, especially with their heavy investment in OpenAI. But we all know, GPT-4’s too power-hungry. So now, Microsoft is creating this new team to make AI more economical. They've pulled in their top AI brains from research teams into a GenAI team to develop these SLMs, including leading researcher Sébastien Bubeck. 14 | 15 | The GenAI team is different from Microsoft’s Turing team which focuses on making larger models. GenAI is led by Corporate Vice President Misha Bilenko and answers to Kevin Scott, the CTO. 16 | 17 | ## Why should I care? 18 | 19 | These SLMs are nifty because, with some optimizations and a narrow application range, they can do much of what the larger models do but on a diet of less computing juice. Microsoft’s Phi models, its attempt at SLMs, has gained much attention among companies and fine-tuning hackers. The Phi series of models is already being tested by Goldman Sachs. Nice, coz I’m all in for AI to become more accessible and cheaper to run for us all. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/microsoft-researchers-accidentally-exposed-38tb-private-data.md: -------------------------------------------------------------------------------- 1 | # Microsoft researchers accidentally exposed 38TB of private data 2 | 3 | [Microsoft researchers accidentally exposed 38TB of private data](https://www.wiz.io/blog/38-terabytes-of-private-data-accidentally-exposed-by-microsoft-ai-researchers?utm_source=bensbites\&utm_medium=referral\&utm_campaign=microsoft-researchers-accidentally-exposed-38tb-of-private-data), includingpersonal computer backups of Microsoft employees. It contains secrets, passwords, and**30,000 internal messages**. The culprit?**Misconfiguring access controls**when publishing open-source training data on GitHub. 4 | 5 | ## What's going on here? 6 | 7 | Microsoft researchers accidentally leaked private data while trying to share open AI models on GitHub. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/a811c2f6-feae-428e-9bc8-501619a3f07e/1695041345-04-teams.jpeg) 10 | 11 | ## What does this mean? 12 | 13 | Microsoft's AI team published some open-source AI code on GitHub and told people to download training data from a link. But that link accidentally gave access to 38TB of private Microsoft data, including employee computer backups with passwords, keys, Teams chats, etc. This happened because the link used a feature called SAS tokens, which can share Azure storage data. The researchers configured the token wrong, so it shared too much. This shows how AI teams now have extra security risks as they work with so much data. 14 | 15 | ## Why I should care? 16 | 17 | This example shows new data security risks as companies use more AI. Sharing lots of data for AI training can lead to leaks if security teams don't guide researchers properly. Also, AI models themselves can be hacked if they aren't reviewed well, which can spread malware. As AI spreads, security teams need to work closer with researchers to prevent issues like this leak. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/microsoft-surface-ai-event-nyc-expect.md: -------------------------------------------------------------------------------- 1 | # Microsoft Surface and AI event in NYC - What to expect 2 | 3 | [Microsoft is holding a “special event” in New York City on September 21st.](https://www.theverge.com/2023/9/19/23878352/microsoft-surface-ai-event-nyc-laptop-studio-2-go-laptop-go-rumors-leaks?utm_source=bensbites\&utm_medium=referral\&utm_campaign=microsoft-surface-and-ai-event-in-nyc-what-to-expect) The rumours hint at\*\*three new Surface devices.\*\*Obviously, Surface devices aside,**AI updates across the board**, including Windows Copilot, Bing Chat, and Microsoft 365 are expected. 4 | 5 | ## What’s going on here? 6 | 7 | The event will be held in person with no live streaming. We'll see Microsoft's vision for integrating AI into its products. 8 | 9 | ## What does this mean? 10 | 11 | Three new Surface devices are expected - refreshed Laptop Studio, Go (tablet), and Laptop Go models. Leaks show more focus on software changes and lesser on the form factor. Updates to Windows copilot are also expected to be in the pipeline. More AI-powered features are already coming to Windows apps like Photos and Snipping Tool. Surface devices may also get new AI capabilities in the camera and mics. 12 | 13 | ## Why should I care? 14 | 15 | Surface laptops and tablets launched in the event will reflect how Microsoft is looking at both software and hardware in the context of AI. They’ve previously used ARM’s NPU for offloading AI tasks in their devices, but now both Intel and AMD have those offerings. This event will showcase Microsoft's progress towards its stated goal of reinventing Windows and Surface with AI. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/microsofts-next-copilots-copilot-sales-service-now-grabs.md: -------------------------------------------------------------------------------- 1 | # Microsoft’s next copilots: Copilot for Sales and Service are now up for grabs. 2 | 3 | [Microsoft just rolled out Copilot for Sales and Service](https://cloudblogs.microsoft.com/dynamics365/bdm/2024/02/01/microsoft-copilot-for-sales-and-copilot-for-service-are-now-generally-available?utm_source=bensbites\&utm_medium=referral\&utm_campaign=microsoft-s-next-copilots-copilot-for-sales-and-service-are-now-up-for-grabs), integrating AI into the daily grind of sales and customer service pros. 4 | 5 | ## What's going on here? 6 | 7 | Microsoft’s next copilots: Copilot for Sales and Service are now up for grabs. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5c5c5ec6-e6d7-479c-bfef-ea50606d9e33/image.png?t=1706880245) 10 | 11 | ## What does this mean? 12 | 13 | For the sales warriors, Copilot will prep your meeting briefs, summarise those never-ending email threads, and even nudge your CRM updates—all from Microsoft apps like Word, Outlook and OneNote. 14 | 15 | Over in the service arena, Copilot for Service is cutting through the noise. AI-powered conversations across data sources as it comes with pre-built integrations with Salesforce, ServiceNow, and Zendesk. There’s also an option to have an embedded copilot in agent desktops for easier support and info access. 16 | 17 | Both of these are $50 per user/month. But if you’re already on the Copilot for Microsoft 365 train, you can add on these new tools for $20 more per user/month. 18 | 19 | ## Why should I care? 20 | 21 | A big part of sales and customer service jobs (like logging emails) are tasks that AI can do faster, and better and leave the human to do more meaningful stuff like connecting with the customers. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/microsofts-plans-beyond-open-ai.md: -------------------------------------------------------------------------------- 1 | # Microsoft's plans beyond Open AI 2 | 3 | [Microsoft is focusing on its plans beyond Open AI](https://www.theinformation.com/articles/how-microsoft-is-trying-to-lessen-its-addiction-to-openai-as-ai-costs-soar?utm_source=bensbites\&utm_medium=referral\&utm_campaign=microsoft-s-plans-beyond-open-ai) by directing internal researchers and resources to make smaller models that can take the place of GPT-4 in relevant situations. 4 | 5 | ## What’s going on here? 6 | 7 | Microsoft wants to build its own advanced AI to reduce costs and dependence on OpenAI. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/25c8e587-1f0f-4cd3-9462-4f436e0c35cd/02d68d3a-f324-452e-8875-6662599ee2b0.jpeg) 10 | 11 | ## What does this mean? 12 | 13 | Microsoft is directing its researchers to create conversational AI models that perform nearly as well as OpenAI's but are much smaller and cheaper to run. This in-house AI is already being tested by Microsoft's Bing team for features similar to ChatGPT. Orca and Phi models from Microsoft are likely the results of the same push. 14 | 15 | The goal is to save on the ballooning compute costs of large AI models. Even though Microsoft has invested over $10 billion in OpenAI for exclusive access to its tech, unchecked costs from widespread use could get out of hand fast. Smaller "distilled" models that mimic the capabilities of behemoths like GPT-4 can help control expenses while providing customers with powerful AI features. 16 | 17 | ## Why should I care? 18 | 19 | This gives Microsoft options beyond OpenAI to deliver performant and affordable AI products. It's a smart business move to avoid overreliance on external partners, however fruitful the relationship may be currently. Plus having in-house alternatives puts Microsoft in a better negotiating position with OpenAI down the road. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/microsofts-plans-dev-day.md: -------------------------------------------------------------------------------- 1 | # Microsoft's Plans after Dev Day 2 | 3 | At OpenAI’s first DevDay conference, Microsoft CEO Satya Nadella made a surprise appearance during Sam Altman’s keynote. While Sam kept poking him, Satya was calm in repeating the fact that they are building the infrastructure that OpenAI needs. To which Sam replied: we have the [best partnership in tech.](https://www.microsoft.com/en-us/microsoft-cloud/blog/2023/11/07/come-build-with-us-microsoft-and-openai-partnership-unveils-new-ai-opportunities/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=microsoft-s-plans-after-dev-day) 4 | 5 | ## What’s going on here? 6 | 7 | Microsoft is again leveraging Open AI’s new announcements and plans to announce more features based off them. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1d960f93-cd9f-47e3-ac4b-422212dbcd71/image.png) 10 | 11 | ## What does that mean? 12 | 13 | The new GPT-4 Turbo offers lower pricing, longer prompts, and JSON formatting for improved efficiency. These Turbo models will be available on Azure OpenAI Service by the end of 2023. On Azure, pricing will match OpenAI's prices. 14 | 15 | The new custom chatbots, GPTs will allow anyone to create tailored ChatGPT versions for specific tasks or audiences. Microsoft is planning a deep ecosystem support for GPTs that they’ll announce soon. 16 | 17 | Microsoft will provide seamless integration of OpenAI and support infrastructure, meeting all enterprise needs for the new custom models program. Microsoft also gave DevDay attendees free 90-day GitHub Enterprise access. 18 | 19 | ## Why should I care? 20 | 21 | They might seem like they are competing for enterprise attention, but both of them respect their mutual dependency to keep marching ahead in the race of AI. (even though Google is nowhere to be seen). 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/nadella-testifies-google-apple-deal.md: -------------------------------------------------------------------------------- 1 | # Nadella testifies against Google and Apple deal 2 | 3 | [Satya Nadella, Microsoft's CEO, testified in an antitrust trial against Google](https://www.wsj.com/tech/ai/microsoft-ceo-testifies-about-limits-of-ai-at-google-antitrust-trial-32d032a9?utm_source=bensbites\&utm_medium=referral\&utm_campaign=nadella-testifies-against-google-and-apple-deal) about anti-competitive search practices. Nadella thinks it will take more than AI to check Google’s advantage in internet search. 4 | 5 | ## What's going on here? 6 | 7 | Nadella argued Google uses distribution deals and dominance to trap Microsoft's Bing in an unwinnable "vicious cycle." 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5e1a8bae-7c1f-4913-957c-306b41b652f9/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Google's search defaults on Apple products like Safari reinforce user habits that maintain its 90% market share. This lets Google continually improve its product while disadvantaging Microsoft. Nadella called switching search engines practically impossible given Google's entrenched position. 14 | 15 | Nadella was skeptical new AI could disrupt this cycle. He worries defaults will still favor Google, even if tools like ChatGPT create new search angles. Earlier optimism about transforming search with AI was possibly "exuberance" from leading a small player. 16 | 17 | ## Why should I care? 18 | 19 | The trial reveals how tech giants leverage dominance. Google's agreements locked in market share gains from a better search product. This insight matters if you believe search needs more competition or that Google unfairly restricts rivals. Understanding the "no fly zone" Google created around search helps explain the industry's dynamics. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/normal-people-define-rules-ai.md: -------------------------------------------------------------------------------- 1 | # What if normal people define the rules for AI 2 | 3 | [Anthropic and the Collective Intelligence Project ran a public input process to create an AI constitution.](https://www.anthropic.com/index/collective-constitutional-ai-aligning-a-language-model-with-public-input?utm_source=bensbites\&utm_medium=referral\&utm_campaign=what-if-normal-people-define-the-rules-for-ai) They discovered areas of agreement and disagreement with their in-house constitution. 4 | 5 | ## What's going on here? 6 | 7 | The public input resulted in a moderately different constitution for AI than Anthropic's internal one. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/041ed9de-d7c8-4cbf-af06-fc2165373c6c/image.png) 10 | 11 | ## What does this mean? 12 | 13 | There was about 50% overlap between the public and Anthropic constitutions. The public constitution focused more on promoting desired behaviour rather than avoiding undesired behaviour. Some public statements were excluded due to a lack of consensus or being problematic. 14 | 15 | Training a model on the public-sourced constitution versus Anthropic's own reduced certain biases, and had similar political opinions without any loss in performance. 16 | 17 | ## Why should I care? 18 | 19 | This work on sourcing the rules from the public is great step in building trust around AI models and what they generate. More people getting a sense of belonging and control over the models can increase the adoption as well. At the same time, this still also involves many judgment calls from Anthropic as in participant selection, platform, seed statements, moderation, and more. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/nvidia-cooking-sushai-japan.md: -------------------------------------------------------------------------------- 1 | # Nvidia is cooking “sushAI” in Japan 2 | 3 | [Nvidia will collaborate with Japanese companies](https://www.japantimes.co.jp/business/2023/12/04/nvidia-japan-ai-processors-priority/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=nvidia-is-cooking-sushai-in-japan), research institutions, and startups to create "factories of AI" and invest in developing homegrown AI talent and infrastructure. 4 | 5 | ## What’s going on here? 6 | 7 | Nvidia CEO says he will try to prioritize Japan for AI processors. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/7cc3b96a-ae3f-4fdd-9384-6b2df715ce13/image.png?t=1701768969) 10 | 11 | ## What does this mean? 12 | 13 | Nvidia's CEO Jensen Huang stated Japan has no need to export its data or AI, implying it has the technical expertise and industrial capability to nurture a self-sufficient AI ecosystem. Huang also met with the Japanese Prime Minister and said that he has promised the PM to prioritise Japan’s request for GPUs. 14 | 15 | ## Why should I care? 16 | 17 | We are going to see this more: countries wanting to get AI companies to work with them (especially the hardware makers). And at the same time, companies going outside just the US/EU to find more flexible business terms (like China-related regulations). 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-building-two-type-ai-agents.md: -------------------------------------------------------------------------------- 1 | # Open AI is building two type of AI agents. 2 | 3 | [OpenAI's new project after ChatGPT is crafting AI agents](https://www.theinformation.com/articles/openai-shifts-ai-battleground-to-software-that-operates-devices-automates-tasks?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-is-building-two-type-of-ai-agents) that can literally take the wheel of your digital tasks. Think virtual buddies who are ace at handling your digital errands. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI is working on AI agents that not just chat but actually get stuff done on your computer and online. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5bc01b55-6c90-4b80-821d-dbebd8d5e52a/image.png?t=1707398577) 10 | 11 | ## What does this mean? 12 | 13 | OpenAI is working on two types of agents for now. One type of agent is like a digital puppeteer for your computer, doing all the clicky-typey stuff you'd rather not. Need to shuffle data from a doc to a spreadsheet or sort out pesky expense reports? This agent's got you. 14 | 15 | The other type is your go-to web ninja, scouting data, planning trips, or snagging flight deals without you breaking a sweat. But, and it's a big but, these agents need to learn about you to be truly helpful - which means letting them in on your digital secrets, with your say-so, of course. 16 | 17 | ## Why should I care? 18 | 19 | Because who doesn't want a personal digital whiz? Beyond just cutting down your task list, agents could seriously change most professional work that happens on a screen. And as these agents evolve, they'll get smarter at understanding our needs, making instances like [this](https://twitter.com/andreasklinger/status/1755490423777394888?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-is-building-two-type-of-ai-agents) more common. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-calls-experts-red-teaming.md: -------------------------------------------------------------------------------- 1 | # Open AI calls experts for red teaming 2 | 3 | OpenAI announced an open call for experts to join the [OpenAI Red Teaming Network](https://openai.com/blog/red-teaming-network?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-calls-experts-for-red-teaming)to help assess risks of AI models before deployment. This expands their safety efforts beyond internal testing to continuous input from external researchers and civil society. 4 | 5 | ## What's going on here? 6 | 7 | OpenAI is formalizing ongoing collaborations with outside experts into a network for iterative red teaming of AI systems. 8 | 9 | ## What does this mean? 10 | 11 | Rather than one-off engagements before launches, OpenAI will maintain a network of trusted experts in diverse domains that can provide perspectives on potential harms throughout development. Members will sign NDAs and provide at least 5-10 hours annually. OpenAI will select experts case-by-case for red-teaming new models based on their expertise. This complements their other safety initiatives like the Researcher Access Program. 12 | 13 | ## Why should I care? 14 | 15 | This network enables broader, continuous input on AI risks from diverse experts vs. just internal testing. As AI grows more capable and influential, thoughtful oversight mechanisms like this could help align it with human values. While details are sparse, the open call signals a meaningful commitment to cooperative safety efforts. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-gives-employees-time-cash-shares.md: -------------------------------------------------------------------------------- 1 | # Open AI gives employees more time to cash out their shares. 2 | 3 | OpenAI is moving forward with the tender offer after briefly putting it on hold when CEO Sam Altman was briefly ousted, and then reinstated. This means employees can sell shares in a deal valuing OpenAI at $86 billion. The tender offer [timeline has been extended by a month](https://www.bloomberg.com/news/articles/2023-11-30/openai-tender-for-employee-shares-is-on-and-extended-to-jan-5?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-gives-employees-more-time-to-cash-out-their-shares) to January 5th due to recent turmoil. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI’s tender offer is being extended to give employees more time. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/46690362-6c30-4596-93dd-18ba1bfadd10/image.png?t=1701432775) 10 | 11 | ## What does this mean? 12 | 13 | By proceeding with the tender offer, OpenAI is sticking to its plan despite the leadership turmoil. Some investors had dropped out when Altman was fired, but there's now enough demand to cover the deal. OpenAI sees enough interest to cash out employees at an eye-popping $86 billion valuation. The extension gives employees more time to opt in after the chaos. Essentially, early employees can unlock the value of their equity. 14 | 15 | ## Why should I care? 16 | 17 | The tender offer signals that OpenAI has stabilized after its leadership crisis and remains on track to go public eventually. The soaring valuation cements its status as a hugely valuable AI leader. Safe to say that despite the speed bumps, investor interest in OpenAI is still through the roof. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-introduces-data-partnerships.md: -------------------------------------------------------------------------------- 1 | # Open AI introduces data partnerships 2 | 3 | [OpenAI is introducing Data Partnerships](https://openai.com/blog/data-partnerships?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-introduces-data-partnerships) to work with organizations to produce public and private datasets for training AI models. It wants to broaden the training data for AI to cover more subjects, industries, cultures and languages so models can better understand and serve all of humanity. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI is on the lookout for more data for its models via its data partnership program. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/92647377-ac5b-4c8f-89ff-08ebd886b12d/image.png) 10 | 11 | ## What does this mean? 12 | 13 | OpenAI is already partnering with organizations and governments to contribute data representing specific industries or countries. It has two stated goals for this program: Create an open-source archive of diverse datasets and private datasets to train proprietary OpenAI models. 14 | 15 | In turn, Open AI is promising world-class OCR (for PDFs, images etc.) and ASR (for audio and video files) to turn your data future-ready. 16 | 17 | ## Why should I care? 18 | 19 | AI works if it helps everyone. To make that a reality diverse datasets are a need. Data partnerships can get more representation from areas that don’t have an active presence on the open web. For countries, it’s crucial to have a footprint in the systems that are being used globally. 20 | 21 | OpenAI says if you have data you wish to keep private but want AI to better understand, you can partner for private datasets. But do you really want to share what might be the only edge a couple of years in the future? 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-planning-features-developers.md: -------------------------------------------------------------------------------- 1 | # Open AI is planning more features for developers 2 | 3 | Open AI is holding its first developer conference in November and Reuters’ report say that [OpenAI plans to roll out major updates](https://www.reuters.com/technology/openai-plans-major-updates-lure-developers-with-lower-costs-sources-2023-10-11/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-is-planning-more-features-for-developers) soon in order to attract more developers. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI wants to make its AI models cheaper and easier for developers to build applications with. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/474c348f-5e02-4a1e-a1d0-c506d9216dca/image.png) 10 | 11 | ## What does this mean? 12 | 13 | First, OpenAI could add memory storage capabilities to reduce computing costs for developers (with estimates that this could cut expenses by up to 20 times). Big deal! 14 | 15 | Second, the new GPT-4V is likely to be offered in the API. That’ll open up a huge number of new use cases, attracting developers to build on it. allow developers to build apps that can analyze images and describe them. This unlocks many potential use cases beyond just text generation. 16 | 17 | ## Why should I care? 18 | 19 | OpenAI has the most developer-friendly interface to work with. They recently improved the UI for their cookbook, launched a fine-tuning dashboard and some other minor updates in a similar direction. The se updates from OpenAI hint that devs are still one of their key focus areas (despite the consumer success of ChatGPT). 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-releases-dalle-3.md: -------------------------------------------------------------------------------- 1 | # Open AI releases DALL-E 3 2 | 3 | [Open AI has announced its next image model, DALL-E 3.](https://openai.com/dall-e-3?utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-releases-dall-e-3) DALL-E 3 understands significantly **more nuance and detail** than DALL-E 2. Achieving this nuance doesn’t require extensive prompt engineering with DALL-E 3. It is **built natively on ChatGPT**and Plus users would be able to use it similar to a ChatGPT plugin starting in October. 4 | 5 | ## What's going on here? 6 | 7 | OpenAI’s latest image model, DALL-E 3 is now available in research preview. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/260da4e5-ac4e-4f86-9161-d7df196c80b3/image.png) 10 | 11 | ## What does this mean? 12 | 13 | DALL-E 3 aims to solve the tendency of image models to ignore secondary details in the prompt. Integration with ChatGPT means that you can use OpenAI’s language models along with DALL-E 3 which allows more detailed brainstorming. The new model will also be available via API and Open AI labs soon. 14 | 15 | ## Why should I care? 16 | 17 | ChatGPT lacks visual AI skills. The addition of DALL-E 3 to ChatGPT kills two birds with one stone: making the product better and taking on Google whose upcoming Gemini model is also multimodal (images and text together). API and Labs availability of this improved model will also attract developers who had been looking towards Midjourney and SDXL for their projects. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-seeking-a-90-billion-valuation.md: -------------------------------------------------------------------------------- 1 | # Open AI seeking a $90 Billion valuation 2 | 3 | OpenAI, the AI startup behind ChatGPT, is seeking a [new valuation of up to $90 billion](https://www.wsj.com/tech/ai/openai-seeks-new-valuation-of-up-to-90-billion-in-sale-of-existing-shares-ed6229e0?mod=tech_lead_story\&utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-seeking-a-90-billion-valuation) by allowing employees to sell existing shares. 4 | 5 | ## What's going on here? 6 | 7 | OpenAI aims to triple its valuation from earlier this year through a new share sale. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/4ad195af-8bb0-44ed-ae70-8400316dc5f3/image.png) 10 | 11 | ## What does this mean? 12 | 13 | OpenAI's valuation is skyrocketing thanks to ChatGPT's popularity and its expected billion-dollar revenue this year. OpenAI is in talks to allow employees to sell existing shares to investors, which could value the company at $80-90 billion—almost triple its earlier $29 billion valuation. 14 | 15 | ## Why should I care? 16 | 17 | This highlights the exploding value of AI due to chatbots like ChatGPT. ChatGPT's widespread popularity is driving OpenAI's revenue, making it one of the most valuable global startups. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/open-ai-wants-google-researchers-even-pay-millions.md: -------------------------------------------------------------------------------- 1 | # Open AI wants Google Researchers, even if it has to pay millions 2 | 3 | The artificial intelligence talent war between OpenAI and Google is intensifying as [OpenAI tries to lure top AI researchers from Google with multimillion dollar pay packages](https://www.theinformation.com/articles/openais-new-tack-in-talent-war-with-google-promising-recruits-a-quick-stock-bump?rc=bdorru\&utm_source=bensbites\&utm_medium=referral\&utm_campaign=open-ai-wants-google-researchers-even-if-it-has-to-pay-millions). OpenAI plans to triple its valuation to over $80 billion through an employee share sale, so it is telling researchers who join now they can benefit from the upcoming increase. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI is aggressively recruiting top AI talent from Google with heavy paychecks. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/3d1bec26-6ec2-4d86-8dfe-1cf867f7008d/image.png) 10 | 11 | ## What does that mean? 12 | 13 | In the “Game of AI Thrones”, you need many houses to support you. The GPU house is often talked about but Open AI is (and has been) looking at the talent house. It is offering senior researchers from Google pay packages worth around $5-10 million in stock compensation. The rumour is that OpenAI is guaranteeing them huge paydays if they get in before OpenAI's valuation spikes from $27 billion to over $80 billion. 14 | 15 | ## Why should I care? 16 | 17 | Open AI already includes many top researchers who were previously at Google. The transformer architecture was created by researchers at Google. Open AI has been “attracting” talent like a super magnet and if this move ends up shifting more researchers to their side, it’ll be a hard game for Google to keep pace. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/openai-adds-election-guardrails-chatgpt.md: -------------------------------------------------------------------------------- 1 | # OpenAI adds election guardrails to ChatGPT 2 | 3 | If you [ask ChatGPT about US elections](https://dmicz.github.io/machine-learning/chatgpt-election-update/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-adds-election-guardrails-to-chatgpt) now, it won't discuss it and will refer you to CanIVote.org instead. This new tool lets OpenAI set policies on what ChatGPT can and can't talk about. 4 | 5 | ## What's going on here? 6 | 7 | OpenAI recently added a new tool to ChatGPT that limits what it can say about US elections. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/924ff6e4-5447-4e45-aacd-2a3116eab536/image.png?t=1705309501) 10 | 11 | ## What does this mean? 12 | 13 | OpenAI quietly put a "guardian\_tool" function into ChatGPT’s content policy that stops it from talking about voting and elections in the US. It now tells people to go to [CanIVote.org](http://CanIVote.org?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-adds-election-guardrails-to-chatgpt) for that info. OpenAI is being proactive about ChatGPT spreading misinformation before the 2024 US elections. 14 | 15 | The tool isn't just for elections either - OpenAI can add policies to restrict other sensitive stuff too. Since it's built-in as a function, ChatGPT will automatically know when to use it based on the conversation. It goes beyond the previous ways OpenAI trained ChatGPT. 16 | 17 | ## Why should I care? 18 | 19 | In 2024, half of the world will be going through elections. OpenAI is taking steps to use AI responsibly as ChatGPT is getting more popular. Hallucinations are still present in chatGPT (and other LLM systems). Restricting election info and redirecting to resources that have human-verified information is a safe way to deal with the current state of the world and these systems—for people and OpenAI both. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/openai-adds-metadata-ai-images-chatgpt.md: -------------------------------------------------------------------------------- 1 | # OpenAI adds metadata to AI images in ChatGPT 2 | 3 | [OpenAI's stepping up its game by embedding C2PA magic into images](https://help.openai.com/en/articles/8912793-c2pa-in-dall-e-3?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-adds-metadata-to-ai-images-in-chatgpt) made by DALL·E 3. That means any pic you get from ChatGPT or the API comes with an ID card you can check out. 4 | 5 | ## What’s going on here? 6 | 7 | Images generated via ChatGPT and OpenAI API now include metadata using C2PA specifications. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c0505fdc-fcda-4723-b36e-dc23b6f70432/image.png?t=1707311267) 10 | 11 | ## What does this mean? 12 | 13 | C2PA is a techie standard that's all about making sure the media we see and share is legit. Think of it as a digital ID card for images, telling you where they came from and what's their story. 14 | 15 | So, every image you get from DALL·E 3 now comes with this C2PA tag - it's like a seal of authenticity. But here's the kicker: not everything's foolproof. This tag can get lost in the digital shuffle, like when images hit social media or get screenshot. OpenAI's clear that while C2PA's cool, it's not the end-all for proving where an image came from. 16 | 17 | ## Why should I care? 18 | 19 | Knowing what's real and what's not is getting tougher. There’s a flood of AI images coming that we’ll have to deal with every day (most of us already are). Adding C2PA tags on AI images is a try to sift through that noise, even if it's not perfect. 20 | 21 | Also read: [Meta is trying to label AI content on its platforms.](https://bensbites.beehiiv.com/p/meta-trying-label-ai-content-platforms) 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/openai-exploring-making-ai-chips.md: -------------------------------------------------------------------------------- 1 | # OpenAI is exploring making its own AI chips. 2 | 3 | [Open AI is exploring making its own AI chips](https://www.reuters.com/technology/chatgpt-owner-openai-is-exploring-making-its-own-ai-chips-sources-2023-10-06/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-is-exploring-making-its-own-ai-chips) to solve its shortage of expensive AI chips that power its software and products. The company relies heavily on Nvidia's graphics processing units (GPUs) and Microsoft’s partnership to access them. 4 | 5 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/e36694f7-47eb-45c7-8405-b08fad0715c6/image.png) 6 | 7 | ## What does this mean? 8 | 9 | Open AI sees in-house chip development as key to tackling two big problems: the global scarcity of high-end AI chips and the massive computing costs of running systems like ChatGPT. CEO Sam Altman has made acquiring more AI chips a top priority given Open AI's reliance on Nvidia, which controls over 80% of the AI chip market. 10 | 11 | ## Why should I care? 12 | 13 | Custom AI chips could significantly reduce Open AI's hardware costs and chip shortage issues, allowing it to scale its AI technologies more easily. However, building custom AI chips is a complex and expensive endeavour that could cost hundreds of millions of dollars with no guarantee of success. Even if Open AI acquires a chip company, it would still take years to develop its own chips. 14 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/openai-hustles-beat-google-launch-multimodal-llm.md: -------------------------------------------------------------------------------- 1 | # OpenAI hustles to beat Google to launch ‘Multimodal’ LLM 2 | 3 | [OpenAI hustles to beat Google to launch ‘Multimodal’ LLM](https://www.theinformation.com/articles/openai-hustles-to-beat-google-to-launch-multimodal-llm?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-hustles-to-beat-google-to-launch-multimodal-llm), as reported by The Information. With all the reports of Gemini being released soon and potentially better than GPT-4, Open AI is trying to keep its lead intact. The multimodal features will be launched under the name **“GPT-vision.”** Also, they are training a **multimodal LLM from scratch codenamed Gobi.** 4 | 5 | ## What’s going on here? 6 | 7 | Google and OpenAI are competing to release the next generation of AI models that can understand both text and images. 8 | 9 | ## What does this mean? 10 | 11 | These new “multimodal” models will be able to generate code from website sketches and analyze visual data, among other capabilities. Google is close to releasing its Gemini model, while OpenAI is rushing to add multimodal features to GPT-4. There are concerns around potential misuse, but both companies are taking steps to ensure responsible development. This race parallels big tech platform competitions like iPhone vs Android. 12 | 13 | ## Why should I care? 14 | 15 | These more advanced AI systems will likely enable new applications and expand what computers can do. However, their capabilities also raise important questions around ethics, bias, and proper oversight that impact society. Following this innovation race helps understand the pace of AI progress and emergence of new powers that could affect daily life. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/openai-kills-chat-pdf-apps.md: -------------------------------------------------------------------------------- 1 | # OpenAI kills Chat with PDF apps. Or did it? 2 | 3 | Some people are getting [updates to GPT-4](https://twitter.com/ldjconfirmed/status/1718456393026490523?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-kills-chat-with-pdf-apps-or-did-it) where you can use, Vision, ADA (code interpreter), Browsing and DallE-3 in the same chat. It’s still rolling out, and I need candy because I haven’t got it yet. And the other update is making faux investors lose their sleep. 4 | 5 | ## What’s going on here? 6 | 7 | ChatGPT is getting upgrades to how it uses files and tools. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/0aeeee95-0122-4ab4-8698-6029124452f7/image.png) 10 | 11 | ## What does that mean? 12 | 13 | ChatGPT can now accept multiple files of different types at the same time. Many Chat with PDF apps were building their product around these features. As always OpenAI adding this feature was expected and the equally expected was the outrage that “Oh! Chat with PDF apps are dead.” 14 | 15 | The tool usage appears as “all tools” in a chat using multiple tools. There are some rumours about another mode called “Magic Create,” but no deets about that. 16 | 17 | ## Why should I care? 18 | 19 | Though the drama about your app being dead is not true, you’ve got to care a bit extra after this. Add additional features to your app, innovate on UI, focus on a niche, double down on marketing, you get the idea. There’s a lot to do. 20 | 21 | As a ChatGPT user, I’ve been waiting for the all tools consolidation for a while. Try out the use cases that needed you to switch threads and lose context, it might perform better now. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/openai-wants-pay-publishers-not-much.md: -------------------------------------------------------------------------------- 1 | # OpenAI wants to pay publishers. But not too much. 2 | 3 | [OpenAI and Apple are negotiating content licensing deals](https://www.theinformation.com/articles/openai-offers-publishers-as-little-as-1-million-a-year?utm_source=bensbites\&utm_medium=referral\&utm_campaign=openai-wants-to-pay-publishers-but-not-too-much) with news publishers to train their AI systems. But according to sources the deal sizes are smaller than the industry expects. Some publishers feel they aren't getting fair value. 4 | 5 | ## What’s going on here? 6 | 7 | OpenAI deal size for access to publishers’ content is smaller than expected. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/d920bb16-0ac2-41f3-9879-c908af394a84/image.png?t=1704451297) 10 | 11 | ## What does that mean? 12 | 13 | We’re talking about offers as low as $1-5 million annually to some publishers. It’s like getting the assistant a coffee to peek inside the boss’s laptop. Apple is saying the “dinner’s on me” but wants to copy stuff to its hard drive. Basically more flexibility in how it uses the content. Google feels behind in the race (again) but let’s not forget that Google already has TeamViewer access with its products like Google News. 14 | 15 | Although OpenAI has made some big-name deals earlier (Alex Springer and Associated Press), more publishers will want to hold out and see how the NYT’s case against OpenAI turns out. 16 | 17 | ## Why should I care? 18 | 19 | OpenAI (and other AI companies) will push harder to secure these contracts with publishers to a) get their hands on unique data and b) signal their “friendship” with creators. At the same time, it’s likely that more publisher will sue AI companies over the use of their materials. 20 | 21 | I wonder if there’s an opportunity for smaller media houses here to go with the AI companies, instead of against them. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/optout-google-ai-ranking-google-search.md: -------------------------------------------------------------------------------- 1 | # Opt-out of Google AI while ranking on Google Search 2 | 3 | [Google has launched a new tool to allow websites to opt out](https://blog.google/technology/ai/an-update-on-web-publisher-controls/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=opt-out-of-google-ai-while-ranking-on-google-search) of having their content used to train AI models while still being indexed by Google Search. 4 | 5 | ## What's going on here? 6 | 7 | Google has created an opt-out tool for websites to control whether their content helps improve Google's AI models like Bard and Vertex. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/a4e36ffd-cc99-4602-a83c-f5f893a49d46/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The tool, called Google-Extended, allows sites to block their content from training Google's AI while still being crawled and indexed by Googlebot. This gives publishers more control over access to their content. 14 | 15 | ## Why should I care? 16 | 17 | If you run a website, this tool gives you greater say over how your content is used, especially for sensitive domains where AI training could be problematic. Having a choice helps build trust with Google. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/perplexity-launches-low-latency-api-open-source-llms.md: -------------------------------------------------------------------------------- 1 | # Perplexity launches low latency API for open source LLMs 2 | 3 | [Perplexity launches pplx-api,](https://blog.perplexity.ai/blog/introducing-pplx-api?utm_source=bensbites\&utm_medium=referral\&utm_campaign=perplexity-launches-low-latency-api-for-open-source-llms) its new, fast and efficient API for accessing open-source large language models like Mistral 7B, Llama2 13B, Code Llama 34B, and Llama2 70B. 4 | 5 | ## What's going on here? 6 | 7 | pplx-api enables blazingly fast LLM inference, up to 2-3x lower latency than competing solutions. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c5154b8c-b17f-489b-88df-6f7daa38446c/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The key benefit is ease of use. Developers can harness powerful language models without needing deep expertise in deploying and managing them. Costs are another big factor. For Perplexity, themselves, migrating one core feature from external APIs to pplx-api cut costs by ~4x for the company. 14 | 15 | The API is proven in production at scale, reliably serving ~1 billion tokens daily. With a roadmap covering custom models, embeddings, and more, pplx-api is positioned to be a go-to solution as demand for industrial-grade LLM inference grows. Data transmitted via the API is automatically deleted after 30 days and Perplexity does not train on user data. 16 | 17 | ## Why should I care? 18 | 19 | With pplx-api, you can easily integrate cutting-edge LLMs into their applications with a familiar REST API, without needing to manage complex model deployment and inference infrastructure. pplx-api abstracts away this complexity through an optimized system leveraging NVIDIA TensortRT-LLM on AWS infrastructure. Low latency and cost savings are factors to look at if you are already using any API from other open source LLM aggregators. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/phind-claims-beats-gpt4.md: -------------------------------------------------------------------------------- 1 | # Phind claims to beats GPT-4 2 | 3 | [Phind has released a new model for code generation](https://www.phind.com/blog/phind-model-beats-gpt4-fast?utm_source=bensbites\&utm_medium=referral\&utm_campaign=phind-claims-to-beats-gpt-4) that matches or exceeds GPT-4 in quality while being much faster. Beyond the benchmarks, Phind claims real users find the Phind Model as helpful as GPT-4 for answering technical coding questions. Phind is a search engine + pair programmer using AI to help developers. 4 | 5 | ## What's going on here? 6 | 7 | Phind's new code generation model beats GPT-4 at coding tasks while running 5x faster at 100 tokens/second. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5a52fc0e-e1ca-47d1-833b-5ba2ae9d1ec3/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The Phind Model V7 achieves a 74.7% accuracy on the HumanEval benchmark. The model is built on top of Phind’s earlier finetunes of CodeLlama-34B. After additional fine-tuning on 70B tokens of code, the Phind Model V7 now provides GPT-4 level responses in 10 seconds rather than 50. This speed boost comes from optimizing the model to run on NVIDIA H100s using the TensorRT-LLM library. 14 | 15 | The Phind Model also supports up to 16k context, allowing for more detailed coding questions. Currently, 12k tokens are enabled for input on the website with 4k reserved for search results. 16 | 17 | ## Why should I care? 18 | 19 | Phind agrees there are still some rough edges, for example on certain challenging questions, the Phind Model might take more generations to get to the right answer than GPT-4. 20 | 21 | Even with those limitations, getting GPT-4 level answers in 5x less time is impressive. Access to a larger context also allows including more code samples and details in your questions. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/purple-llama-meta-evals-models-open-source-safety.md: -------------------------------------------------------------------------------- 1 | # Purple Llama by Meta - Evals and Models for Open Source Safety 2 | 3 | [Meta is announcing Purple Llama](https://ai.meta.com/blog/purple-llama-open-trust-safety-generative-ai/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=purple-llama-by-meta-evals-and-models-for-open-source-safety), an open source project to provide trust and safety tools and evaluations for developing responsible generative AI models. 4 | 5 | ## What's going on here? 6 | 7 | Meta is open-sourcing tools and benchmarks focused on cybersecurity and content safety for generative AI to enable developers to build responsibly. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/4f84e0f4-5fc2-4b00-99b3-c38778928185/image.png?t=1702036142) 10 | 11 | ## What does this mean? 12 | 13 | To start Purple Llama, Meta is releasing CyberSec Eval, a set of cybersecurity benchmarks for evaluating potential risks in language models. You can test your LLM’s tendency to recommend insecure code and comply with malicious requests with CyberSec Eval. 14 | 15 | Additionally, Meta is providing Llama Guard, a content safety classifier to filter risky outputs. It is a pre-trained model to help defend against generating potentially risky outputs. 16 | 17 | ## Why should I care? 18 | 19 | Open-source models are great. At the same time, open-source eval systems are also needed. Purple Llama is an umbrella project for such efforts. Even if you want to write your own evals, having a base set to rely on is great. The best way to ensure people follow safety standards for their deployments is by making it easier to do so. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/quora-raises-new-funding-ai-chatbot-platform-poe.md: -------------------------------------------------------------------------------- 1 | # Quora raises new funding for its AI chatbot platform Poe. 2 | 3 | [Quora has raised $75 million in funding](https://quorablog.quora.com/New-Funding-from-Andreessen-Horowitz?utm_source=bensbites\&utm_medium=referral\&utm_campaign=quora-raises-new-funding-for-its-ai-chatbot-platform-poe) from Andreessen Horowitz's growth fund. The funding will accelerate Poe's growth and pay bot creators on Poe through its creator monetization program. 4 | 5 | ## What’s going on here? 6 | 7 | Quora raises new funding for its AI chatbot platform Poe. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/11006d80-59f4-4228-8fd3-90c567c1e2d2/image.png?t=1704892755) 10 | 11 | ## What does that mean? 12 | 13 | Quora is now valued at $500M, much lower than its peak valuation of $2B+ in 2021. But excluding Poe, Quora is cash flow positive, so all of this new funding will be used on Poe. 14 | 15 | Exactly where? Adam says on the creator monetization program for its program where developers can [create bots on Poe.](http://developer.poe.com/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=quora-raises-new-funding-for-its-ai-chatbot-platform-poe) 16 | 17 | ## Why should I care? 18 | 19 | OpenAI’s GPT store is launching this week. That also aims to offer creator monetization. Poe also provides models other than OpenAI’s GPTs for developers to use. This feels like the browser wars of the previous era, it’ll be interesting to see who gets the attention from different types of consumers (developers, non-technical users, etc.) 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/replit-ai-everyone.md: -------------------------------------------------------------------------------- 1 | # Replit AI is for everyone. 2 | 3 | [Replit is bringing the AI treats](https://blog.replit.com/ai4all?utm_source=bensbites\&utm_medium=referral\&utm_campaign=replit-ai-is-for-everyone) early this Halloween. AI features like code completion and assistance will now be available for all users on their platform, not just paid Pro accounts. 4 | 5 | ## What’s going on here? 6 | 7 | Replit is releasing its AI capabilities to all its 23M+ users for free. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/19d898ad-2b2d-4c47-8b58-4cc5c22a51d0/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The common coder just hit the jackpot baby! Chaps on the free plan will have access to our basic AI features. Replit has also trained a new model, replit-code-v1.5-3b to power these new features in this massive rollout. This 3B parameter model is optimized for programming in 30 languages. Replit is claiming SOTA performance with this model among the 3B parameter models. 14 | 15 | For the Pro users, it’s business as usual: they will get exclusive access to the most powerful AI models(which Amjad has said is GPT-4 right now) and advanced features. 16 | 17 | ## Why should I care? 18 | 19 | Replit is letting go of the Ghostwriter brand to make AI and Replit synonymous. It is treating AI completion as autocorrect in a docs editor. Sure the free version may not be as juiced up, but it's a start. The native experience of coding will AI change how most developers create software and Replit be hookin' everybody up! 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/responsible-innovation-lab-one-ai-protocol.md: -------------------------------------------------------------------------------- 1 | # Responsible Innovation Lab - One more AI protocol 2 | 3 | Investor Hemant Taneja (General Catalyst) has founded [Responsible Innovation Labs](https://www.rilabs.org/news/responsible-ai-protocol-launch?utm_source=bensbites\&utm_medium=referral\&utm_campaign=responsible-innovation-lab-one-more-ai-protocol) with support from 35 VC firms with a protocol for responsible AI development. Though, Hemant and the firms on the list are now getting roasted left and right. 4 | 5 | ## What’s going on here? 6 | 7 | A new non-profit has popped up to create the “framework for regulating AI”. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/8bf9aff5-1fff-4ef2-b170-41d176f67310/image.png) 10 | 11 | ## What does that mean? 12 | 13 | The introductory protocol put forward by RIL has 5 key action steps: 14 | 15 | - Secure organizational buy-in on responsible AI 16 | 17 | - Foster trust through transparency 18 | 19 | - Forecast AI risks & benefits 20 | 21 | - Audit and test to ensure product safety 22 | 23 | - Make regular and ongoing improvements 24 | 25 | ## Why should I care? 26 | 27 | Let’s get away from the X drama about “Oh! the list of VCs to not raise from has dropped” or “It’s a play to get in the good graces of old geezers”. That might be true, but as [Balaji points](https://twitter.com/balajis/status/1724620547584360572?utm_source=bensbites\&utm_medium=referral\&utm_campaign=responsible-innovation-lab-one-more-ai-protocol) the firms and Hemant are all respected, doing great work across different things. The primary problem is with the same limiting approach to testing and auditing, vague definitions and the general lack of tech-optimism. 28 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/rewind-pendant-new-ai-wearable-town.md: -------------------------------------------------------------------------------- 1 | # Rewind Pendant - The new AI wearable in town. 2 | 3 | Rewind, the personalized AI app that remembers everything you do on your computer released a waitlist for their [wearable device called Rewind Pendant.](https://www.rewind.ai/pendant?utm_source=bensbites\&utm_medium=referral\&utm_campaign=rewind-pendant-the-new-ai-wearable-in-town)Rewind Pendant is the hardware extension of “remember everything you do.” This comes following a number of AI wearable previews last week. 4 | 5 | ## What’s going on here? 6 | 7 | Rewind Pendant is a wearable that captures what you say and hear in the real world and then transcribes, encrypts, and stores it entirely locally on your phone. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/39fcf499-96cb-442c-a2a9-b5fae2c92a24/image.png) 10 | 11 | ## What does this mean? 12 | 13 | The Rewind Pendant allows the AI app to learn from a user's real world conversations and sounds, beyond what they experience on screens. There’s no demo or launch date but you can pre-order for $59 and secure your spot. 14 | 15 | ## Why should I care? 16 | 17 | In the past week we have seen three wearables: Meta-Rayban Smart Glasses, [Humane’s Ai Pin](https://bensbites.beehiiv.com/p/humane-reveals-ai-pin-paris-fashion-week) and Avi Schiffman’s Tab (also in a locket form). Then we have rumours about Open AI building an AI device with ex-iPhone designer. There are clear signs that the next path for AI integration is physical devices. AI hardware is heating up. 18 | 19 | With this heat, there are concerns about privacy and wiretapping laws that these companies still need to address. Someone joked on X: who’s gonna build a detector for these AI recording devices now? 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/samsung-unveils-llm-samsung-gauss.md: -------------------------------------------------------------------------------- 1 | # Samsung unveils its LLM: Samsung Gauss 2 | 3 | [Samsung recently unveiled its new generative AI system called Samsung Gauss](https://www.zdnet.com/article/samsung-unveils-its-generative-ai-model-samsung-gauss/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=samsung-unveils-its-llm-samsung-gauss), which consists of language, code, and image models. Gauss currently helps employees via productivity features but will expand to consumer products. 4 | 5 | ## What's going on here? 6 | 7 | Samsung Gauss represents Samsung's attempt to leverage AI to enhance its products and services. While still in early stages, Samsung plans to integrate Gauss across various offerings. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/71fed0d5-1409-40de-93ed-6b821195a5db/image.png) 10 | 11 | ## What does this mean? 12 | 13 | By launching Gauss, Samsung is signaling its serious commitment to AI. Given its massive consumer reach, Samsung integrating Gauss across its ecosystem could significantly accelerate everyday people's adoption of generative AI. 14 | 15 | Samsung seems to be taking a measured approach by having the AI models strengthen its own operations before consumer rollout. This allows time to address risks like security and data privacy. 16 | 17 | ## Why should I care? 18 | 19 | For consumers, Samsung launching Gauss could mean AI assistants in your next Samsung phone or fridge. It also represents tech giants embracing AI, which will likely impact services and products industry-wide. 20 | 21 | Broader adoption of AI has huge implications. It could enhance convenience but also disrupt industries and raise concerns around misinformation. As a tech leader, how Samsung navigates the AI landscape will set an important precedent. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/scale-ais-new-research-lab.md: -------------------------------------------------------------------------------- 1 | # Scale AI's new research lab 2 | 3 | Scale AI is unveiling a new research initiative called the [Safety, Evaluations, and Analysis Lab](https://scale.com/blog/safety-evaluations-analysis-lab?utm_source=bensbites\&utm_medium=referral\&utm_campaign=scale-ai-s-new-research-lab) (SEAL) to establish safety benchmarks and evaluation products for deploying large language models (LLMs). 4 | 5 | ## What’s going on here? 6 | 7 | Scale is ramping up investments in advanced red teaming and evaluation methods to enhance transparency and standardization around LLM safety. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ec9f4e85-3db6-4e30-8606-b45dc36f572c/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Currently, each AI company establishes safety guidelines in-house, which can be inefficient and overlook key risks. Scale plans to collaborate with regulators and the AI community to develop comprehensive LLM safety evaluation products and benchmarks applicable across the industry. This aims to mitigate common safety risks outlined in the executive order on AI, like cybersecurity and deceptive content. The SEAL team will conduct research to improve evaluation reliability, apply red teaming techniques, and develop LLM-based automated rating systems. 14 | 15 | ## Why should I care? 16 | 17 | Standardized safety benchmarks will increase accountability and transparency for companies deploying LLMs. More rigorous evaluation methods can help identify and mitigate risks early. As AI becomes deeply integrated into products and services, consumers need assurance LLMs align with ethical principles. More open collaboration on safety practices will also accelerate AI progress responsibly. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/scale-collaborates-nvidia-nemo-steerlm.md: -------------------------------------------------------------------------------- 1 | # Scale collaborates with NVIDIA with NeMo SteerLM 2 | 3 | [Scale AI is collaborating with NVIDIA](https://scale.com/blog/nvidia-steerlm?utm_source=bensbites\&utm_medium=referral\&utm_campaign=scale-collaborates-with-nvidia-with-nemo-steerlm) to create the next generation of generative AI models using NVIDIA's NeMo SteerLM technique and Scale's high-quality training datasets. 4 | 5 | ## What's going on here? 6 | 7 | SteerLLM from Nvidia and Scale AI will help developers make apps more aligned towards their outcomes. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/3ada2cc6-b0c3-4f65-8ef8-f2588e422d15/image.png?t=1701171939) 10 | 11 | ## What does this mean? 12 | 13 | SteerLM comes with an open-source evaluation dataset with 37k samples and different dimensions like helpfulness, correctness, complexity and verbosity. 14 | 15 | SteerLM allows developers to dynamically customize model behaviour through easy-to-adjust attributes instead of full retraining. For example, in education applications, SteerLM can tailor model complexity and verbosity to individual learning needs. In gaming, SteerLM enables shaping NPC personality and emotional range. 16 | 17 | ## Why should I care? 18 | 19 | Making LLMs perform well in industry-specific scenarios is the target for many providers at the moment. Two of the major roadblocks in achieving that are a) access to industry-specific, custom-labelled, high-quality data and b) readymade frameworks with options to crank up/down certain behaviours. 20 | 21 | Scale and Nvidia’s partnership is one solution for both of these. I’d bet on getting more of such options to come out of LLM fine-tuning/deployment companies in the next months. 22 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/secret-talks-among-us-companies-chinese-experts-ai-safety.md: -------------------------------------------------------------------------------- 1 | # Secret talks among US companies and Chinese experts on AI safety 2 | 3 | Some big US AI companies like OpenAI, Anthropic and Cohere [met up secretly](https://www.ft.com/content/f87b693f-9ba3-4929-8b95-a296b0278021?utm_source=bensbites\&utm_medium=referral\&utm_campaign=secret-talks-among-us-companies-and-chinese-experts-on-ai-safety) with Chinese government AI experts in Geneva last year to chat about keeping AI safe and avoiding harms. This rare team-up comes while the US and China are competing hard in cutting-edge tech like AI. 4 | 5 | ## What’s going on here? 6 | 7 | Some US and Chinese stakeholders put aside their tech race to quietly work together on AI safety. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ddb641e8-b838-4dc3-9ed6-a7651ac73b5c/image.png?t=1705056778) 10 | 11 | ## What does this mean? 12 | 13 | These previously secret meetings in July and October 2022 were among American AI leaders and big Chinese institutions like Tsinghua University. Their goal was to find common ground on AI safety and standards to ensure AI develops in a safe, ethical way. 14 | 15 | White House, UK and Chinese government officials knew about the organization of these Geneva meetings. Shaikh Group which helps with talks in regions of conflict mediated these talks. Chinese AI companies like ByteDance, Tencent, Baidu, and Google DeepMind didn't participate, although DeepMind was briefed. 16 | 17 | ## Why should I care? 18 | 19 | Keeping aside the future innovations, AI as we have it now is going to impact geopolitics and international narratives massively. Among that, it’s nice to have key stakeholders discuss how they can deal with this stuff and maybe, not blow it all up. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/sequoia-act-two-generative-ai.md: -------------------------------------------------------------------------------- 1 | # Sequoia: Act two of Generative AI 2 | 3 | [Sequoia recaps the hype cycle of generative AI](https://www.sequoiacap.com/article/generative-ai-act-two/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sequoia-act-two-of-generative-ai) in the past year and reflects on lessons learned. They argue **we are entering "Act 2"** where companies must **solve real customer problems**, not just showcase cool tech demos. 4 | 5 | ## What's going on here? 6 | 7 | After an initial explosion of hype and usage, generative AI companies are struggling with poor retention. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/97fe69fa-8f5b-4a82-a5d7-a47f507d4f3c/image.png) 10 | 11 | [Credits: Sequoia Capital](https://www.sequoiacap.com/article/generative-ai-act-two?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sequoia-act-two-of-generative-ai) 12 | 13 | ## What does this mean? 14 | 15 | The post argues the main challenge now is proving enduring value to users, not finding demand. Early generative AI products wowed people as novel demos, but they don't yet solve whole problems well enough for most people to use them daily. However, techniques like fine-tuning models on custom data are emerging to close the "expectations vs. reality gap." 16 | 17 | ## Why should I care? 18 | 19 | This is a pivotal time for generative AI. The tech has captured people's imagination, but it needs to mature from flashy demos into indispensable tools. As both an AI builder and user, I should care because this period will determine whether generative AI fizzles out as a fad or becomes as foundational as mobile and cloud computing. The playbook for creating lasting value with AI is still being written, so it's an exciting time to get involved. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/sequoia-implications-gpu-overinvestment.md: -------------------------------------------------------------------------------- 1 | # Sequoia: Implications of GPU overinvestment 2 | 3 | This article analyses the implications of the massive increase in AI model training driven by [surging demand for GPUs.](https://www.sequoiacap.com/article/follow-the-gpus-perspective/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sequoia-implications-of-gpu-overinvestment) The GPU and data center overbuilding will likely \*\*waste capital in the short term but lower costs long-term,\*\*enabling more experimentation. However, startups must shift focus from infrastructure to concrete use cases. 4 | 5 | ## What's going on here? 6 | 7 | There is a disconnect between the AI hype fueling infrastructure overbuilding and the actual end-customer value creation needed to justify it. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c5ab5d5d-78d9-443c-8e48-5ca1954dc2ed/image.png) 10 | 11 | [Credits: Sequoia Capital](https://www.sequoiacap.com/article/follow-the-gpus-perspective/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=sequoia-implications-of-gpu-overinvestment) 12 | 13 | ## What does this mean? 14 | 15 | Based on a conservative estimate of Nvidia guiding $50B in GPU revenue, $100B will be spent on data centres annually. For investors to earn a return, $200B in value must be created by AI products leveraging these GPUs. But current known revenue from major tech companies utilizing AI is only around $75B, leaving a $125B+ gap. 16 | 17 | ## Why should I care? 18 | 19 | The infrastructure will most likely get cheap with time. Startups need to understand the implications of GPU overinvestment versus end-user value creation and find a balance that works for them and investors. One path forward is to identify specific customer pain points and build narrowly focused AI solutions, rather than building general-purpose LLMs. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/snap-releases-lens-studio-50-chatgpt-integration.md: -------------------------------------------------------------------------------- 1 | # Snap releases Lens Studio 5.0 with ChatGPT integration. 2 | 3 | [Snap has released Lens Studio 5.0](https://techcrunch.com/2023/11/09/snaps-latest-version-of-its-ar-development-tool-includes-a-chatgpt-api-boosted-productivity-and-more/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=snap-releases-lens-studio-5-0-with-chatgpt-integration), a major update to its augmented reality development platform. The rewrite aims to help creators build lenses faster and work better together. 4 | 5 | ## What’s going on here? 6 | 7 | Snap has released a new rewrite of its AR studio with AI integrations. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/95305d8b-28fc-40ea-b714-1f77222c1149/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Lens Studio 5.0 brings big performance improvements, cutting project load times by 18x. There’re new features like AI-powered texture generation along with OpenAI's ChatGPT integration, and PBR materials from Meshy. The update also adds workflow optimizations like pinning inspectors, opening multiple projects, and Git version control support. The move is to modernize Lens Studio's dated architecture. 14 | 15 | ## Why should I care? 16 | 17 | Lens Studio 5.0's speed and workflow improvements will let you iterate and ship faster. The expanded toolset through AI integrations provides new options to easily generate assets and interactivity. Built-in support for Git and other collaboration features might be useful for teams to reduce friction when working together. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/spotify-translates-podcasts-new-languages.md: -------------------------------------------------------------------------------- 1 | # Spotify translates podcasts into new languages. 2 | 3 | Spotify is launching [Voice Translations for podcasts.](https://newsroom.spotify.com/2023-09-25/ai-voice-translation-pilot-lex-fridman-dax-shepard-steven-bartlett/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=spotify-translates-podcasts-into-new-languages) It’s a new pilot program for AI-powered voice translations of podcasts into different languages **while retaining the original speaker's voice.** Lex Fridman, Steven Bartlett, and a few other podcasters are part of this pilot program. 4 | 5 | ## What's going on here? 6 | 7 | Spotify is launching a pilot program to translate podcast episodes into different languages using AI to mimic the original speaker's voice. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5e61aa5e-61da-4560-9119-3bfe3eb72339/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Spotify’s new Voice Translation tool uses OpenAI's new voice generation capabilities, which can translate podcast episodes into other languages while matching the speaking style and voice of the original podcaster. This allows podcasts to be available in more languages globally while retaining the speaker's unique voice and authenticity. 14 | 15 | ## Why should I care? 16 | 17 | This technology helps break down language barriers for podcast listening, allowing more people around the world to discover podcasts in their native language. As a podcast listener, this means you may be able to listen to your favorite shows in multiple languages, making podcasts more accessible across different cultures. For podcast creators, this expands their potential reach and audience. Overall, AI voice translation can help facilitate deeper connections between listeners and creators worldwide. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/taking-week-off-exploding-ai-trends.md: -------------------------------------------------------------------------------- 1 | # Taking the week off + some exploding AI trends 2 | 3 | ![Author](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/user/profile_picture/fc858b4d-39e3-4be1-abf4-2b55504e21a2/thumb_uJ4UYake_400x400.jpg) 4 | 5 | [Ben Tossell](https://bensbites.beehiiv.com/authors/fc858b4d-39e3-4be1-abf4-2b55504e21a2)\ 6 | December 28, 2023 7 | 8 | 👋Hey, this is Ben with a**🔒**subscriber-only issue**🔒**of Ben’s Bites Pro.A weekly newsletter covering AI trends, ideas, business breakdowns and how companies are using it internally. 9 | 10 | Happy holidays and happy new year - we’ll be back next week with a new Pro-only post. 11 | 12 | I’ve spent the Christmas period planning content for 2024 and have some interesting conversations that will be covered here. 13 | 14 | But instead of nothing this week, here are some exploding AI trends that I expect to carry on into 2024. 15 | 16 | ## Subscribe to Ben's Bites Pro to read the rest. 17 | 18 | Become a paying subscriber of Ben's Bites Pro to get access to this post and other subscriber-only content. 19 | 20 | [Upgrade](https://bensbites.beehiiv.com/upgrade) 21 | 22 | Already a paying subscriber? [Sign In](https://bensbites.beehiiv.com/login) 23 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/uk-government-announces-ai-advisory-service.md: -------------------------------------------------------------------------------- 1 | # UK government announces AI advisory service 2 | 3 | The UK government announced a [new advisory service to help businesses](https://www.gov.uk/government/news/new-advisory-service-to-help-businesses-launch-ai-and-digital-innovations?utm_source=bensbites\&utm_medium=referral\&utm_campaign=uk-government-announces-ai-advisory-service) quickly and safely bring AI and digital innovations to market. Backed by £2 million in funding, it will run as a pilot for a year starting next year. Firms can apply to get support on specific AI/digital proposals. If successful, regulators will advise them on compliance. 4 | 5 | ## What's going on here? 6 | 7 | The UK is launching a streamlined regulatory advisory service so companies can accelerate compliant AI innovation. 8 | 9 | ## What does this mean? 10 | 11 | The service will provide tailored guidance so companies can meet regulations when launching new digital and AI products. It brings together regulators like the ICO and FCA to give coordinated advice across sectors. This helps businesses navigate complex regulatory landscapes to build compliant innovations faster. 12 | 13 | ## Why should I care? 14 | 15 | The advisory service removes friction. Rather than getting bogged down in regulatory complexity, UK startups can focus on building safely with regulators' input upfront. The quicker path to market can allow AI companies to maximize their competitive advantage. They can lead in rolling out next-gen compliant innovations before others navigate regulatory red tape. 16 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/universal-music-sues-anthropic.md: -------------------------------------------------------------------------------- 1 | # Universal Music sues Anthropic. 2 | 3 | Universal, Concord and ABKCO have filed a [copyright lawsuit against AI startup Anthropic](https://www.ft.com/content/0965d962-5c54-4fdc-aef8-18e4ef3b9df5?utm_source=bensbites\&utm_medium=referral\&utm_campaign=universal-music-sues-anthropic), alleging that Anthropic's chatbot Claude scrapes and regurgitates the studios’ lyrics without permission. 4 | 5 | ## What’s going on here? 6 | 7 | Universal Music group sues Anthropic for generating identical song lyrics. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/d3f0ac53-d53c-4556-ab35-67dcd5bc5f7d/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Universal alleges Anthropic's Claude chatbot copies lyrics word-for-word when prompted, violating copyrights. Anthropic has not properly licensed or attempted to license the lyrics, Universal claims. 14 | 15 | Universal is separately working with Google to license its catalogue for AI songs and partnered with BandLab for ethical AI use of its copyrights. It has also asked Spotify to cut off AI developers' access to its catalogue. 16 | 17 | ## Why should I care? 18 | 19 | This reminds the industry of the Napster days. But much of the perspective isn’t anti-AI, it’s more anti-piracy. The other deals and the claims in the lawsuit hint that the group just want to be paid for the license. But how much is the big question, I guess. 20 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/youtube-wants-use-ai-creating-videos.md: -------------------------------------------------------------------------------- 1 | # YouTube wants you to use AI in creating videos 2 | 3 | [YouTube announced new AI-powered tools](https://blog.youtube/news-and-events/made-on-youtube-2023?utm_source=bensbites\&utm_medium=referral\&utm_campaign=youtube-wants-you-to-use-ai-in-creating-videos) at Made On YouTube to make video creation and discovery easier, including**AI-generated backgrounds**, an app to **simplify editing,** and **AI insights** to get video ideas. 4 | 5 | ## What's going on here? 6 | 7 | YouTube is launching multiple AI features to simplify and automate video creation so anyone can easily produce content. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/253047f0-5e8a-4bc5-8d16-daa577bd76b5/image.png) 10 | 11 | ## What does this mean? 12 | 13 | YouTube will begin testing Dream Screen later this year, allowing users to generate AI video or image backgrounds by typing prompts. A new mobile editing app called YouTube Create is in beta and offers easy editing tools and effects. Starting next year, YouTube Studio will use AI insights to suggest personalized video ideas, help find music faster, and dub your videos in other languages using Aloud. 14 | 15 | ## Why should I care? 16 | 17 | YouTube is the first major social platform to heavily invest in AI creation tools, staying ahead of competitors. While startups have launched similar AI features, YouTube is uniquely positioned to integrate AI natively while others play catchup. This investment by Youtube shows AI-assisted content creation is going mainstream soon. The new mobile editor will also compete with TikTok’s CapCut which many creators swear by. 18 | -------------------------------------------------------------------------------- /fixtures/bensbites.beehiiv.com/zapconnect-2023-building-ai-easier-zapier.md: -------------------------------------------------------------------------------- 1 | # ZapConnect 2023 - Building with AI is easier with Zapier. 2 | 3 | [Zapier launched new AI features at ZapConnect](https://zapier.com/blog/zapconnect-product-updates-2023/?utm_source=bensbites\&utm_medium=referral\&utm_campaign=zapconnect-2023-building-with-ai-is-easier-with-zapier) to help businesses solve challenges and build custom solutions. The highlight is their new product: Zapier Canvas - A new AI-powered diagramming tool. 4 | 5 | ## What's going on here? 6 | 7 | Zapier is announcing new AI and automation features to help users build one-of-a-kind solutions. 8 | 9 | ![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/2dc7a646-1f43-4ecf-bf46-03da60da3234/image.png) 10 | 11 | ## What does this mean? 12 | 13 | Zapier redesigned their editor with drag-and-drop and zoom features for easier workflow building. New AI suggestions map fields automatically and generate notes. Users can now edit test records and test in bulk. 14 | 15 | They launched the database tool Tables and App Builder Interfaces to create full solutions. AI Chatbot lets users build customized bots, and Canvas uses AI to diagram workflows visually. The focus is making automation more visual, collaborative, and AI-powered. 16 | 17 | ## Why should I care? 18 | 19 | These updates can help your business automate routine tasks, gain insights from data, and scale processes without adding headcount - freeing up time and resources to focus on strategic work. 20 | 21 | Much of the talk about the business use case of AI feels you need to be too tech-savvy. These new features from Zapier can help those businesses automate routine processes, gain insights from data, and scale operations without adding headcount, potentially freeing up resources and time for strategic work that drives growth. 22 | -------------------------------------------------------------------------------- /license: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2023 Travis Fischer 4 | 5 | Permission is hereby granted, free of charge, to any person obtaining a copy 6 | of this software and associated documentation files (the "Software"), to deal 7 | in the Software without restriction, including without limitation the rights 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 | copies of the Software, and to permit persons to whom the Software is 10 | furnished to do so, subject to the following conditions: 11 | 12 | The above copyright notice and this permission notice shall be included in all 13 | copies or substantial portions of the Software. 14 | 15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 21 | SOFTWARE. 22 | -------------------------------------------------------------------------------- /next.config.mjs: -------------------------------------------------------------------------------- 1 | /** @type {import('next').NextConfig} */ 2 | const nextConfig = { 3 | reactStrictMode: false 4 | } 5 | 6 | export default nextConfig 7 | -------------------------------------------------------------------------------- /postcss.config.cjs: -------------------------------------------------------------------------------- 1 | module.exports = { 2 | plugins: { 3 | 'postcss-import': {}, 4 | tailwindcss: {}, 5 | autoprefixer: {} 6 | } 7 | } 8 | -------------------------------------------------------------------------------- /public/favicon.ico: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/transitive-bullshit/bens-bites-ai-search/dbbacc69b958751305c92a54453d232a62ee2e29/public/favicon.ico -------------------------------------------------------------------------------- /public/icon.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/transitive-bullshit/bens-bites-ai-search/dbbacc69b958751305c92a54453d232a62ee2e29/public/icon.png -------------------------------------------------------------------------------- /public/logo-light.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/transitive-bullshit/bens-bites-ai-search/dbbacc69b958751305c92a54453d232a62ee2e29/public/logo-light.png -------------------------------------------------------------------------------- /public/social.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/transitive-bullshit/bens-bites-ai-search/dbbacc69b958751305c92a54453d232a62ee2e29/public/social.jpg -------------------------------------------------------------------------------- /public/social.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/transitive-bullshit/bens-bites-ai-search/dbbacc69b958751305c92a54453d232a62ee2e29/public/social.png -------------------------------------------------------------------------------- /src/bin/export-newsletter-links-csv.ts: -------------------------------------------------------------------------------- 1 | import * as fs from 'node:fs/promises' 2 | 3 | import papaparse from 'papaparse' 4 | 5 | import * as config from '@/server/config' 6 | import * as types from '@/server/types' 7 | import * as utils from '@/server/utils' 8 | 9 | async function main() { 10 | const urls: types.NewsletterLink[] = await utils.readJson( 11 | config.newsletterLinksPath 12 | ) 13 | 14 | console.log( 15 | `writing ${urls.length} links to ${config.newsletterLinksCSVPath}\n` 16 | ) 17 | 18 | await fs.writeFile( 19 | config.newsletterLinksCSVPath, 20 | papaparse.unparse(urls, { 21 | columns: [ 22 | 'linkText', 23 | 'url', 24 | 'title', 25 | 'site', 26 | 'description', 27 | 'author', 28 | 'category', 29 | 'shortlink', 30 | 'canonical', 31 | 'date', 32 | 'author_url', 33 | 'thumbnail', 34 | 'thumbnailWidth', 35 | 'thumbnailHeight', 36 | 'icon', 37 | 'iconWidth', 38 | 'iconHeight', 39 | 'dead', 40 | 'postTitle', 41 | 'postDate', 42 | 'postId', 43 | 'postUrl' 44 | ] 45 | }), 46 | 'utf-8' 47 | ) 48 | } 49 | 50 | main().catch((err) => { 51 | console.error('error', err) 52 | process.exit(1) 53 | }) 54 | -------------------------------------------------------------------------------- /src/bin/scratch.ts: -------------------------------------------------------------------------------- 1 | import * as beehiiv from '@/server/beehiiv' 2 | import * as config from '@/server/config' 3 | import { resolveLink } from '@/server/url-utils' 4 | 5 | async function main() { 6 | const page = await beehiiv.resolveBeeHiivPostContent( 7 | // 'https://www.bensbites.co/p/wordle-prompts', 8 | 'https://bensbites.beehiiv.com/p/open-ai-fights-back-court', 9 | { baseUrl: config.newsletterUrl } 10 | ) 11 | console.log(page) 12 | return 13 | 14 | const l = 15 | 'https://www.google.com/url?q=https://www.google.com/url?q%3Dhttps://www.google.com/url?q%253Dhttps://www.google.com/url?q%25253Dhttps://www.google.com/url?q%2525253Dhttps://www.google.com/url?q%252525253Dhttps://t.co/YV3pzUbYOr%2525252526sa%252525253DD%2525252526source%252525253Deditors%2525252526ust%252525253D1671786460642148%2525252526usg%252525253DAOvVaw2B9pdCd2RP4w2hvWRQDJVp%25252526sa%2525253DD%25252526source%2525253Deditors%25252526ust%2525253D1671786489689699%25252526usg%2525253DAOvVaw0B_LmNeZNLctOGV9nyZS5Y%252526sa%25253DD%252526source%25253Deditors%252526ust%25253D1671786495435250%252526usg%25253DAOvVaw3wmOEGOJ_xgcdTk1tNkV8k%2526sa%253DD%2526source%253Deditors%2526ust%253D1671786500545138%2526usg%253DAOvVaw2zPLkJ-n3MOjRiiFE5fyZ6%26sa%3DD%26source%3Deditors%26ust%3D1671786545920663%26usg%3DAOvVaw2UHsbWeuGYHkaM7judECvm&sa=D&source=editors&ust=1671786550827059&usg=AOvVaw0w7D1CHb4tT3G1iXcptuY5' 16 | const u = await resolveLink(l) 17 | console.log(u) 18 | return 19 | } 20 | 21 | main().catch((err) => { 22 | console.error('error', err) 23 | process.exit(1) 24 | }) 25 | -------------------------------------------------------------------------------- /src/components/Button/Button.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | import cs from 'clsx' 3 | 4 | import styles from './styles.module.css' 5 | 6 | export const Button: React.FC< 7 | { 8 | className?: string 9 | buttonClassName?: string 10 | children: React.ReactNode 11 | isLoading?: boolean 12 | ref?: any 13 | } & React.AnchorHTMLAttributes 14 | > = React.forwardRef(function Button( 15 | { className, buttonClassName, children, style, isLoading, ...buttonProps }, 16 | ref 17 | ) { 18 | return ( 19 |
24 | 25 |
{children}
26 | 27 |
28 | ) 29 | }) 30 | -------------------------------------------------------------------------------- /src/components/Button/styles.module.css: -------------------------------------------------------------------------------- 1 | .buttonWrapper { 2 | position: relative; 3 | } 4 | 5 | .button { 6 | position: relative; 7 | cursor: pointer; 8 | 9 | background: var(--fg-color); 10 | color: var(--bg-color); 11 | 12 | border: 1px solid transparent; 13 | box-shadow: 0 4px 4px 0 #00000010; 14 | transition-property: color, background-color, box-shadow; 15 | transition-duration: 0.15s; 16 | transition-timing-function: ease; 17 | padding: 12px 24px; 18 | line-height: 1.5em; 19 | border-radius: 5px; 20 | max-width: 100%; 21 | font-weight: 400; 22 | font-size: 1rem; 23 | display: flex; 24 | justify-content: center; 25 | align-items: center; 26 | user-select: none; 27 | outline: none; 28 | } 29 | 30 | .buttonContent { 31 | text-overflow: ellipsis; 32 | white-space: nowrap; 33 | overflow: hidden; 34 | display: inline-block; 35 | } 36 | 37 | .button:hover, 38 | .button:focus { 39 | border-color: var(--fg-color); 40 | background-color: var(--bg-color); 41 | color: var(--fg-color); 42 | } 43 | 44 | .button:active { 45 | background-color: var(--bg-color-1); 46 | } 47 | 48 | .buttonWrapper:has(.button:disabled) { 49 | opacity: 0.3; 50 | cursor: not-allowed; 51 | } 52 | 53 | .buttonWrapper:has(.button:disabled) * { 54 | pointer-events: none; 55 | } 56 | -------------------------------------------------------------------------------- /src/components/Footer/Footer.tsx: -------------------------------------------------------------------------------- 1 | 'use client' 2 | 3 | import * as React from 'react' 4 | import cs from 'clsx' 5 | 6 | import * as config from '@/lib/config' 7 | import { Discord, GitHub, Twitter } from '@/icons/index' 8 | 9 | import styles from './styles.module.css' 10 | 11 | export const Footer: React.FC<{ className?: string }> = ({ className }) => { 12 | return ( 13 | 62 | ) 63 | } 64 | -------------------------------------------------------------------------------- /src/components/Footer/styles.module.css: -------------------------------------------------------------------------------- 1 | .footer { 2 | width: 100%; 3 | display: flex; 4 | flex-direction: row; 5 | justify-content: space-between; 6 | align-items: center; 7 | gap: 1rem; 8 | } 9 | 10 | .settings, 11 | .copyright, 12 | .poweredBy, 13 | .social { 14 | flex: 1; 15 | } 16 | 17 | .settings, 18 | .poweredBy { 19 | justify-content: center; 20 | text-align: center; 21 | } 22 | 23 | .social { 24 | justify-content: flex-end; 25 | } 26 | 27 | .copyright, 28 | .poweredBy { 29 | font-size: 0.8rem; 30 | padding: 0.5rem 0; 31 | } 32 | 33 | .settings, 34 | .social { 35 | display: flex; 36 | flex-direction: row; 37 | align-items: center; 38 | gap: 1rem; 39 | user-select: none; 40 | } 41 | 42 | .action { 43 | cursor: pointer; 44 | width: 2rem; 45 | height: 2rem; 46 | display: inline-flex; 47 | padding: 0.25rem; 48 | transition: color 250ms ease-out; 49 | } 50 | 51 | .footer .darkModeToggle { 52 | width: 1.75em; 53 | height: 1.75em; 54 | } 55 | 56 | .action:hover { 57 | transition: color 50ms ease-out; 58 | } 59 | 60 | .twitter:hover { 61 | color: #2795e9; 62 | } 63 | 64 | .github:hover { 65 | color: #c9510c; 66 | } 67 | 68 | .discord:hover { 69 | color: #5766f2; 70 | } 71 | 72 | @media only screen and (max-width: 500px) { 73 | .footer { 74 | flex-direction: column; 75 | } 76 | 77 | .footer .settings { 78 | justify-content: center; 79 | order: 1; 80 | } 81 | 82 | .footer .social { 83 | justify-content: center; 84 | order: 2; 85 | } 86 | 87 | .footer .poweredBy { 88 | justify-content: center; 89 | order: 3; 90 | } 91 | 92 | .footer .copyright { 93 | justify-content: center; 94 | order: 4; 95 | } 96 | } 97 | -------------------------------------------------------------------------------- /src/components/Header/Logo.tsx: -------------------------------------------------------------------------------- 1 | 'use client' 2 | 3 | import * as React from 'react' 4 | import cs from 'clsx' 5 | import Image from 'next/image' 6 | 7 | import LogoLight from '@/public/logo-light.png' 8 | 9 | import styles from './styles.module.css' 10 | 11 | export const Logo: React.FC<{ className?: string }> = ({ className }) => { 12 | return ( 13 | Logo 20 | ) 21 | } 22 | -------------------------------------------------------------------------------- /src/components/HeroButton/HeroButton.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | import cs from 'clsx' 3 | 4 | import styles from './styles.module.css' 5 | 6 | export type HeroButtonVariant = 'orange' | 'blue' | 'purple' 7 | 8 | export const HeroButton: React.FC< 9 | { 10 | variant?: HeroButtonVariant 11 | className?: string 12 | buttonClassName?: string 13 | children: React.ReactNode 14 | } & React.AnchorHTMLAttributes 15 | > = ({ 16 | variant = 'purple', 17 | className, 18 | buttonClassName, 19 | children, 20 | style, 21 | ...buttonProps 22 | }) => { 23 | return ( 24 |
25 | {variant === 'blue' && ( 26 | 27 | )} 28 | 29 | {variant === 'purple' && ( 30 | 31 | )} 32 | 33 | {variant === 'orange' && ( 34 | 35 | )} 36 | 37 | 38 |
{children}
39 | 40 |
41 | ) 42 | } 43 | -------------------------------------------------------------------------------- /src/components/Layout/Layout.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | import cs from 'clsx' 3 | import { Inter } from 'next/font/google' 4 | 5 | import { Footer } from '@/components/Footer/Footer' 6 | import { Header } from '@/components/Header/Header' 7 | 8 | import { RootLayoutProviders } from './providers' 9 | import styles from './styles.module.css' 10 | 11 | const inter = Inter({ subsets: ['latin'] }) 12 | 13 | export function Layout({ children }: { children: React.ReactNode }) { 14 | return ( 15 |
16 | 17 |
18 | 19 |
{children}
20 | 21 |
22 | 23 |
24 | ) 25 | } 26 | -------------------------------------------------------------------------------- /src/components/Layout/providers.tsx: -------------------------------------------------------------------------------- 1 | 'use client' 2 | 3 | import * as React from 'react' 4 | import { MotionConfig } from 'framer-motion' 5 | 6 | export function RootLayoutProviders({ 7 | children 8 | }: { 9 | children: React.ReactNode 10 | }) { 11 | return {children} 12 | } 13 | -------------------------------------------------------------------------------- /src/components/Layout/styles.module.css: -------------------------------------------------------------------------------- 1 | .container { 2 | flex: 1; 3 | display: flex; 4 | flex-direction: column; 5 | justify-content: space-around; 6 | align-items: center; 7 | margin: 0 auto; 8 | max-width: var(--max-width); 9 | min-height: 100vh; 10 | padding: 0; 11 | } 12 | 13 | .header { 14 | padding: 12px 12px 0; 15 | } 16 | 17 | .footer { 18 | padding: 0 12px 12px; 19 | } 20 | 21 | .main { 22 | flex: 1; 23 | width: 100%; 24 | min-height: 100vh; 25 | max-width: var(--max-body-width); 26 | display: flex; 27 | flex-direction: column; 28 | align-items: center; 29 | margin-bottom: var(--gap-h); 30 | padding: 0 12px; 31 | padding-top: 72px; 32 | } 33 | 34 | @media (max-width: 800px) { 35 | .header { 36 | padding: 0; 37 | } 38 | } 39 | -------------------------------------------------------------------------------- /src/components/LoadingSpinner/LoadingSpinner.tsx: -------------------------------------------------------------------------------- 1 | 'use client' 2 | 3 | import * as React from 'react' 4 | import Loader from 'react-spinners/BeatLoader' 5 | 6 | import styles from './styles.module.css' 7 | 8 | export const LoadingSpinner: React.FC<{ loading?: boolean }> = ({ 9 | loading = true 10 | }) => { 11 | return 12 | } 13 | -------------------------------------------------------------------------------- /src/components/LoadingSpinner/styles.module.css: -------------------------------------------------------------------------------- 1 | .loading { 2 | position: absolute; 3 | top: 50%; 4 | left: 50%; 5 | transform: translate(-50%, -50%); 6 | } 7 | -------------------------------------------------------------------------------- /src/components/Markdown/Markdown.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | import cs from 'clsx' 3 | 4 | import styles from './styles.module.css' 5 | 6 | /** 7 | * `content` is assumed to have already been transformed into HTML via remark/rehype. 8 | */ 9 | export const Markdown: React.FC<{ className?: string; content: string }> = ({ 10 | className, 11 | content 12 | }) => { 13 | return ( 14 |
18 | ) 19 | } 20 | -------------------------------------------------------------------------------- /src/components/SearchResults/styles.module.css: -------------------------------------------------------------------------------- 1 | .searchResults { 2 | flex: 1; 3 | width: 100%; 4 | display: flex; 5 | flex-direction: column; 6 | max-width: 800px; 7 | } 8 | 9 | .detail { 10 | flex: 1; 11 | position: relative; 12 | width: 100%; 13 | /* margin: 0 auto var(--gap); */ 14 | } 15 | 16 | .emptyResults { 17 | flex: 1; 18 | display: flex; 19 | flex-direction: column; 20 | justify-content: flex-start; 21 | align-items: center; 22 | } 23 | 24 | .emptyResults p { 25 | margin-bottom: 1em; 26 | } 27 | 28 | .desc { 29 | text-align: center; 30 | } 31 | 32 | .socialImageWrapper { 33 | width: 100%; 34 | flex: 1; 35 | display: flex; 36 | flex-direction: column; 37 | justify-content: center; 38 | align-items: center; 39 | } 40 | 41 | .socialImage { 42 | width: 600px; 43 | max-width: 100%; 44 | border-radius: 8px; 45 | } 46 | -------------------------------------------------------------------------------- /src/components/SearchResultsList/SearchResultsList.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | 3 | import * as types from '@/types' 4 | import { SearchResult } from '@/components/SearchResult/SearchResult' 5 | 6 | import styles from './styles.module.css' 7 | 8 | export const SearchResultsList: React.FC<{ 9 | results: types.SearchResult[] 10 | }> = ({ results }) => { 11 | // console.log( 12 | // 'results', 13 | // results.map((r) => r.metadata) 14 | // ) 15 | 16 | return ( 17 |
18 | {results.map((result) => ( 19 | 20 | ))} 21 |
22 | ) 23 | } 24 | -------------------------------------------------------------------------------- /src/components/SearchResultsList/styles.module.css: -------------------------------------------------------------------------------- 1 | .searchResultsList { 2 | width: 100%; 3 | display: flex; 4 | flex-direction: column; 5 | justify-content: center; 6 | align-items: center; 7 | margin-bottom: var(--gap-h); 8 | gap: 0; 9 | } 10 | 11 | @media (max-width: 700px) { 12 | .searchResultsList { 13 | margin-bottom: 2em; 14 | gap: 3em; 15 | } 16 | } 17 | -------------------------------------------------------------------------------- /src/icons/Clear.tsx: -------------------------------------------------------------------------------- 1 | export const ClearIcon = ({ 2 | width = 20, 3 | height = 20, 4 | className 5 | }: { 6 | width?: number 7 | height?: number 8 | className?: string 9 | }) => ( 10 | 17 | 18 | 19 | ) 20 | -------------------------------------------------------------------------------- /src/icons/External.tsx: -------------------------------------------------------------------------------- 1 | export const External = ({ className }: { className?: string }) => ( 2 | 11 | 12 | 13 | ) 14 | -------------------------------------------------------------------------------- /src/icons/GitHub.tsx: -------------------------------------------------------------------------------- 1 | export const GitHub = ({ className }: { className?: string }) => ( 2 | 3 | 4 | 5 | ) 6 | -------------------------------------------------------------------------------- /src/icons/Globe.tsx: -------------------------------------------------------------------------------- 1 | export const Globe = ({ className }: { className?: string }) => ( 2 | 8 | 9 | 10 | ) 11 | -------------------------------------------------------------------------------- /src/icons/Search.tsx: -------------------------------------------------------------------------------- 1 | export const SearchIcon = ({ className }: { className?: string }) => ( 2 | 9 | 15 | 16 | ) 17 | -------------------------------------------------------------------------------- /src/icons/Twitter.tsx: -------------------------------------------------------------------------------- 1 | export const Twitter = ({ className }: { className?: string }) => ( 2 | 3 | 4 | 5 | ) 6 | -------------------------------------------------------------------------------- /src/icons/index.ts: -------------------------------------------------------------------------------- 1 | export * from './GitHub' 2 | export * from './Discord' 3 | export * from './Twitter' 4 | export * from './Clear' 5 | export * from './Search' 6 | export * from './Globe' 7 | export * from './External' 8 | -------------------------------------------------------------------------------- /src/lib/bootstrap.ts: -------------------------------------------------------------------------------- 1 | import { githubRepoUrl, isSafari, isServer } from './config' 2 | 3 | const detail = `This webapp is open source: ${githubRepoUrl}` 4 | const banner = ` 5 | 6 | ████████╗██████╗ █████╗ ███╗ ██╗███████╗██╗████████╗██╗██╗ ██╗███████╗ ██████╗ ███████╗ 7 | ╚══██╔══╝██╔══██╗██╔══██╗████╗ ██║██╔════╝██║╚══██╔══╝██║██║ ██║██╔════╝ ██╔══██╗██╔════╝ 8 | ██║ ██████╔╝███████║██╔██╗ ██║███████╗██║ ██║ ██║██║ ██║█████╗ ██████╔╝███████╗ 9 | ██║ ██╔══██╗██╔══██║██║╚██╗██║╚════██║██║ ██║ ██║╚██╗ ██╔╝██╔══╝ ██╔══██╗╚════██║ 10 | ██║ ██║ ██║██║ ██║██║ ╚████║███████║██║ ██║ ██║ ╚████╔╝ ███████╗ ██████╔╝███████║ 11 | ╚═╝ ╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═══╝╚══════╝╚═╝ ╚═╝ ╚═╝ ╚═══╝ ╚══════╝ ╚═════╝ ╚══════╝ 12 | 13 | ${detail} 14 | ` 15 | 16 | export async function bootstrap() { 17 | if (isServer) return 18 | 19 | if (isSafari) { 20 | console.log(detail) 21 | } else { 22 | console.log(banner) 23 | } 24 | } 25 | -------------------------------------------------------------------------------- /src/lib/config.ts: -------------------------------------------------------------------------------- 1 | export const environment = process.env.NODE_ENV || 'development' 2 | export const isDev = environment === 'development' 3 | export const isServer = typeof window === 'undefined' 4 | export const isSafari = 5 | !isServer && /^((?!chrome|android).)*safari/i.test(navigator.userAgent) 6 | 7 | export const title = "Ben's Bites AI Search" 8 | export const description = 9 | "Search across all of the best resources in AI. Powered by the Ben's Bites AI Newsletter and updated daily, with over 60,000 subscribers from companies like Google, a16z, Sequoia, Amazon, and Meta." 10 | 11 | export const domain = 'search.bensbites.co' 12 | 13 | export const author = 'Travis Fischer' 14 | export const twitter = 'transitive_bs' 15 | export const twitterUrl = `https://twitter.com/${twitter}` 16 | export const discordUrl = 'https://discord.gg/ben-s-bites-1039224571376701510' 17 | export const newsletterTitle = "Ben's Bites AI Newsletter" 18 | export const newsletterUrl = 'https://www.bensbites.co/' 19 | export const githubRepoUrl = 20 | 'https://github.com/transitive-bullshit/bens-bites-ai-search' 21 | export const githubSponsorsUrl = 22 | 'https://github.com/sponsors/transitive-bullshit' 23 | export const copyright = `Copyright 2023 ${author}` 24 | export const madeWithLove = 'Made with ❤️ in Brooklyn, NY' 25 | 26 | export const port = process.env.PORT || '3000' 27 | export const prodUrl = `https://${domain}` 28 | export const url = isDev ? `http://localhost:${port}` : prodUrl 29 | 30 | export const apiBaseUrl = 31 | isDev || !process.env.NEXT_PUBLIC_VERCEL_URL 32 | ? url 33 | : `https://${process.env.NEXT_PUBLIC_VERCEL_URL}` 34 | 35 | // these must all be absolute urls 36 | export const socialImageUrl = `${prodUrl}/social.png` 37 | -------------------------------------------------------------------------------- /src/lib/markdown-to-html.ts: -------------------------------------------------------------------------------- 1 | import rehypeFormat from 'rehype-format' 2 | import rehypeRaw from 'rehype-raw' 3 | import rehypeStringify from 'rehype-stringify' 4 | import remarkGfm from 'remark-gfm' 5 | import remarkParse from 'remark-parse' 6 | import remarkRehype from 'remark-rehype' 7 | import { unified } from 'unified' 8 | 9 | const processor = unified() 10 | .use(remarkParse) 11 | .use(remarkGfm) 12 | .use(remarkRehype, { 13 | allowDangerousHtml: true 14 | }) 15 | .use(rehypeRaw) 16 | .use(rehypeFormat) 17 | .use(rehypeStringify) 18 | 19 | export async function markdownToHtml(markdown: string) { 20 | const result = await processor.process(markdown) 21 | return result.toString() 22 | } 23 | -------------------------------------------------------------------------------- /src/lib/readme.md: -------------------------------------------------------------------------------- 1 | The `lib/` folder is shared between frontend and backend. 2 | -------------------------------------------------------------------------------- /src/pages/_app.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | import { Analytics } from '@vercel/analytics/react' 3 | import type { AppProps } from 'next/app' 4 | import Head from 'next/head' 5 | 6 | import { bootstrap } from '@/lib/bootstrap' 7 | import { isServer } from '@/lib/config' 8 | import '@/styles/globals.css' 9 | 10 | if (!isServer) { 11 | bootstrap() 12 | } 13 | 14 | export default function App({ 15 | Component, 16 | pageProps: { session, ...pageProps } 17 | }: AppProps) { 18 | return ( 19 | <> 20 | 21 | 22 | 23 | 24 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | ) 35 | } 36 | -------------------------------------------------------------------------------- /src/pages/_document.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | import { Head, Html, Main, NextScript } from 'next/document' 3 | 4 | export default function Document() { 5 | return ( 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 |
14 | 15 | 16 | 17 | 18 | ) 19 | } 20 | -------------------------------------------------------------------------------- /src/pages/about/styles.module.css: -------------------------------------------------------------------------------- 1 | .aboutPage { 2 | flex: 1; 3 | width: 100%; 4 | display: flex; 5 | flex-direction: column; 6 | justify-content: flex-start; 7 | align-items: center; 8 | gap: var(--gap-2); 9 | } 10 | 11 | .meta { 12 | width: 100%; 13 | display: flex; 14 | flex-direction: column; 15 | align-items: center; 16 | margin: 1em auto; 17 | } 18 | 19 | .title { 20 | font-size: 2.5rem; 21 | } 22 | -------------------------------------------------------------------------------- /src/pages/index.module.css: -------------------------------------------------------------------------------- 1 | .page { 2 | flex: 1; 3 | width: 100%; 4 | display: flex; 5 | flex-direction: column; 6 | justify-content: center; 7 | align-items: center; 8 | padding-top: var(--gap-h); 9 | gap: var(--gap-2); 10 | } 11 | 12 | .body { 13 | position: relative; 14 | flex: 1; 15 | width: 100%; 16 | margin: 0 auto; 17 | display: flex; 18 | flex-direction: column; 19 | justify-content: flex-start; 20 | align-items: center; 21 | gap: var(--gap-h); 22 | } 23 | -------------------------------------------------------------------------------- /src/pages/index.tsx: -------------------------------------------------------------------------------- 1 | import * as React from 'react' 2 | 3 | import { Layout } from '@/components/Layout/Layout' 4 | import { PageHead } from '@/components/PageHead/PageHead' 5 | import { SearchOptions } from '@/components/SearchOptions/SearchOptions' 6 | import { SearchResults } from '@/components/SearchResults/SearchResults' 7 | import { Search } from '@/lib/hooks/search' 8 | 9 | import styles from './index.module.css' 10 | 11 | export default function HomePage() { 12 | return ( 13 | 14 | 15 | 16 |
17 |
18 | 19 | 20 | 21 | 22 | 23 |
24 |
25 | 26 | ) 27 | } 28 | -------------------------------------------------------------------------------- /src/server/config.ts: -------------------------------------------------------------------------------- 1 | import path from 'node:path' 2 | 3 | import dotenv from 'dotenv-safe' 4 | import { z } from 'zod' 5 | 6 | dotenv.config() 7 | 8 | const envSchema = z.object({ 9 | // openai 10 | OPENAI_API_KEY: z.string().min(1), 11 | 12 | // pinecone 13 | PINECONE_API_KEY: z.string().min(1), 14 | PINECONE_BASE_URL: z.string().url(), 15 | PINECONE_NAMESPACE: z.string().min(1) 16 | }) 17 | 18 | // ensure that all required env variables are defined 19 | const env = envSchema.safeParse(process.env) 20 | 21 | if (env.success === false) { 22 | console.error( 23 | '❌ Invalid environment variables:', 24 | JSON.stringify(env.error.format(), null, 4) 25 | ) 26 | 27 | process.exit(1) 28 | } 29 | 30 | export const openaiEmbeddingModel = 'text-embedding-ada-002' 31 | export const openaiEmbeddingModelMaxTokens = 8191 32 | export const protocolAllowList = new Set(['https:', 'http:']) 33 | export const domainAllowList = new Set(['flight.beehiiv.net', 't.co']) 34 | 35 | export const newsletterUrl = process.env.BEEHIIV_URL 36 | export const newsletterParsedUrl = new URL(newsletterUrl) 37 | export const newsletterHostname = newsletterParsedUrl.hostname 38 | export const newsletterDir = path.join('fixtures', newsletterHostname) 39 | export const newsletterMetadataPath = path.join( 40 | newsletterDir, 41 | 'newsletter.json' 42 | ) 43 | export const newsletterLinksCSVPath = path.join(newsletterDir, 'links.csv') 44 | export const newsletterLinksPath = path.join(newsletterDir, 'links.json') 45 | export const twitterDataCachePath = path.join(newsletterDir, 'twitter.json') 46 | 47 | export default env 48 | -------------------------------------------------------------------------------- /src/server/embedding.ts: -------------------------------------------------------------------------------- 1 | import pMemoize from 'p-memoize' 2 | import pRetry from 'p-retry' 3 | import pThrottle from 'p-throttle' 4 | 5 | import * as config from './config' 6 | import { openai } from './services/openai' 7 | 8 | // https://platform.openai.com/docs/guides/rate-limits/what-are-the-rate-limits-for-our-api 9 | // TODO: enforce TPM 10 | const throttleRPM = pThrottle({ 11 | // 3k per minute instead of 3.5k per minute to add padding 12 | limit: 3000, 13 | interval: 60 * 1000, 14 | strict: true 15 | }) 16 | 17 | export const createEmbedding = pMemoize(throttleRPM(createEmbeddingImpl)) 18 | 19 | async function createEmbeddingImpl({ 20 | input, 21 | model = config.openaiEmbeddingModel, 22 | maxTokens = config.openaiEmbeddingModelMaxTokens 23 | }: { 24 | input: string 25 | model?: string 26 | maxTokens?: number 27 | }): Promise { 28 | // TODO: enforce max tokens 29 | 30 | const res = await pRetry( 31 | () => 32 | openai.createEmbedding({ 33 | input, 34 | model 35 | }), 36 | { 37 | retries: 4, 38 | minTimeout: 1000, 39 | factor: 2.5 40 | } 41 | ) 42 | 43 | return res.data.data[0].embedding 44 | } 45 | -------------------------------------------------------------------------------- /src/server/got.ts: -------------------------------------------------------------------------------- 1 | import got from 'got' 2 | 3 | const USER_AGENT = 4 | 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/108.0.0.0 Safari/537.36' 5 | 6 | export default got.extend({ 7 | headers: { 8 | 'user-agent': USER_AGENT 9 | } 10 | }) 11 | -------------------------------------------------------------------------------- /src/server/readme.md: -------------------------------------------------------------------------------- 1 | The `server/` folder is restricited to server-side usage only. 2 | -------------------------------------------------------------------------------- /src/server/services/meilisearch.ts: -------------------------------------------------------------------------------- 1 | import { MeiliSearch } from 'meilisearch' 2 | 3 | export const meilisearch = new MeiliSearch({ 4 | host: process.env.MEILISEARCH_URL, 5 | apiKey: process.env.MEILISEARCH_API_KEY 6 | }) 7 | -------------------------------------------------------------------------------- /src/server/services/openai.ts: -------------------------------------------------------------------------------- 1 | import { Configuration, OpenAIApi } from 'openai' 2 | 3 | import '../config' 4 | 5 | export const openai = new OpenAIApi( 6 | new Configuration({ 7 | apiKey: process.env.OPENAI_API_KEY 8 | }) 9 | ) 10 | -------------------------------------------------------------------------------- /src/server/services/storage.ts: -------------------------------------------------------------------------------- 1 | import { Storage } from '@google-cloud/storage' 2 | 3 | import * as types from '@/server/types' 4 | 5 | export interface UploadFileToBucketOptions { 6 | name: string 7 | path: string 8 | storage: types.GCPStorage 9 | bucket: string 10 | contentType?: string 11 | } 12 | 13 | export const storage = new Storage() 14 | 15 | export async function uploadFileToBucket({ 16 | path, 17 | name, 18 | storage, 19 | bucket, 20 | contentType = 'image/jpeg' 21 | }: UploadFileToBucketOptions) { 22 | return storage.bucket(bucket).upload(path, { 23 | destination: name, 24 | contentType 25 | }) 26 | } 27 | -------------------------------------------------------------------------------- /src/server/services/twitter.ts: -------------------------------------------------------------------------------- 1 | import { TwitterApi } from 'twitter-api-v2' 2 | 3 | import '../config' 4 | 5 | const twitterApi = new TwitterApi({ 6 | appKey: process.env.TWITTER_API_KEY, 7 | appSecret: process.env.TWITTER_API_SECRET_KEY, 8 | accessToken: process.env.TWITTER_API_ACCESS_TOKEN, 9 | accessSecret: process.env.TWITTER_API_ACCESS_SECRET 10 | }) 11 | const { v1: twitterV1 } = twitterApi 12 | export { twitterV1 } 13 | -------------------------------------------------------------------------------- /src/server/twitter.test.ts: -------------------------------------------------------------------------------- 1 | import fs from 'node:fs/promises' 2 | 3 | import test from 'ava' 4 | 5 | import { 6 | maxTwitterId, 7 | minTwitterId, 8 | tweetComparator, 9 | tweetIdComparator 10 | } from './twitter' 11 | 12 | test('maxTwitterId', (t) => { 13 | t.is(maxTwitterId('123', '456'), '456') 14 | t.is(maxTwitterId('1230', '999'), '1230') 15 | t.is(maxTwitterId('', '999'), '999') 16 | t.is(maxTwitterId('999', ''), '999') 17 | t.is(maxTwitterId('', null), null) 18 | t.is(maxTwitterId('', undefined), null) 19 | t.is(maxTwitterId('948392', '948392'), '948392') 20 | }) 21 | 22 | test('minTwitterId', (t) => { 23 | t.is(minTwitterId('123', '456'), '123') 24 | t.is(minTwitterId('1230', '999'), '999') 25 | t.is(minTwitterId('', '999'), '999') 26 | t.is(minTwitterId('999', ''), '999') 27 | t.is(minTwitterId('', null), null) 28 | t.is(minTwitterId('', undefined), null) 29 | t.is(minTwitterId('948392', '948392'), '948392') 30 | }) 31 | 32 | test('tweetIdComparator', (t) => { 33 | t.is(tweetIdComparator('100', '200'), -1) 34 | t.is(tweetIdComparator('3000', '999'), 1) 35 | t.is(tweetIdComparator('3001', '3001'), 0) 36 | }) 37 | 38 | test('tweetComparator', (t) => { 39 | t.is(tweetComparator({ id: '100' }, { id: '200' }), -1) 40 | t.is(tweetComparator({ id: '3000' }, { id: '999' }), 1) 41 | t.is(tweetComparator({ id: '3001' }, { id: '3001' }), 0) 42 | 43 | t.deepEqual( 44 | [ 45 | { id: '5' }, 46 | { id: '1000' }, 47 | { id: '9999' }, 48 | { id: '5' }, 49 | { id: '15' }, 50 | { id: '500' } 51 | ].sort(tweetComparator), 52 | [ 53 | { id: '5' }, 54 | { id: '5' }, 55 | { id: '15' }, 56 | { id: '500' }, 57 | { id: '1000' }, 58 | { id: '9999' } 59 | ] 60 | ) 61 | }) 62 | -------------------------------------------------------------------------------- /src/server/unfurl-tweet.test.ts: -------------------------------------------------------------------------------- 1 | import fs from 'node:fs' 2 | 3 | import test from 'ava' 4 | 5 | import { unfurlTweet } from './unfurl-tweet' 6 | 7 | const resolvedTwitterUser = JSON.parse( 8 | fs.readFileSync('fixtures/transitive_bs.json', 'utf8') 9 | ) 10 | 11 | const fixtures = [ 12 | { id: '1625435055706308608', label: 'basic reply' }, 13 | { id: '1603341229047021568', label: 'basic retwet' }, 14 | { id: '1628578673585709056', label: 'media w/ alt text' }, 15 | { id: '1622150427885326337', label: 'media' }, 16 | { id: '1628578692707532800', label: 'media + quote tweet' }, 17 | { id: '1628219587748589568', label: 'url' }, 18 | { id: '1628143012847943680', label: 'quote tweet' }, 19 | { id: '1628118176524533760', label: 'media + url (google drive)' }, 20 | { id: '1628088218364297216', label: 'url + url + media' }, 21 | { id: '1628065270245003264', label: 'url + url' }, 22 | { id: '1627984478743236609', label: 'media (gif)' }, 23 | { id: '1628578702421557248', label: 'recursive (tweet edit)' }, 24 | { id: '1596617311867359232', label: 'error' }, 25 | { id: '1572186724117221376', label: 'error tweet special character' }, 26 | { id: '1376571894493704194', label: 'error tweet special character' }, 27 | { id: '1616716417700892674', label: 'ampersand' } 28 | ] 29 | 30 | for (const fixture of fixtures) { 31 | test(`unfurlTweet ${fixture.id} – ${fixture.label}`, async (t) => { 32 | const tweet = resolvedTwitterUser.tweets[fixture.id] 33 | t.truthy(tweet) 34 | 35 | const text = unfurlTweet(tweet, { resolvedTwitterUser }) 36 | t.truthy(text) 37 | 38 | const expanded = unfurlTweet(tweet, { 39 | resolvedTwitterUser, 40 | unfurlUrls: true 41 | }) 42 | t.truthy(expanded) 43 | 44 | console.log({ 45 | id: fixture.id, 46 | label: fixture.label, 47 | full_text: tweet.full_text, 48 | text, 49 | expanded 50 | }) 51 | // t.snapshot(text) 52 | }) 53 | } 54 | -------------------------------------------------------------------------------- /src/server/utils.ts: -------------------------------------------------------------------------------- 1 | import crypto from 'node:crypto' 2 | import fs from 'node:fs/promises' 3 | 4 | import * as types from './types' 5 | 6 | export async function readJson(filePath: string): Promise { 7 | return JSON.parse(await fs.readFile(filePath, 'utf-8')) 8 | } 9 | 10 | export async function writeJson(filePath: string, json: T) { 11 | return fs.writeFile(filePath, JSON.stringify(json, null, 2), 'utf-8') 12 | } 13 | 14 | export function pick(obj: T, ...keys: string[]): U { 15 | return Object.fromEntries( 16 | keys.filter((key) => key in obj).map((key) => [key, obj[key]]) 17 | ) as U 18 | } 19 | 20 | export function omit(obj: T, ...keys: string[]): U { 21 | return Object.fromEntries( 22 | Object.entries(obj).filter(([key]) => !keys.includes(key)) 23 | ) as U 24 | } 25 | 26 | export function hash(d: Buffer | string): string { 27 | const buffer = Buffer.isBuffer(d) ? d : Buffer.from(d.toString()) 28 | return crypto.createHash('sha256').update(buffer).digest('hex') 29 | } 30 | 31 | export function sanitizePineconeString(input: string): string { 32 | if (!input) return '' 33 | return input.replace(/[\ud800-\udfff]/g, '').trim() 34 | } 35 | 36 | export function getNewsletterLinkId(link: types.NewsletterLink): string { 37 | return `${link.postId}-${hash(link.url).slice(0, 16)}` 38 | } 39 | -------------------------------------------------------------------------------- /src/types/index.ts: -------------------------------------------------------------------------------- 1 | import { z } from 'zod' 2 | 3 | import type { PineconeMetadata } from '@/server/types' 4 | 5 | export { PineconeMetadata } 6 | export type SearchResult = PineconeMetadata & { 7 | id: string 8 | } 9 | 10 | export const SearchQuerySchema = z.object({ 11 | query: z.string(), 12 | limit: z.coerce.number().nonnegative().int().min(1).max(100).optional() 13 | }) 14 | 15 | export const SearchOptionsSchema = z.object({ 16 | orderBy: z.string().optional() 17 | }) 18 | 19 | export type SearchQuery = z.infer 20 | export type ISearchOptions = z.infer 21 | -------------------------------------------------------------------------------- /tailwind.config.cjs: -------------------------------------------------------------------------------- 1 | const { fontFamily } = require('tailwindcss/defaultTheme') 2 | 3 | /** @type {import('tailwindcss').Config} */ 4 | module.exports = { 5 | darkMode: 'class', 6 | content: [ 7 | './src/pages/**/*.{js,ts,jsx,tsx}', 8 | './src/components/**/*.{js,ts,jsx,tsx}' 9 | ], 10 | theme: { 11 | extend: { 12 | fontFamily: { 13 | sans: ['system-ui', ...fontFamily.sans] 14 | }, 15 | colors: { 16 | fg: { 17 | 0: 'rgb(var(--color-fg-0) / )', 18 | 1: 'rgb(var(--color-fg-1) / )', 19 | 2: 'rgb(var(--color-fg-2) / )', 20 | 3: 'rgb(var(--color-fg-3) / )', 21 | 4: 'rgb(var(--color-fg-4) / )' 22 | }, 23 | bg: { 24 | 0: 'rgb(var(--color-bg-0) / )', 25 | 1: 'rgb(var(--color-bg-1) / )', 26 | 2: 'rgb(var(--color-bg-2) / )', 27 | 3: 'rgb(var(--color-bg-3) / )', 28 | 4: 'rgb(var(--color-bg-4) / )' 29 | } 30 | } 31 | } 32 | }, 33 | plugins: [require('@tailwindcss/typography')] 34 | } 35 | -------------------------------------------------------------------------------- /tsconfig.json: -------------------------------------------------------------------------------- 1 | { 2 | "compilerOptions": { 3 | "target": "es2016", 4 | "lib": ["dom", "dom.iterable", "esnext"], 5 | "allowJs": true, 6 | "skipLibCheck": true, 7 | "strict": false, 8 | "forceConsistentCasingInFileNames": true, 9 | "esModuleInterop": true, 10 | "module": "esnext", 11 | "moduleResolution": "node", 12 | "resolveJsonModule": true, 13 | "isolatedModules": true, 14 | "jsx": "preserve", 15 | "noEmit": true, 16 | "incremental": true, 17 | "baseUrl": ".", 18 | "paths": { 19 | "@/*": ["./src/*"], 20 | "@/public/*": ["./public/*"], 21 | "@/styles/*": ["./styles/*"] 22 | }, 23 | "plugins": [ 24 | { 25 | "name": "next" 26 | }, 27 | { 28 | "name": "typescript-plugin-css-modules" 29 | } 30 | ] 31 | }, 32 | "include": ["next-env.d.ts", "**/*.ts", "**/*.tsx"], 33 | "exclude": ["node_modules", "build"] 34 | } 35 | --------------------------------------------------------------------------------