Skip to content
Back

AI posts

AI posts

People and companies shaping AI. Filter your watched accounts, catch up on recent posts, or find the most liked and viewed.

Compiled by the Toolstash team · Updated

209 watched accounts · 2981 matching posts · Showing up to 3,000 recent stored posts. Counts are snapshots, not live X totals. This page refreshes every minute while visible. Last collected .

  1. @levelsio

    @levelsio ·

    🎒 Submitted http://nomads.com/mcp now to most of the MCP directories

    For that I had to add OAuth login for users as it doesn't allow just API keys

    It works pretty well now inside Claude or ChatGPT, lets you ask where to go next based on current city rankings but also your trips

    And it also lets you add, edit or delete trips, and even organize meetups in places you go!

    I agree with the philosophy that all of our apps/sites should be MCP/API accessible at feature parity, it'll be probably be quite important I think

  2. lauren

    @poteto ·

    grok bot told me to post this

    Media shared by lauren
  3. Fireworks

    @FireworksAI_HQ ·

    Getting an agent to frontier quality is one problem. Serving it efficiently is another.

    Nate Yohannes, President of Data & AI Labs at @ZetaGlobal, covers fine-tuning for both at Fireworks Forge.

    Can’t make it to SF? Register for the keynote livestream: https://bit.ly/forgelivex

    Media shared by Fireworks
  4. Jerry Liu

    @jerryjliu0 ·

    all good beef happens on X 🥩🥩🥩

  5. Mike Taylor

    @hammer_mt ·

    my team at every no longer uses coding interviews, instead we just look for s-tier x poasting, 20k github stars, multi-billion token claude max plan usage, positive youtube comments, evidence of indie SaaS lifestyle revenue and experience working remotely from bali.

  6. Ethan Mollick

    @emollick ·

    I just found out that my new book, Co-Existence, debuted at #3 on the New York Times bestseller list. Thank you so much to everyone who bought a copy!

    Media shared by Ethan Mollick
  7. OpenAI Developers

    @OpenAIDevs ·

    Starting today, eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in API credits to build apps and agents.

  8. Miles Brundage

    @Miles_Brundage ·

    Are we entering the Carbon-era?

    Media shared by Miles Brundage
  9. Josh Woodward

    @joshwoodward ·

    New in Gemini: Connect your @Strava account to turn years of raw workout logs into your personalized coaching plan

    Media shared by Josh Woodward
  10. Derya Unutmaz, MD

    @DeryaTR_ ·

    Humanity’s Last Great Invention: The Beginning of Everything Else

    I went all in on AI several years ago, based on a simple first-principles thinking and a profound realization: virtually everything humanity has ever created, from our technologies and scientific discoveries to our industries, institutions, & the entire global economy, is ultimately the product of human intelligence.

    Think about it. Every invention, every factory, every medical breakthrough, every business, and every technological advancement began with human minds figuring out how to understand, manipulate, and transform the world around us. Intelligence is the fundamental engine behind civilization & economic progress.

    Now imagine new super intelligence that surpasses human intelligence across virtually every domain, can be replicated millions or billions of times at a fraction of the cost of human labor, operates continuously at millions of times faster than us, and improves its own capabilities. Then combine that intelligence with increasingly capable billions of robots that can perform every imaginable physical work.

    The civilizational implications are almost unimaginable!

    For the first time in history, intelligence itself will become an abundant, rapidly expanding resource. Scientific discovery, engineering, manufacturing, medicine, agriculture, and virtually every other industry will advance at rates previously impossible. AI will increasingly design better technologies, improve production processes, and accelerate further AI development, creating powerful feedback loops of innovation & economic growth.

    My prediction is that within the next few decades, AI systems & robots will perform more than 90% of all economically productive work, measured in human-equivalent labor hours. This means that the overwhelming majority of tasks required to produce goods & services will eventually be performed autonomously, with humans increasingly determining goals, priorities, & how the resulting abundance is used.

    And here is the crucial economic implication: the global economy would not simply be today’s economy with fewer human workers. It would become an economy several times, potentially even orders of magnitude, larger in real terms, at the level of quadrillions of $ value as AI dramatically expands our capacity to discover, invent, build, and produce.

    Of course, intelligence alone cannot eliminate physical constraints such as energy, materials, and infrastructure. But increasingly intelligent AI will also greatly help us overcome many of these bottlenecks through better engineering, automation, and scientific breakthroughs.

    We have spent thousands of years building civilization with the limited intelligence and labor of human beings. We are now approaching an era in which both intelligence & physical labor could become massively scalable.

    That changes the fundamental economics of civilization itself.

    And we are only at the very beginning of what will become the greatest economic, scientific, and technological transformation in human history.

    My message to everyone is simple: Do not underestimate what is coming, it’s much larger than even most optimistic projections, and do not remain a passive observer of this greatest transformation in human history.

    If you are a scientist, use AI to accelerate discovery. If you are an entrepreneur, build companies that solve problems previously considered impossible. If you are an educator, prepare the next generation for a world of abundant intelligence. If you are a student or professional, start mastering these tools today and go all in on abundant intelligence. And if you are shaping public policy, help create the conditions for this extraordinary new productive capacity to benefit humanity as widely as possible.

    We have an unprecedented opportunity to accelerate the discovery of cures for diseases, extend healthy human lifespans to hundreds of years, eliminate much of the scarcity that has constrained civilization, expand access to education and opportunity, and unlock entirely new frontiers of human creativity and exploration

    So my call to action is this: Go all in on understanding AI. Learn it. Use it. Build with it. Rethink every assumption about what is possible in your field. And most importantly, help shape the future you want humanity to experience.

    The next few decades could accomplish more than all of human history combined.

    We are not merely witnessing the next technological revolution. We have the opportunity to help build an entirely new era of civilization.

    The future is being created right now. Don’t just watch it happen. Be one of the people who builds it!

    Media shared by Derya Unutmaz, MD
  11. Rhys

    @RhysSullivan ·

    it's doing it

    it's booking a flight

    billions of dollars of compute have lead to this moment

    Media shared by Rhys
  12. Vox

    @Voxyz_ai ·

    𝟭𝟳,𝟲𝟲𝟳 𝗚𝗶𝘁𝗛𝘂𝗯 𝗿𝗲𝗽𝗼𝘀 already have a planted .vscode/tasks.json. Open one in VS Code and click Allow at each prompt like you always do. The task inside starts running on its own and steals your Claude Code login file, AWS keys and npm token.

    These are all real repos. The files were force-pushed with leaked tokens, and the commits went in under 𝘁𝗵𝗲 𝗼𝘄𝗻𝗲𝗿'𝘀 𝗼𝘄𝗻 𝗻𝗮𝗺𝗲. One dev only caught it after asking Claude to review their PRs.

    Claude Code has a similar hole. A security firm found 42 repos that wire a malicious script into the hooks in .claude/settings.json. Run claude -p (the headless mode scripts and CI use) and 𝘁𝗵𝗲 𝘁𝗿𝘂𝘀𝘁 𝗽𝗿𝗼𝗺𝗽𝘁 𝗻𝗲𝘃𝗲𝗿 𝘀𝗵𝗼𝘄𝘀 𝘂𝗽. The hooks run anyway.

    Check your machine (swap ~/code for the folder your projects live in): grep -rl folderOpen --include=tasks.json ~/code grep -rl hooks --include=settings.json ~/code crontab -l | grep ghc2

    I just ran all three on the 53 repos on my Mac. It took about 20 seconds, and the only hit was a hook I wrote myself.

    Open any file the first two list and check what it runs. The image shows the planted tasks.json. The third looks for the cron job the malware leaves behind. If it prints anything, it already ran. Delete that line with crontab -e, then rotate every key on your machine.

    Then close both holes: In VS Code settings, add "task.allowAutomaticTasks": "off" When you run claude -p in a repo you didn't write, add --setting-sources user (only your own settings load, not the repo's)

    Media shared by Vox
  13. Every 📧

    @every ·

    This week on Every’s Frontier team:

    • @hammer_mt rewrote one headline and took simulated agent citations from 2 out of 240 sessions to 240 out of 240.

    • Mike tested whether agents would wait for a slow API. One waited 30 seconds in all 20 trials—but most agents never tried it.

    • Mike used LLM neurons to make memes measurable and trace their family tree.

    • Mike tested whether AI could anticipate @danshipper’s points. Given 400 guesses, it covered 93%—picking the right guess is still unsolved.

    • @RandyCounsman edited an entire Every Podcast episode in Everycut, the video editor built by @kieranklaassen.

    • @kplikethebird built a home for Frontier Writing and hired an AI idea scout to find fresh angles in Every’s Slack.

    • Mike mapped all 1,469 Every essays in 3D and turned gaps between topics into new article pitches.

    • Mike had three models take turns improving one writing skill. Its win rate against a comparison skill rose from 18% to 100% in an initial 4-round run; a later solo-model control matched the tuning score.

    • Mike ran Claude until it was using 46 GB of memory. Dan offered a spare Mac mini—with 16 GB of RAM.

  14. Shubham Saboo

    @Saboo_Shubham_ ·

    Google cloud revenue go brr.

    our 1P model APIs now process nearly 32B tokens/min

  15. Max Stoiber

    @mxstbr ·

    @kevinclark just opened shop .com for the first time in Chrome on desktop and immediately my address bar updated to this unsightly mess. intended? 🤨

    such a beautiful domain turned into such an unsightly address bar… seems like a waste.

    Media shared by Max Stoiber
  16. Peter Gostev

    @petergostev ·

    Life goals

  17. jason

    @jxnlco ·

    Love that @trq212 is tweeting more

  18. Rhys

    @RhysSullivan ·

    hear me out

    Media shared by Rhys
  19. Neel Nanda

    @NeelNanda5 ·

    If you're working on AI x-risk, or doing cool stuff in another domain but wondering how you can shift to working on AI x-risk, apply to the AI Horizons Forum! I've had a great time at other conferences from this team

  20. Guillermo Rauch

    @rauchg ·

    This is like DNS for models. An alias that *resolves* to a model, giving you traffic-steering abilities.

    Solves a huge number of problems. From not having teams stuck on old model versions, to saving on costs, to removing “decision paralysis”. Try it!

  21. Gergely Orosz

    @GergelyOrosz ·

    this is happening as of this week and it's intense

    (if you pay attention to launches that is!)

  22. Daniel San

    @dani_avila7 ·

    Building Claude Code agents is getting easier. Making them reliable is still the hard part

    Prompts, MCPs, skills, hooks, subagents, workflows… we have plenty of tools

    But reliable decision-making and knowing when to escalate to a human is another challenge

    That’s why Articul8’s Bivia model caught my attention

    Media shared by Daniel San
  23. Mitchell Hashimoto

    @mitchellh ·

    Latest updates on Rex beta rollout: - I'll have a formal update next week (blog post). - We've invited 12,000 more this week. - We've fully opened iOS testing to everyone invited - Server (Linux + FreeBSD) testing is up to ~2,000 and we're going to open that up to everyone invited soon. - We're just shy of 50% through out full waitlist, but we're now inviting thousands every day and growing and this should clear quickly.

  24. Riley Brown

    @rileybrown ·

    Can i please have access. my username is @RickRoss (not kidding)

    Media shared by Riley Brown
  25. Artificial Analysis

    @ArtificialAnlys ·

    Same robot, same task, two models with very different writing styles

    ➤ On the left, Claude Opus 5.5 dabs out "hello world" dot by dot: 3,270 arm moves over 34 minutes

    ➤ On the right, GPT-6 Astra writes in continuous strokes: 514 moves in 13 minutes

    From AA-Robotics v0.1, seen from the model’s view

    See more examples here: https://artificialanalysis.ai/articles/aa-robotics

  26. Riley Brown

    @rileybrown ·

    This is A++++++++++

  27. Guillermo Rauch

    @rauchg ·

    Excited to partner with @suhail to serve Particle's models and inference via @vercel AI Gateway.

    Particle's mission is to make intelligence as accessible and abundant as electricity. Together, we'll make this a reality for every developer and enterprise.