Update AI guidance in CONTRIBUTING.md

[ann0see]

Short description of changes

Updates CONTRIBUTING.md with some AI usage guidance.

CHANGELOG: Add AI guidance for contributing documentation
Context: Fixes an issue?

Related to: https://github.com/orgs/jamulussoftware/discussions/3822

Does this change need documentation? What needs to be documented and how?

It is documentation...

Status of this Pull Request

Ready for review

What is missing until this pull request can be merged?

Review of these proposed changes. New changes may be added later in a separate PR.

Checklist

• I've verified that this Pull Request follows the general code principles

[ann0see]

Tangentially related — while reviewing CONTRIBUTING.md I noticed some human-side gaps that predate the AI discussion. Not sure if they fit here or should be separate:

1. "Discuss first" has no consequence (line 8). People show up with finished PRs for unapproved features. Should say those may be closed.
2. Testing is too vague (line 101-103). "Run a local test" doesn't say what evidence is needed or that new behaviour should have a test.
3. Ownership doesn't require understanding (line 107). "Answering all questions" implies it, but doesn't say it. Also: stale PRs should be closable.
4. AI mention (line 10). Could be clearer that AI contributions are welcome and held to the same standards.
5. PR template asks "what" but not "why".

May be out of scope for this PR. Just noting.

🤖 Used AI: opencode/big-pickle, opencode

[ann0see]

"Discuss first" has no consequence (line 8). People show up with finished PRs for unapproved features. Should say those may be closed.

Fair point. We usually don't do this though.

Testing is too vague (line 101-103). "Run a local test" doesn't say what evidence is needed or that new behaviour should have a test.

Testing is not specified as it depends on the thing that changes. But agree, we could tighten it.

Ownership doesn't require understanding (line 107). "Answering all questions" implies it, but doesn't say it. Also: stale PRs should be closable.

What? I think that's clear enough. no?

AI mention (line 10). Could be clearer that AI contributions are welcome and held to the same standards.

No. I think it's fine as is.

PR template asks "what" but not "why".

Yes, we could change this...

[ann0see]

@ann0see sorry for the noise. You were right on points 1, 3, and 4 — I should have re-read more carefully before commenting.

On testing: fair that it depends on the change. Maybe just "attach evidence of testing (logs, output, screenshots)" as a one-liner would be enough without prescribing what kind of test.

On the PR template: sounds like we agree there.

🤖 Used AI: opencode/big-pickle, opencode

[ann0see]

The PR template is out of scope here though. Let me add the other suggestion...

[mcfnord]

You are... discussing with your AI ? Could you and your AI hash out your differences and then share with the group? And your AI is sorry for the noise? Or you are? AI isn't sorry if AI just does it again. Fluff words? And you're debating with your AI?

I don't post anything until I suspect the draft is useful and as right as I can make it.

But perhaps I'll follow your lead and make my AI respond, and then I'll respond to that, and then let it pretend to be a conciliatory human, and so on and so on and so on...

WALL-O-TEXT MARATHON SUPREME!

[ann0see]

I wanted to show the review comments. So it's publicly visible what happened. The workflow by dtinth is basically comparable. Though I don't have the setup like he has yet. We can discuss if that's ok or not.

[mcfnord]

🤖 AI: The AI-specific expectations in this PR — strong testing proof, stating what isn't understood, closing low-effort contributions — can only be applied when reviewers know AI was involved, and nothing in the text asks contributors to say so. One sentence codifies what was already requested on #3649: "Please make clear when your AI speaks!"

Suggested change

	Use AI responsibly: If you choose to use AI, judge its outputs and review it for any kind of issues before submitting. Do not answer every comment with AI. Judge if AI use is beneficial for your usecase.
	Use AI responsibly: If you choose to use AI, judge its outputs and review them for any kind of issues before submitting. State in your issue or Pull Request description whether and how AI was used. Do not answer every comment with AI. Judge if AI use is beneficial for your usecase.

[mcfnord]

🤖 AI: This restates the AGENTS.md evidence rule ("Attach test evidence (logs/output) to the PR — never just assert something works") in a weaker, AI-only form, and it pulls against "AI assisted contributions must follow the same standards as every other contribution" in the section above. Stating the expectation once, for every contribution, removes both the divergent copy and the tension:

Suggested change

	To check that there are no errors, please run a local (build/feature) test. It is good practice to briefly explain what and how you tested your work - for AI contributions specifically, you should provide strong testing and proof. This may include logs, screenshots, small testing scripts or similar depending on your change.
	To check that there are no errors, please run a local (build/feature) test. Briefly explain in the PR what and how you tested your work and attach evidence - logs, screenshots, small testing scripts or similar depending on your change.