Viral 'the harness is the machine' thread touts unverified Claude Fable/Mythos 5.2 9-equation architecture
blaizedsouza · x · 2026-11-11
A widely shared Harness Engineering thread claims Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, plugged into a Claude Code harness formalized in 9 equations: Opus for architecture, Sonnet 5.5 writing code, Haiku intercepting tool calls in <50ms, Fable 5.2 running 3,500 machine-checked proofs, Mythos 5.2 red-teaming diffs for zero-day vulnerabilities.
The claimed benchmarks (FrontierCode merge rate 49.1%, task completion 18%→95%, cryptographic receipts required before git commit) and the task-contract + isolated-worktree + adversarial-verification architecture guide have spread widely. Caveat: none of these model names or benchmark numbers are confirmed by Anthropic — treat with skepticism.
The core thesis — "the model is the engine, the harness is the machine; 99% of developers still run raw models in an open chat window" — aligns with current agent-engineering consensus.
More from coding & agent
- Stop burning tokens: a Claude Code orchestrator prompt that routes tasks to cheaper models — TawohAwa · 2026-11-11
- Amazon paper: multi-agent setups burn 1.4-4.7x tokens for small gains; single agent is the default — rohanpaul_ai · 2026-11-11
- Dev shows off building a retro FPS entirely via Codex Mobile — 503 tests passing — Dimillian · 2026-11-11
- Explosion launches Ellf directory: 52 NLP templates plus multi-agent annotation workflow — _inesmontani · 2026-11-11
- Meme: 'Stochastic AI Agent Architecture' Is Just a Markov Chain in Disguise — PrasoonPratham · 2026-11-11
- OpenCode adds Gemini 4 Carbon and Claude Fable 5.2 support — dean_rie · 2026-11-11