OpenHands

image.png

Wow OpenHands Planning has strict boundaries

yeah, tbh I probably won’t use it because I find planning w/o action of limited value. but if you’re using openhands locally it might have benefits sometimes

I saw it available so i was like oh i gotta try it but usually i use our coding agent to plan tbh

Recently, Sonnet 3.7 was retired, and I had a nostalgia day looking around. I think this was the LLM that had most of a built-in need to write plans, docs, .md files, to plan a "comprehensive" thing before doing and then jump all over the place doing it

It was the best coding LLM last year, last year at this time. I think maybe that's when planning like this was more valuable, because it kinda mapped to that SOTA LLM, it was the part of the harness that seemed to help it

Maybe planning this way will see a revival with sub-agents, though, idk, maybe it already sort of does

:sweat_smile: I just did it again.
• start conversation without repo
• switch to Plan Mode
• "clone yourself these three repos, understand the first two, we want x"
• GPT-5.4 started to look around a lot, including searching deepwiki and browsing :joy:

does anyone know if it is possible to use planning mode in the CLI? I am late to realizing that there is a planning mode, but I can't figure out how to enter it

<@U0AJC23HB8T> If this is something you want to see in the CLI I made an issue with a possible implementation outlined <https://github.com/OpenHands/OpenHands-CLI/issues/613>

that's perfect! thank you very much <@U0AM1NT36R2> -- really pumped to try this out