Altman's optimism manifesto (10148 likes): 'I would rather be an optimist and work hard than a pessimist posting about why things won't work. It's much more difficult and the most likely path is failure, but society fails if people don't try. No amount of "it will never work" essays will drive society forward.' Sottiaux's cyber hire (467 likes): 'Better Cyber. Excited to welcome Halvar Flake to the team soon.' Plus his open-ended question post (1464 likes): 'What is my title at OpenAI' — driving massive engagement.

Rauchg's Vercel milestone (398 likes): a cryptic post showing massive numbers. Plus DeepSeek v4 savings (238 likes): '1 line of code in @aisdk saves you 90% or more in DeepSeek v4 Flash AI Gateway tokens.' Plus (188 likes): 'Vercel is the Vercel for backends.' Josh Woodward's Notebook launch (341 likes): 'Notebook is built for thinking, not toggling. While others add more modes, Notebook keeps everything in a single, unified prompt bar and just does the things you want.'

Levie on enterprise AI diversity (145 likes): 'Unlike the early innings of cloud where there were really only a couple deployment patterns available, AI has a much wider array of implementation strategies. Every company is doing something meaningfully different.'

Zara on tech adoption psychology (90 likes): 'Most people will not adopt a new technology just because it makes them more efficient. They'll adopt because someone similar to them adopted it and got positive results.' Plus her AI training tip (44 likes): 'The best AI training isn't a course. It's pulling an agent into your team's group chat and letting people watch it work.' Plus her meeting efficiency post (74 likes): 'An efficient meeting has no to-do list left behind, because all actions are done DURING the meeting.'

Dan Shipper on agency rupture aftermath (70 likes): 'Once the agency rupture heals and the AI goes back to being invisible, we'll think only about the humans and what they've done. AI use will be assumed and unimportant.' Nikunj's discrete meetup spot (221 likes). Matt Turck's SaaS exit reality (117 likes): 'Everyone on X: damn, Airtable sold for so low! Many SaaS founders, secretly: damn, I'd take that any day.' Garry Tan on Prop 40 (1140 likes): 'It's insane that the CA Dem Party is endorsing Prop 40, the asset seizure tax that will destroy the California tax base.' Plus his housing post (241 likes).

Training Data: Sonya Huang interviews Chai Discovery co-founders Josh Alman (ex-OpenAI early team) and Matt Cagle (math/CS background turned protein structure researcher). Chai is engineering drug molecules with AI, treating biology as a scaling problem. Key topics: from 0.1% antibody binding rate to 15% with Chai-2 model, the bitter lesson applied to biology (simplicity + scale > bespoke modules), why diffusion models unlocked protein generation, building models from scratch (not fine-tuning LLMs), the 23-submodule problem in Chai-1 and how simplification drives scaling laws, and why drug discovery should become drug design.