Developers of popular open source projects are inundated with pull requests. What about a solution somewhere in between a user feature request and a code diff? User writes a feature request, AI checks the project code and turns the idea into a summary that highlights pros/cons based on the codebase.

@manton Great insight. Taking this a step further, if there were a layer that turns a user request into a feature plan and lets you apply it in one tap, GitHub could become a whole new kind of platform. Maybe managing open-source projects wouldn’t be just for geeks anymore.

@manton LLMs have turned me into a much more frequent contributor, which I generally think is a positive. But I often wrestle with trying to make sure my changes are addressing general problems so it doesn’t seem out of left field or bespoke when I open the PR.

I’m not sure if creating an issue first would be better or not, but for now just do my best to keep things narrow and well documented (though there’s tension with LLM verbosity). The waiting game can be stressful though.

@dongli.aeyo.app GitHub is nerdy but I’ve often thought more non-developers should use it. It’s a great collaboration platform.

@manton I have found that by forcing the model to be concise, fact-checked against source and by using a human skill to write the PR, the result is much much better, less mental overload for the maintainer/reviewer, it allows me, for example in the case of indiekit, to contribute back to an open source project 6 months of fixes, features I have been developing on my forks, before LLM’s I just couldn’t contribute in any meaningful manner and it was very frustrating

@manton I have also seen in other repo’s where there is some variant of OpenClaw or other Agent at work, doing more or less what you suggest, I have also seen repo’s where if you don’t abide to a specific template, if you don’t provide concrete line to changes, diffs in the comment of the PR, the PR is simply discarded by an Agent, this force contributors to abide to upstream repo best practice and once you do it, Agent will auto-review your PR, improve on it or sometimes rewrite the fix on their own based on specific in-house guidelines, which allows the contribution to move forward.

@fgtech You might be right. I was thinking it would help to have a concise summary based on implementation details but without the mountain of code to review.

@manton Ah, I see what you mean. I think the issue is more along the lines of too many requests that also happen to be not good enough for the code base. Each project is unique, though! What you suggest is probably useful for someone.