10 people, $3.1M monthly tokens, 77k changes: fact-checking the Spawn AI startup story
sujingshen · x · 2026-11-01
Sagasu fact-checks the viral Spawn story (10 people, $3.1M/month in tokens, 77k changes in 30 days): the company and timeline are real and Simon Willison blogged about it, but the numbers come with filters, and no revenue is disclosed.
- Core argument: teams fail with AI not because models are dumb but because they stuff AI into human-era processes—after generation speeds up, bottlenecks shift to communication tax, review queues, and exec sign-off.
- Counter-evidence: METR found devs felt 20% faster but were 19% slower; GitClear shows duplicated code blocks up 81% and refactored/moved code down from 21% to 3.8%.
- The author, a non-technical PM who burned 1B tokens building a macOS agent client, warns against being an AI's full-time sponsor.
- Three copyable practices for small teams: real acceptance testing, minimal judge rules, and deleting more than you add.
More from coding & agent
- GLM-5.3 shows strikingly different chain-of-thought under Claude Code vs Codex scaffolds — sebkrier · 2026-11-01
- Indie hacker's viral video made with a Skill that auto-builds social-proof walls — yihui_indie · 2026-11-01
- Vibe Kanban's Shutdown Teaches a Lesson: Keep Agent Coordination MCP-Portable — onitsoga · 2026-11-01
- Kimi autonomously used a 4x3090 rig and YuE2 to score its own 45-second launch film — minchoi · 2026-11-01
- Kimi models a walkable 3D forest in Blender, rendered in-browser with Three.js — minchoi · 2026-11-01
- Kimi K3.1 generates a 3D turbo engine visual in about 40 minutes from one prompt — minchoi · 2026-11-01