Developer slams Claude's invisible sub-calls and opaque eval harness

Developer kalomaze criticized Anthropic's agent harness, where invisible small-model calls misinterpret /goal instructions and let stop hooks kill completed tasks, and slammed the company's opaque eval harnesses for undermining third-party reproducibility of benchmarks.

2026-11-14 ~ 2026-11-14 · 4 related posts

• kalomaze slams Anthropic over 'invisible nontransparent harness' in evals — kalomaze · 2026-11-14
• kalomaze slams Anthropic's opaque agent harness: sub-model misreads /goal and blocks the main model — kalomaze · 2026-11-14
• Subagent misreads an abbreviation, stop hook blocks Claude Code after goal is met — kalomaze · 2026-11-14
• kalomaze: 'subagents' are really invisible model calls, and one just rejected a completed task — kalomaze · 2026-11-14