Florian Laurent

Excited to share we’re launching @gen_intuition today, alongside our $133.7M Seed round! We’re a research lab building novel foundation models dedicated to games - and the real world

why did we call it hallucination and not confabulation en.wikipedia.org/wiki/Confabula…

@Snixtp Now replicate the /fast command to dynamically change it!

More efficiency tests on a single 3090 TL;DR: - I tested 8 local LLMs on a single RTX 3090, power limit from 100W to 450W. - Across the full set, the best efficiency was consistently around 225-250W. - At 225W, the card averaged 90.4 tok/s at 0.4167 tok/s/W. - At 450W, it reached 107.1 tok/s, but efficiency dropped to 0.2731 tok/s/W. - So max power added only ~16.7 tok/s, while using about 184W more.

@charles_maddock It was the best of times, it was the worst of times

looking back at old 100k+ LOC projects that I wrote with my bare hands back in my prime

@badlogicgames First thing i implemented: keeping all tool call outputs on a single line

"pi, there's an example extension called minimal.ts or so. i want that, but cute. do the thing. make no mistakes"

@olv "Biscuit"? 🤔 119 W 24th St

bruh

tried out our good old Cedille model in Pi "je n'ai pas envie de le faire" 😬

Reminder that I wrote a little book about deep-learning, which is phone-formatted, entirely free, and nearing the 1M download: fleuret.org/francois/lbdl.…

looking forward!

@Kpaxs Specialization is for insects

Here’s a secret that every genuinely original thinker knows: there are no original ideas. there are only original combinations. Every “breakthrough” is two existing ideas from different domains meeting for the first time inside someone’s head. The person who reads only within their field will only ever have ideas that their field has already had. This is why the most interesting people are almost always polymaths. Go wider. Read the thing that has nothing to do with your work. Talk to the person who has nothing in common with you. Visit the place that makes no sense on your itinerary. The irrelevant input is the one that will combine with everything else and produce something nobody’s ever seen.

Claude watching me write code manually after I hit the daily limit

We’re hiring Engineers and leaders in SF! If you’re excited about hard problems in proactive AI, context graphs, and memory, DMs are open! Open roles include Head of Engineering; Full Stack, Backend, ML/AI Engineers; Head of ML/AI. If you’re a builder in the Bay, let’s talk.

@wandb is there a way to point it to a specific project? instead of trying to guess from cwd?

BREAKING: Meek Mill has abandoned the W&B web dashboard entirely. Says he only monitors training runs in the terminal now. "I don't need a browser to know loss is converging," sources close to the rapper confirm.

Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords. LiteLLM itself has 97 million downloads per month which is already terrible, but much worse, the contagion spreads to any project that depends on litellm. For example, if you did `pip install dspy` (which depended on litellm>=1.64.0), you'd also be pwnd. Same for any other large project that depended on litellm. Afaict the poisoned version was up for only less than ~1 hour. The attack had a bug which led to its discovery - Callum McMahon was using an MCP plugin inside Cursor that pulled in litellm as a transitive dependency. When litellm 1.82.8 installed, their machine ran out of RAM and crashed. So if the attacker didn't vibe code this attack it could have been undetected for many days or weeks. Supply chain attacks like this are basically the scariest thing imaginable in modern software. Every time you install any depedency you could be pulling in a poisoned package anywhere deep inside its entire depedency tree. This is especially risky with large projects that might have lots and lots of dependencies. The credentials that do get stolen in each attack can then be used to take over more accounts and compromise more packages. Classical software engineering would have you believe that dependencies are good (we're building pyramids from bricks), but imo this has to be re-evaluated, and it's why I've been so growingly averse to them, preferring to use LLMs to "yoink" functionality when it's simple enough and possible.

mapped lyrics I heard while running (each word placed where I heard it) using spotify + strava APIs

i checked if my running pace correlates with song BPM answer is no (~500 songs over ~100 runs)

“The deepest value isn’t physics. It’s human intuition. A physics engine models how a drone moves; it can’t model how a skilled operator reacts when surprised. In surgery, it’s the feel for how the tissue responds to the scalpel. Train on human decision-making and you capture expertise that can’t be described with words, only shown, felt.” KV Partner Nicole Fraenkel on world labs and why we bet on @gen_intuition last year fortune.com/2026/03/06/ai-…

there is a game called "data center" on steam which let's you build and manage your own data center. this is lowkey genius, the best way to educate people on a new trait. hyperscalers should learn a thing or two from "edutainment".

Claude: “I estimate this will take 1-2 weeks to complete” Me: