Symptom: Qwen3.8-Flash-Next and GLM-5.3-Flash repeat plans in reasoning, rethinking again and again in each block without apparent awareness of prior turns and being slow to progress under Hermes Agent.
Cause: By default, Hermes Agent strips reasoning from assistant replays for custom providers, whatever the backend and the chat template might have set it to. This can cause problems for models that accept (or indeed expect) preserved reasoning. Other frameworks may have similar defaults, though I am not familiar with any specific examples.
Fix: Set model.reasoning_echo totrue in Hermes Agent config.
Just thought this might be non-obvious and useful for anyone else having loopy local models in agentic harnesses. That would be all.
source & further reading
forum.level1techs.com — original article
When Is It Worth Personally Preparing for AI Disasters?
What should the EA community be doing now that the (AIS) Overton window has opened up again?
Money Pumps for Agents Who Are Ambiguity Averse or Ambiguity Seeking