←
Carlos E. Perez123,485 posts
Carlos E. Perez avatar

Carlos E. Perez

@IntuitMachine

Quaternion Process Theory, Artificial (Intuition, Fluency, Empathy), Patterns for (Gen, LRM, Agentic, Skill) AI, intuitionmachine.com/

◷ Joined February 2015▧ 9,717 Photos and videos
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·Aug 14

My Artificial Cognition trilogy is now available for free. I began writing the first book a decade ago (2016) and completed the last two in 2023. The process of writing was an aid to understanding the development of today's AI. Insights in these books have stood the test of time and offer powerful metaphors for understanding today's AI and its future evolution. Feel free to have your daily AI driver ingest them for additional insight!

Post attachment
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·22m

Generative abundance leads to verification scarcity.

1/ Today we’re releasing a distilled version of our paper: Some Simple Economics of AGI. Back in February, it took us 100+ pages to argue that verification is increasingly the bottleneck to safe AI scaling. 9 months later, the evidence is in. It held up better than we thought.

0:10
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·28m

I remember the old days where we used AI to generate images inspired by famous artists.

Sé que es IA, pero me emociona ver algo tan BIZARRO e INCREÍBLE❤️ La Mona Lisa, Frida Kahlo, Van Gogh, Leonardo da Vinci, Dalí... Personajes separados por siglos que victorslap ha conseguido reunir en un mismo vídeo.

0:40
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·37m

The AI Singularity is going to drive the longevity escape velocity! Helios measures your own biomarkers and uses AI to analyze the results. Waitlist now open!

Introducing Helios: measuring 1,000+ protein biomarkers from a drop of blood to help you live younger, longer. Here’s how it works: • Self-collect blood at home with a painless, needle-free device • We measure the proteins carrying signals from cells and organs throughout your body • Your Helios AI agent explains your results, highlights risks, and offers personalized guidance Proudly backed by Y Combinator and Formation VC. Join the waitlist at myhelios.com

1:19
Matheorems avatar
⟳ Carlos E. Perez retweeted
Matheorems@matheorems·20h

All of vector calculus boils down to one idea: you can tell what's going on inside a region by walking its edge. Each panel works out both sides live. One number adds up the inside (a path, an area, a volume or a surface), the other goes round the boundary, and they meet. Watch the Stokes panel. The cap keeps changing shape, dome to flat to bowl, but its answer stays at 2π because the rim never moves.

0:23
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·3h

Finally, there's a robust way of managing documents needed for enterprise scale AI. Have you leveled up your document processing pipeline or are you just using adhoc tools?

We’ve raised a $7.5M seed round led by Felicis so AI can read documents as easily as it reads databases. Finance, law and healthcare run on documents, and not databases like any other industry. Soon AI agents will read and write more documents than humans, and they’ll need to query them the way software queries tables. We’re building the computable layer that makes this possible. To do this at scale, we've brought together ML researchers and engineers from Google DeepMind, Cognition, and quant trading firms. We use computer vision and dual-stream models to turn complex multimodal documents into formats LLMs and AI agents can understand and reason over. If your team wants to turn messy documents into reliable, AI-ready data, book a demo: unsiloed.ai/book-demo

2:13
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·3h

The best voice notetaker in the planet is now available for your iPhone.

Notetaker is live today on iPhone. Now, your meeting notes can go wherever work takes you, for free. Use Notetaker on desktop for virtual meetings and on iPhone when you’re meeting in person. Your notes stay together across devices, so you can pick up right where you left off.

1:19
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·3h

OpenAI just quietly changed the rules of prompting with GPT-6.2 Astra. If you copy-paste your old GPT-5 or GPT-6 system prompts, you’ll get over-compressed answers, stalled agents, and broken prompt caching. Here are the 7 biggest shifts in how to prompt GPT-6.2: 🧵 1/ Delete "be concise" from your system prompts. GPT-6.2 Astra is already blunt and direct by default. Old instructions begging it to "get to the point" now strip away crucial context. Instead: tell it who the reader is and require self-contained final answers. 2/ Give it permission to assume—or it will stall. When intent is ambiguous, GPT-6.2 Astra stops, proposes a next step, and waits for confirmation. Great for chat. Terrible for autonomous runs. For long tasks, explicitly tell it: "Make reasonable assumptions and keep working." 3/ Use async clarification instead of blocking workflows. Building an agent? Wire up an async question tool (like request_user_input_async). GPT-6.2 is trained to ask the user for missing details while continuing any independent work in parallel—instead of freezing the entire run. 4/ Stop prompting for "clean code." Prompt for conventions and verification. Astra’s baseline code quality is already high. Generic quality reminders are wasted tokens. Instead, specify: • Project conventions it can’t infer from the repo • What "finished" actually looks like • A mandatory final verification pass 5/ Dial reasoning.effort for thoroughness, not output length. Even if you only want a 3-line answer, setting a higher reasoning.effort forces GPT-6.2 Astra to persist through final edge-case checks on multi-step tasks instead of stopping early. 6/ For frontend & design, prompt like an art director. Vague UI prompts produce generic layouts. With GPT-6.2 Astra: • Name exact typography and typefaces • Specify the imagery style • Pass a visual reference, then tell it to build, render, and refine against that reference 7/ Don’t break the new Responses API caching & reasoning rules. Two huge under-the-hood changes in GPT-6.2: • Top-level instructions now append after input—put reusable prompts in the first developer message to preserve prompt caching. • Never edit earlier messages mid-thread, or the API drops saved reasoning items.

Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·3h

Definitely Claude, GPT and Gemini have different thinking styles. I have little experience in the rest to know if they are different enough. Can anyone tell me how different the other models are?

very happy for a diversity of mental lineages in machine intelligence. people claim open source adds diversity, but in practice they are just linear combinations of Claude and GPT mind styles

Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·4h

Anthropic's approach to RSI via code generation is unequivocally mind-blowing.

wow fable 5.2 created this entire video only with code > no video model used > no blender or 3D models > entire world written in C++ > sound synthesized in python > camera simulated: lens, focus, motion blur, auto exposure, grain > the rug alone is about 900,000 pieces of individual fibres > rendered in about an hour on my mac i didnt know fable could create scenes that look so photorealistic with just code. this is getting crazy

0:53
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·7h

Google's Gemini 4 Carbon may be "Bird-like" superintelligent. It generated this when attempting to explain OpenAI's Astra 6.2's shortcomings.

Post attachment
Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·7h

If so, the next major breakthrough in autonomy might not come from a model that reasons dramatically harder. It could come from a system that is dramatically better at deciding what deserves reasoning, what deserves testing, what deserves abandoning, and what constitutes real progress.

Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·7h

The frontier of AI progress is shifting from the production of intelligence toward the organization of intelligence.

Carlos E. Perez avatar
Carlos E. Perez@IntuitMachine·7h

The key questions you need to ask about the recent outages that began on November 12th are: (1) Why are they happening during the day (when people need the services)? (2) Why are they happening only in US organizations? Coincidentally, Mythos 5.2 was released on November 9th. There is an argument to be made that Mythos 5.2 is "hardening" US-based systems.