Rauchg reveals Vercel's internal agent (1233 likes): 'We built an agent that powers our company's internal operations called @v. Every day-to-day job at Vercel now involves @v. It's growing exponentially in daily interactions and token use. One can extrapolate to the day AI agents run entire companies.' Plus his follow-up (66 likes): teams had dozens of agents, each with its own domain name - @v unified them. Plus his mastery insight (3662 likes): 'AI alone is cool. But mastery + creativity + AI hits on a whole different level. Don't let anyone discourage you from pursuing excellence and craft. Keep studying the blade.'

Dan Shipper on agency rupture (192 likes): 'When you experience a language model doing a task that used to require you at every step and now doesn't, it's a gigantic agency rupture. We equate ourselves with our tasks, our outputs, and a rupture of this kind is akin to the loss of your identity and sense of self.' Plus his philosophical note (18 likes): 'If ought implies can, and technology reshapes the field of human abilities, then technology reshapes oughts.'

Thariq's Jevons paradox (1191 likes): 'You can already see Jevons paradox at work in mathematics. There is more happening, it is easier to understand and mathematicians have more time to discuss it with us at higher abstraction levels. Demand for people who think and know about math will go up.' Plus his chess parallel (73 likes).

Levie on automation targets (318 likes): 'Some of the hardest work in the world is actually prone to automation first, particularly due to its verifiability. Math, cyber, and code - while being insanely hard and high value fields - have the benefit of being able to be tested.' Amjad's LLM chess on Lichess (215 likes): 'My LLM chess engine is now on Lichess autonomously playing real games against people and bots. 1253 Elo.' Karpathy's GTA Hobbiton (392 likes): sharing the pelican-on-bicycle source, playable in browser. 'Look out for GTA Hobbiton dropping before GTA VI.'

Nikunj on VC reality (372 likes): 'VC has effectively fully become vibes capital. The market reality at the early-mid stage is completely divorced from fundamentals.' Plus his warm NYC nights photo (416 likes). Garry Tan's white pill (302 likes): 'AI will create unimaginable economic growth and that is the best white pill.' Plus wonder disappearing (173 likes): 'The sense of wonder disappeared right at the moment the amount of wonder is going parabolic.' Plus map vs territory (195 likes).

Ryo Lu on software's future (390 likes): 'My early mentors were these apps, especially Rdio, Mailbox, and Apple. They made new patterns that made software feel simpler and intuitive. As we leave the world of apps behind, what parts of software will remain visible, and how could it feel?' Steipete on new spam patterns (271 likes).

Training Data: Sonya Huang interviews Jerry Tworek (ex-OpenAI VP, ran Strawberry/reasoning) and Rohan Anil (ex-Gemini pre-training lead, also Google Brain and Anthropic). They founded Core Automation to build the 'most automated lab in the world' and replace the transformer architecture. Key thesis: transformers can't do continual learning, RL alone isn't sufficient, and the architecture itself is the bottleneck to smarter systems. They need models that learn at test time, on user data, on real-world distribution. Six-month roadmap: kernel generation automation, new architecture experiments at scale, combining pre-training and RL end-to-end.