This is interesting. It would be even better if ones coding agent also got better at its coding abilities every night. Otherwise, agents might re-introduce the same class of bugs that your nightly improvements keep reacting to.
This would require logging agent actions during the day, analyzing it at night, identifying things the coding agent could do better, and update its AGENTS.md memory (relatively easy) or fine-tune the agent (if local LLM, and the fine-tuning cost was acceptable).
It would be nice that when a person corrects an agent's mistake this was captured and fed back into a nightly agent improving process.
Yeah, what even is the point of a "chronometer" (you don't have a clock elsewhere?), "data stream" (???), "radar visualization" (???), "depth field" (???)
That's a nice idea but I hate how every instruction step in the README is just "Ask your agent to do it". Not only does it make it harder and more time-consuming to understand how the general system is implemented, but it also dumbs down users. Do you think people checking your repo wouldn't be able to copy/paste a few commands or find an equivalent on their system?
that's the neat part, there is no prescribed system or code. your agent decides the implementation which you control, the repo is mostly natural language (save for the presentational tui)
Earlier this summer I vibe-coded a harness to keep grinding at a research question I'm working on over the weekend. I asked Claude to name it something catchy, and it chose the name Nightshift.
This is interesting. It would be even better if ones coding agent also got better at its coding abilities every night. Otherwise, agents might re-introduce the same class of bugs that your nightly improvements keep reacting to.
This would require logging agent actions during the day, analyzing it at night, identifying things the coding agent could do better, and update its AGENTS.md memory (relatively easy) or fine-tune the agent (if local LLM, and the fine-tuning cost was acceptable).
It would be nice that when a person corrects an agent's mistake this was captured and fed back into a nightly agent improving process.
great instinct. that's exactly what happens today! it learns from closed PRs and feedback on what changes are high value vs not
The UI looks fun but reminds me of the hours and hours I spent tweaking my linux distro as a younger nerd.
It didn't help me get anything done, but it sure did look computery!
Yeah, what even is the point of a "chronometer" (you don't have a clock elsewhere?), "data stream" (???), "radar visualization" (???), "depth field" (???)
That's a nice idea but I hate how every instruction step in the README is just "Ask your agent to do it". Not only does it make it harder and more time-consuming to understand how the general system is implemented, but it also dumbs down users. Do you think people checking your repo wouldn't be able to copy/paste a few commands or find an equivalent on their system?
that's the neat part, there is no prescribed system or code. your agent decides the implementation which you control, the repo is mostly natural language (save for the presentational tui)
It would be neat if you can give a general idea what it tries - sets up a python web dashboard on this port and launchd to run off-hours etc.
https://imgflip.com/i/b0vp88
systemd service to write "opencode run --session <id> continue, make no mistakes"
...were you not prompting .e
Earlier this summer I vibe-coded a harness to keep grinding at a research question I'm working on over the weekend. I asked Claude to name it something catchy, and it chose the name Nightshift.
Does it have nightmares? =)
occassionally when you forget to git commit/stash before sleeping
how would it be different from a skill plugin?
This is really cool!
Jarvis is that u?
what are the odds we post a Show HN on the same day with the same opening title lmao: https://news.ycombinator.com/item?id=49643048
great minds. we gotta collab, dm