kalomaze slams Anthropic's opaque agent harness: sub-model misreads /goal and blocks the main model
kalomaze · x · 2026-11-14
Local-LLM dev kalomaze calls out Claude's agent harness: a smaller, shorter-context sub-model misinterprets an abbreviation in /goal as something entirely different, then prevents the main model from finishing — and it's on by default. He says he hates "invisible nontransparent harness bullshit," implying Anthropic is a big fan of it. A pointed critique of observability in commercial coding agents: sub-agent interactions users can't see or control.
Related event: Developer slams Claude's invisible sub-calls and opaque eval harness(4 posts)→
More from coding & agent
- Inside the MI300X: 23-min visual guide from compute chiplets to a Gluon GEMM — saurabhtwq · 2026-11-14
- Job hunter's trick: fix a real issue in a startup repo, send the PR as evidence — minhnguyen · 2026-11-14
- Screenshot shows multi-agent, multi-user chat with realtime 3D video and long-term memory — nptacek · 2026-11-14
- DeepSeek opens its entire app stack via API: search, voice, OCR, ASR, TTS — churchkey · 2026-11-14
- Garry Tan's 52-task eval: Fable 5.2 matches Opus 5.5 on pass rate but costs 2.80x more — garrytan · 2026-11-14
- YC's Garry Tan open-sources agent harness evals: pass rates tie, costs differ 2.5x — garrytan · 2026-11-14