Anthropic had its most consequential week yet. Andrej Karpathy joined to lead AI-accelerated pre-training research, new financials indicate the company may already be profitable, and a deepening compute partnership with SpaceX signals serious infrastructure ambitions. Three separate developments, one directional signal: Anthropic is not running a catch-up strategy anymore.
The piece worth reading is not the headline hiring or the profit numbers. It is NLW's argument about recursive research, where AI accelerates the research that improves AI, and why compute constraints are the actual bottleneck this changes. That framing resets how to interpret every other lab's roadmap, not just Anthropic's.
In context: OpenAI is moving toward an IPO, Cursor launched a leaner coding model, and markets are still pricing AI on the old boom narrative. The argument here is that Anthropic's week forces a structural reread of where leverage actually sits in the AI stack going into the next 12 months.
[WATCH ON YOUTUBE →]