zulip added slop to their codebase a long time ago (1, 2) but now they’ve released this bullshit blog post with some choice nonsense:

I seriously considered banning LLM use for Zulip contributions. But our view is that contributors should be allowed to use modern tools in the service of producing great, reviewable work. AI-assisted work is of course subject to the same rigorous review processes we’ve always used for community contributions.

So we decided to invest in creating, refining, and enforcing a new AI use policy, which has the following key tenets:

• End-to-end human responsibility for work and the communication around it. You always need to understand, test, and explain the changes you’re proposing to make, whether or not you used an LLM as part of your process to produce them.
• Clear and concise communication about points that actually require discussion. While we allow carefully edited AI-generated PR descriptions, we’ve had to ban AI-generated chat messages in the development community as too disruptive. Manual enforcement of this policy has been rough, with far more PRs closed without review, stern warnings, and outright bans of repeat offenders than we’ve ever had to apply before. (What do you do when someone apologizes for submitting AI slop… by copy-pasting an apology from ChatGPT, including surrounding quotation marks?) We expect that next fall, automation or other major changes will be required for the PR triage process to be manageable.

The results [of using Claude] were promising (and far better than just a few months prior) — enough for us to start investing in teaching Claude Code how to self-review its work, and how to produce PRs that are easy for maintainers to review. This has largely been an AI-supported process of digesting our contributor documentation into CLAUDE.md, and iterating when we see the model struggle.

i liked zulip 😞