Cursor's $0.50 Coding Model and Anthropic's $900B Round
On May 18, 2026, Cursor shipped Composer 2.5: a coding agent that ties Claude Opus 4.7 on SWE-Bench Multilingual and edges it on CursorBench v3.1, at $0.50 per million input tokens and $2.50 per million output — roughly one tenth the per-token price of Opus. The base model is Moonshot AI's open Kimi K2.5; 85% of the compute went into Cursor's own post-training pipeline. The same week, Anthropic is closing a $30B round at a $900B valuation, annualized revenue is going from $9B at end of 2025 to over $30B in April 2026, and Andrej Karpathy announced he's joining Anthropic's pre-training team.
If frontier coding got commoditized in seven days, what exactly is anyone paying premium for? That's the cold open. From there, Oscar and Matt filter the week's AI noise — Karpathy to Anthropic, Google I/O 2026, Gemini Spark, Antigravity 2.0, Hark's $700M Series A — debate the tools they actually adopted, and close with two hot takes you won't hear anywhere else.
Like how we think about AI?
Human In the Loop is me and Matt thinking out loud. Putting that thinking to work inside a real company is the day job. If you're a founder or team trying to ship AI that survives production, let's talk.
Free 30-minute call · No pitch, just a plan · No commitment until you say go
More from the archive
Meta Compute and the Next AI Startup Bets
Ep 13: Meta Compute and the Next AI Startup Bets Meta spent so much on AI infrastructure that it may have backed into becoming a cloud company. On July 1, Reuters…
The Frontier Just Got a Guest List
OpenAI just shipped its most powerful model to twenty companies the government picked. Not the best twenty. The approved twenty. A week after the US switched off…
The World Cup Is Secretly Run by AI
Last week the US government gave Anthropic's best model a 72-hour public life, then switched it off for every foreign national on earth. This week, three Chinese…