How Anthropic Builds: Lessons from Labs — Mike Krieger, Anthropic

Over a single weekend, Mike Krieger had Claude port a few hundred thousand lines of Python to TypeScript, verify it, and churn on its own output until the thing was deployable. He came back Monday to a finished port. He offers it as an example of a habit he thinks most people still lack, which is being unreasonable in what you ask for. The first generation of AI products boxed models in, granting so little access to tools and environment that ambition felt pointless, and the instinct to ask small has outlived the constraint. He knows the pull of the old way from the inside. He spent two years as Anthropic's chief product officer, watched everyone around him building, and moved into an individual contributor role at Labs because the FOMO became unbearable. Labs runs on two week persevere or pivot reviews, and projects get wound down in nearly every cycle, which is the design rather than a failure. Teams assemble fresh around each bet, so the org chart deliberately does not track projects and bet leads usually manage nobody. The bottleneck he cares about is not review time but whether a human can still hold the change in their head. A 2,000 line pull request now travels with an artifact laying out intent and tradeoffs, and he admits he no longer reads every line. He ends somewhere most technical keynotes do not go, on burnout, and on the reminder that no job is so important you cannot be offline for two days. Speaker info: - https://x.com/mikeyk - https://www.linkedin.com/in/mikekrieger Timestamps: 0:00 - From chief product officer to IC at Labs 2:55 - Be unreasonable, and why people still ask small 4:51 - Porting Python to TypeScript over a weekend 6:48 - Scaling lessons from 2010: measure everything, build knobs 8:06 - Tagging Claude in Slack, and multiplayer delegation 10:37 - The review bottleneck is comprehension, not time 11:53 - Persevere or pivot, and bet leads who manage nobody 13:43 - Where Claude Design goes next 15:37 - Project unhip, and what he would delete 17:34 - Why start a company when the labs ship this fast 20:06 - Vertical AI and finance 22:00 - Burnout, 996, and staying sane 25:04 - Verbalizing emotions as a leader