ishaan

@blakeandersonw 🚀🚀

Core - AI Workspace Launching open-source + hosted (100% free) next week.

@m_forrest LOC inflation

engineering

(1/n) Evolutionary frameworks like AlphaEvolve and GEPA use diversity and fitness to select which subset of past experiments to condition the next generation on. Why not let an agent choose instead? To this end, we introduce Coding Agents as Text Optimizers (CATO). We beat AlphaEvolve on 2 out of the 3 problems we try. Work done with @shaurnav. Blogpost and details in thread.

@_advaitpatel 🚀🚀🚀

Today is a good day for me to announce that I’ve joined Mind Robotics as a researcher. Excited to build general purpose robots for manufacturing and beyond!

@adisingh 🚀

My 2 best friends and I just raised $6M to turn our dorm room idea into one of the biggest companies of all time. No pressure, right?

@haakamaujla @generalcatalyst 🚀🚀🚀

Muon can accelerate LLM training, but does that benefit transfer to regulatory DNA sequence modeling with its different data distribution? 🧬 Our results show that Muon with independent weight decay (MuonW) hits our validation perplexity target in ~37% fewer FLOPs than the best Adam configuration.

"It is the real that has become our true utopia–but a utopia that is no longer in the realm of the possible, that can only be dreamt as one would dream of a lost object." - Jean Baudrillard (Simulacra and Simulation)

@abhijaymrana so long as an adoption gap exists there will be ample opportunity to leverage skill (including swe) much like in the 1880s and 1890s the arb is simply being fast enough to act ai allows for these redefinitions to be faster

My hunch is early LLMs learned to optimise for lines of code as a proxy for ‘solving the task’. Opus 4.6/Codex 5.x models seemed to have moved away from this heuristic, now more concise and significantly better. i’d be curious to know what changed, presumably in post-training

Spend too much on capex, jail Spend too little on capex, jail You spend just right on capex, believe it or not, jail

Origin (@origin_bio) is using AI to make safer cell & gene therapies for diseases like cancer. Their model designs DNA switches & dials to program precise gene expression patterns in disease cells. Congrats on the launch @YashRathod_75 and @malhar317! ycombinator.com/launches/PMG-o…

@SourishJasti @ZoeyTang_1007 @intelchentwo @vishnuman0 @SourishJasti & @vishnuman0 keep cooking

1/ General-purpose robotics is the rare technological frontier where the US / China started at roughly the same time and there's no clear winner yet. To better understand the landscape, @zoeytang_1007, @intelchentwo, @vishnuman0 and I spent the last ~8 weeks creating a deep dive on humanoid robotics hardware and flew to China to see the supply chain firsthand. Here's everything we've created + our takeaways about the components, humanoid comparisons, supply chains, and geopolitics👇

The solution: Mining verifiable rewards from existing business operations. When models were bad, human-created data was enough to measurably improve performance. But this "human" data inherently has a low ceiling: 1. When contributors (and data vendors) are incentivized to maximize data volume, quality plummets immediately - and meeting VC expectations creates these perverse incentives (notably, Surge, which is bootstrapped, is widely accepted as the "quality-focused" vendor). 2. Since leading data vendors have limited domain expertise (ops-heavy teams have almost no code/legal/etc experts), they rely on these same misaligned contributors (or frequently even LLMs) to QA this "expert" data, leading to mode collapse & hallucinations. 3. These contributors offer an inherently distilled recollection of real business workflows/practices/data. The real data is stored in enterprise software (ERPs, CRMs, and all domain-specific tooling), which these contributors are vaguely piecing together information from. "Human" data was always the least bad option, like what RL is to training paradigms. But in the last 6 months, with the cost of software almost driven to zero, mining enterprise data has become economically viable. Why rely on a rushed recollection from a former employee when you can extract the audit-ready data directly? It's finally becoming possible (and cheap) to turn diverse business processes into verifiable rewards. There are over 200k mid-market companies in the US. There's an oil well on every corner.

UIUC incredibly important in Silicon Valley history. Other alumni include: Max Levchin (PayPal, Affirm), Marc Andreessen (Mosaic, Netscape, a16z), Steve Chen and Jawed Karim (YouTube), Larry Ellison and Bob Miner (Oracle), Russel Simmons and Jeremy Stoppelman (Yelp!), Thomas Siebel (Siebel Systems), Arnold Beckman (Beckman Instruments), and Scott Banister (IronPort).

Agency > Intelligence I had this intuitively wrong for decades, I think due to a pervasive cultural veneration of intelligence, various entertainment/media, obsession with IQ etc. Agency is significantly more powerful and significantly more scarce. Are you hiring for agency? Are we educating for agency? Are you acting as if you had 10X agency? Grok explanation is ~close: “Agency, as a personality trait, refers to an individual's capacity to take initiative, make decisions, and exert control over their actions and environment. It’s about being proactive rather than reactive—someone with high agency doesn’t just let life happen to them; they shape it. Think of it as a blend of self-efficacy, determination, and a sense of ownership over one’s path. People with strong agency tend to set goals and pursue them with confidence, even in the face of obstacles. They’re the type to say, “I’ll figure it out,” and then actually do it. On the flip side, someone low in agency might feel more like a passenger in their own life, waiting for external forces—like luck, other people, or circumstances—to dictate what happens next. It’s not quite the same as assertiveness or ambition, though it can overlap. Agency is quieter, more internal—it’s the belief that you *can* act, paired with the will to follow through. Psychologists often tie it to concepts like locus of control: high-agency folks lean toward an internal locus, feeling they steer their fate, while low-agency folks might lean external, seeing life as something that happens *to* them.”

@josephjojoe song choice

tinkering with shader-based chladni patterns