Steve Yegge: only Claude Fable is worth a damn, and it's too expensive — orchestrators won't work for a year
Steve_Yegge · x · 2026-09-16
Steve Yegge's new essay Seats and Sunsets centers on the "fuel apocalypse" — token consumption — in agent orchestration:
- Fuel apocalypse: He argues Claude Fable 5 is broadly the only model in the industry worth trusting as an AI employee — and even it fails at roughly half the job (no self-pacing, spams colleagues with long emails). Weaker models like the Astras and Opus 5s are fine as personal assistants or for single coding tasks, but running a real-world factory needs Fable-tier intelligence in at least some roles, or it eats itself.
- Too expensive to work: Fable is the only good model, but its cost means nobody's orchestrators work today — including his own.
- No relief for a year: he bluntly predicts the situation won't improve for another year.
- He also details Wheelhouse, the more mature successor to Gas Town (both Beads machines) he built for his video game Wyvern, and floats turning his agent/orchestrator writing into a weekly newsletter.
Essential reading on agent engineering economics: the tradeoff between model capability and token cost.
More from coding & agent
- Microsoft's MemSteer turns unlabeled user histories into agent memory supervision — MicrosoftResearch · 2026-11-02
- She fired herself from her own Instagram: AI agent runs the account, views up 82.3% — heyshrutimishra · 2026-11-02
- Sourcegraph: AI writes code faster than teams can review it, driving more outages — AI Engineer · 2026-11-02
- Tsinghua & Microsoft's HarnessMM cuts agent token use by up to 87.93% while raising success rate — rohanpaul_ai · 2026-11-02
- GitHub launches Agent HQ: Anthropic, OpenAI, Google coding agents in Copilot — emmanuelvivier · 2026-11-02
- DeepSeek open-sources Super-Agent, a coding agent that reads team chats to minimize human input — emmanuelvivier · 2026-11-02