Regular participant in itnig tertulias; deep hands-on experience with AI coding tools (Cursor, Claude Code, Codex); tracks model capabilities, compute access, and developer workflow evolution daily.
no scored calls yet — needs a stated position or a categorical verdict, with a matured window vs SPY
Cursor owns the world's largest developer interaction dataset; training own models (Kimi 2.0 based on Chinese open source) and now leveraging xAI's Colossus 2 supercomputer for cheap, fast, code-specialized models; building computer-use agent layer that makes switching costs high for developers.
Anthropic has achieved true enterprise product-market fit via Claude Code/Enterprise layer; Q1 $5B revenue scaling to $10B in Q2 with 30%+ of open roles in enterprise sales at £600k+; enterprises pay full token price vs 10x consumer discount, creating pricing power and stickiness through annual contracts.
ChatGPT has massive MAUs but only 20-30% pay $20/mo — not a comparable business to Anthropic's API/enterprise model; now aggressively hiring enterprise talent (CMO from ServiceNow) and pushing Codex to capture the coding agent layer where enterprises sign annual contracts at full token pricing.