Browser extension enables LLM-to-LLM critique loops

AccurateCartoonist38 · reddit · 2026-08-15

The author built a browser extension to automate interactions between different LLMs, moving beyond manual copy-pasting. The workflow cycles responses (LLM A -> LLM B -> LLM A -> LLM B), allowing models to critique and react to each other's reasoning. The author found it interesting to observe how responses shifted based on the critique, raising the question of when such multi-LLM interaction improves answers versus just adding cost and latency.

Original post →

More from coding & agent

coding & agent channel →