108 #genai #declared

This is a big topic* in our community right now, and as Dribdat now also has recently had a first PR with purely generated code (yes, I'm very late to this party), I'm reaching out for opinion on both how to properly disclose - and support - working with AI at hackathons.

To maintain Dribdat I have relied Dependabot and FOSSA, along with various coding and continuous integration tools for years. Software fertilizer is hard to avoid these days! Codeberg's position seems clear that 'software is about humans', but someone is already working on Forgejo MCP support to allow Agents to interact with such sites. If there is an open source generator they endorse, I would certainly give it preference.

The main issue seems to be the loss of control. As long as a human reads and approves every line of every PR, it's a mitigatable concern. Try dealing with AI agents and their minions turning your backlog into an issue swamp. The other tricky one is the creative disenfranchising of millions of developers through coercive techniques on gamified platforms like StackOverflow and GitHub, the lack of foresight from OSI and other organisations in the legal area. We are only just beginning to use AI opt out - a.k.a. content exclusion controls, and that may feel too late for many people.

Very importantly we need to make clear which parts / commits are from a bot / agent, and take some time to understand the consequences** of code generation. I found the discussion at ghostty about disclosure quite relevant.

I can make sure that in our PRs and releases this is stated prominently. As you can see above, Jules and I recently added support for CC Signals, a prominent initiative in this direction. I'd also like to add a way to make GenAI commits visible and accounted for - see dribdat#426

Hackathons should respond to this. If we have an opportunity in public sprints to give some healthy advice to the next generation, and lead by example, I'd like to take it. So if you have any good examples or "betterHack" practices, I'm all ears.

The Git log doesn't lie.

* e.g. Nancy Law et al 2025 among others referenced in our hackathons library
** Petropakis et al 2024 etc.

Discuss this with us on GitHub: