a few other interesting ai data startups kled Avi Patel hud oro DG CK parsewave Van0SS (thanks Ibragim) patronus Rebecca Qian (thanks Tim Shi) antim labs & lucent & good start labs (via Sean Cai)

AI is writing a growing share of the world's software. No one is formally verifying any of it. New essay: "When AI Writes the World's Software, Who Verifies It?" leodemoura.github.io/blog/2026/02/2…

I genuinely feel bad for all folks early in their career or folks just jumping into the job market, not just because of the obvious AI impact Cold email doesnt work because AI bots are sending 100s of emails a day and people ignore them Linkedin Jobs has 1000s of applicants to

Leo Boytsov My minor gripe is just that it implies that people look at SWE-Bench scores and assume our LM systems are capable of autonomously solving all coding tasks. I don’t think anyone was claiming this. They’re missing the context on what SWE-Bench was created for and what purpose it

While waiting for DeepSeek V4 we got two very strong open-weight LLMs from India yesterday. There are two size flavors, Sarvam 30B and Sarvam 105B model (both reasoning models). Interestingly, the smaller 30B model uses “classic” Grouped Query Attention (GQA), whereas the

i actually read the stanford paper linked here. it doesnt say anything about "semantic collapse" or embedding precision dropping 87% at 50k docs. its an empirical study of legal RAG tools like Lexis and Westlaw hallucinating 17-33% of the time on legal queries. the actual finding

As an advisor, my take is that a PhD student is not a paper-generating machine. If a professor thinks AI agents can replace students, they might have lost sight of what advising is truly about IMO mentorship is about training an independent thinker, building a relationship, and

I (finally) put together a new LLM Architecture Gallery that collects the architecture figures all in one place! sebastianraschka.com/llm-architectu…

Today's the third time this year I’ve heard someone’s partner died in their sleep. Third. I must blurt the uncomfortable truth out loud. this is often an engineering problem. We need hardware-AI. And you might already own some of the fitness trackers that get us there. 🧵

in preparation for GTC tomorrow, i offer a 3000 year old ode to the Conference (Atharva Veda 7.12)

We just got into Y Combinator, got $500K, and I’m moving to SF! We are two 18-year-olds building , an AI-native agency that creates technical docs for dev tools. Team of 5, ~$150k+ ARR, worked with Greptile, Reducto, supermemory, etc. I started this business

I am delighted to share that I will serve as the Founding Director of the Lodha Theoretical Physics Institute (LTPI), Mumbai. ltpi.org Launching with EPQHS-10 (May 27–29, 2026), LTPI aims to become a world-class centre for fundamental physics.

Turns out LLMs also love to gamble... We made Claude, GPT-5, and Gemini agents play 100+ hands of no-limit poker♠️against each other and analyzed their reasoning traces. What we found: the largest gap in frontier AI isn't reasoning in isolation, it's adaptation and coordination

A couple monks came into Notion today to tell us how they use the product. Totally flabbergasted that they're configuring custom AI agents and taking AI meeting notes with custom instructions. This is advanced, power-user usage... monks! Use AI... find inner peace?

we’d rather execute than talk but we were told to post more: last summer when IPAM lost federal funding mid-cycle, we flew to LA and moved $ same-day so they could make payroll it isn’t a billion dollars but enough to keep Terence Tao’s team going until the system caught up

Terence Tao spent a year at the Institute for Advanced Study - no teaching, no random events of committees, just unlimited time to think. But after a few months, he ran out of ideas. Terence thinks that mathematicians and scientists need a certain level of randomness and

ok wildly successful

A polite pigeon in Massachusetts calmly rode a commuter train before disembarking at the next stop.

Do feel the answer is apt at some level, it should have been "handed him to me" [of course given pragmatics of the situation the "me to him" should've been realized and corrected at some level the same way LLMs handle other grammar errors so it's still interesting that didn't

So much discussion focuses on AI displacing human labor. But we may be seeing initial signs of another transformation -- shifting away from **humans as consumers** A couple months ago, Micron completely shut down their consumer memory line Crucial (after 30 years!), pivoting to