Human vs. AI – Diff-based line-level provenance for text under agentic editing (github.com)
46 points by eighttrigrams 13 hours ago
hamandcheese 5 hours ago
I'm curious why people care about such tracking? I hardly use an editor anymore, and I'm not alone. All that matters is who is signing off on the commit.
> Text a human wrote or edited should be considered close to sacred: an agent should be hesitant and have a very good reason to touch it.
Not sure I agree (particularly with code, not prose). If code is risky to change for reasons that aren't obvious, it should be commented as such. It doesn't matter how the bytes were generated.
> Another use case: the README.md, originally generated, where you rewrite the opening paragraphs. The agent should feel free to redo or append parts further downwards but should really think twice changing anything in the opener.
This, I understand.
eighttrigrams 4 hours ago
> I hardly use an editor anymore, and I'm not alone. All that matters is who is signing off on the commit. In such a context I won't distinguish between me and an agent. But same, almost no editor these days.
For coding, this is more useful for heavily vibecoded projects. Thb, for my own projects most likely I will end up using it only for specs and documentation, maybe in rare cases for some tests. But, I use it for agentic memory, where I can edit the memories as well, and the agent should tend to preserve my edits when updating them.
In general, my main use case is tracking texts in knowledge management systems.
nextaccountic 3 hours ago
It may not matter for you, but it could help agents to recover context about some specific code
spuz 9 hours ago
> A git repository is already a history of versions each carrying a provenance marker — every revision of the file, in order, with the author of the change that made it.
Maybe I'm missing how people use AI these days but when I have an agent working locally, all git commits have my authorship attached.
andai an hour ago
Claude's system prompt tells it to sign git commits with "co-authored by Claude [model]".
(Codex doesn't do that.)
GitHub renders this as two authors for those commits.
eighttrigrams 8 hours ago
I take care of that in two ways, 1) is via claude hooks, I make sure these are tagged as claude, 2) my sandboxed agents have git env vars set, with the same effect.
spuz 8 hours ago
Does Claude run git commit on every change it makes? Doesn't that pollute your git history?
eighttrigrams 8 hours ago
drusepth 8 hours ago
A /precommit tool can be very powerful -- not only can it run the tests, do light code review on its own, spin up sub-agents to review performance/architecture/etc impacts, but it can also write hilariously-verbose commit messages once everything passes, detailing not only what changed but why it changed.. tagged with Claude as a co-author.
alansaber 9 hours ago
Cool, we do the same thing, but we also denote when a line is "AI generated but was modified by a human" (aka human made anything upwards of a 1 character change).
eighttrigrams 8 hours ago
I'll probably be annoyed enough soon that I want that level of granularity, too
gritzko 11 hours ago
Interesting. I use annotations for that. But even if agents ignore them 1 out of 100 cases, I have enough cases here to see it happen.