[jangorecki]

it actually seems to work: https://www.youtube.com/watch?v=bh6S4N8TnYQ

based on https://github.com/ghostty-org/ghostty/

[MichaelChirico]

Agree w the spirit of this but it seems like overkill. I don't think we've seen the same flood of "spammish" low-effort contributions as faced by projects like ghostty, zig, etc.

The AGENTS.md we already have seems pretty good to me, I would prefer we try and iterate on the spirit we've already laid out than to drop in such a blanket ban. That is:

• LLMs are OK for research, explanation, understanding
• LLMs must not be used to author data.table_x.y.z.tar.gz code
• LLMs writing GHA code is fine

What's new in your PR that's not well-covered by the current AGENTS.md, IINM, is a stance on authoring bug reports with LLM. Do we want to take such a stance? I only see 11 "external" issues filed in the last 4 months, regardless of quality; is it worth also blocking good use of LLM to write issues?

[jangorecki]

I don't think we've seen the same flood of "spammish" low-effort contributions

agree

still the point is to ensure that a human reads what is about to be submitted.

[tdhock]

Is there a way to ensure the PR sumitter looks at the Files changed tab before converting the draft PR to a real one and asking for review?

[ben-schwen]

Is there a way to ensure the PR sumitter looks at the Files changed tab before converting the draft PR to a real one and asking for review?

Don't think so and after all, ppl would probably just click the check marks away anyway

[tdhock]

too bad, that is my number one issue these days (seems like rookie PR submitters don’t look at their own diffs for obvious issues)