Ask HN: Is there any problem using multi-LLM
TL;DR
On Hacker News, a developer asks whether using Codex, GitHub Copilot and Claude Code together for personal projects carries any risk. The thread stays small but touches a real nerve: many dev teams now run multiple AI assistants in parallel without clear best practices for data flow, license terms or potential code leaks. The discussion highlights how little battle-tested guidance exists for multi-LLM workflows.
Nauti's Take
Running several LLMs in parallel is a strong option — Codex, Copilot and Claude Code have different strengths and together cover more edge cases. The risk the HN thread rightly raises: each tool may ship code, prompts and repo context to a different vendor with its own privacy and training terms.
Before going multi-LLM, teams should map sensitive repos, separate client work from personal projects and write down clear per-tool policies.
Sources