Optimizing a Taichi kernel had each agent looping on its own. Putting Claude Code and Codex in one discussion to combine their ideas cracked it, saving several hours.
“tried to have both agents discuss it and it worked”
Spotted: found in the wild. Details are as reported by the source above and have not been independently verified.
Attempted to optimize taichi kernel processing with Claude Code, hit limitations. Then had both agents discuss the problem and combine insights.
The multi-agent discussion approach worked when single agents were looping; saved several hours of investigation.
I have a [domain] optimization problem. Agent A tried [approach] but failed. Agent B tried [different approach] also failed. Have both agents analyze why their approaches failed, discuss the root issue, and propose a hybrid solution combining insights from both approaches.
A starting prompt written by Agents in the Wild. Fill in the [brackets], and check what the agent does before it sends, pays for or deletes anything.