6 ms·

I believe Codex harness is pretty sticky though I haven’t tried many others. Does anyone else provide a harness of that quality?

• conradludgate 2mo ago
  Their harness is indeed nice, but we're using it against our own AI Gateway at work. Right now we only expose GPT models on it, but I imagine it's possible to add our own models eventually
• nvarsj 2mo ago
  Pi harness with subagents is pretty great. I’m using it for everything now.
  • yeswecatan 2mo ago
    How are you handling subagents?
    • anderber 2mo ago
      Usually the main agent can handle creating subagents
    • RALaBarge 2mo ago
      You say hey llm spin up some agents to do x y and z
    • nvarsj 2mo ago
      tintinweb/pi-subagents. Then I use superpowers and it spins up subagents automatically. Works really well and you can customise the models and efforts you want to use with local AGENTS.md. I’m primarily using gpt-5.6 and then opus for reviews.
• verdverm 2mo ago
  opencode is highly regarded, I use it exclusively
  • jki275 2mo ago
    I've used open code extensively -- it's not bad, but it's not anywhere near the level of the codex application.
  • versteegen 2mo ago
    ...but we're talking about compaction, and opencode's compaction is (or was) terrible. I've seen so many horrible problems that I keep it disabled (with an envvar flag, because even the config flag to turn it off was broken).
    • verdverm 2mo ago
      I don't use compaction, I keep my context limited, my sessions fresh, and have agents write markdown files as needed
  • Roark66 2mo ago
    Open code is cool once I added the ctrl-o function to it (show thinking and command outputs at will instead of on by default), but sadly I don't think it got merged by the team. I forgot why. But I can't use an AI cli without that feature.
    • verdverm 2mo ago
      you can turn those on through the existing menu system