@David_Thomas Gen (David get her on the forum!) Colin @sudokita maybe @Steve-TechMatters and anyone else…
AI is changing coding… like super duper fast. The large positive impact is that people who have domain expertise can now (maybe / probably / ?) begin meaningfully contributing both to the definition and actual coding of real issues.
That should / could / ? super power open source codebases (and potentially overwhelm senior devs as reviewers) and advance those applications really fast. However…
- Non-developers don’t have contribution pathways built for them… and
- Maintainers could be overwhelmed or simply unpreprared for the type of contributions provide
I’ve been thinking about this as David Thomas (a definite domain expert on OFN) into OFN to see what he can do. So I asked Claude, and it gave a very thoughtful and useful answer IMO. Worth reading, not just ai puke.
https://claude.ai/share/c0c69fe8-954f-43ed-8892-e1787d61dc3c
It made interesting points to my original question, and expanded it to suggest that the contribution process as a whole needs revisiting, not just better tagging. It’s suggestions are (with my comments):
- Adding Persona and Functional Area as custom fields in their GitHub Project
- This helps a non-developer user more easily find issues related to them they care enough about to tackle (ie “super admins” or “generating reports for farm hubs”). This is key to a motivating entry point and useful issue queue for them.
- Create 3-4 issue templates that guide non-developers through structured submission
- I would even suggest that you could have an AI (which has access to the code base) initiate a back and forth with the user (like claude terminal does now) to clarify items that may be missing. I find my first query is 50% of the value, and claude’s first round of follow up questions often get me to 90%.
- Propose an
AGENTS.mdorCLAUDE.mdfile that maps functional areas to codebase locations.- This is really really critical. If each project (like OFN) has a default
agents.mdfile, then a non-developer contributors can pull up claude code and claude can utilize that out of the box. It may include best practices, places in the code to look for certain functions, overall architectural / design patterns not explicitly written out, etc. Without this file (which should be written and maintained by the primary OFN developers!!!) contributors are likely to produce crap code out of the box, and ultimately they’ll need to recreate these types of files themselves to meaningfully contribute anyway.
- This is really really critical. If each project (like OFN) has a default
- Suggest a trial period — maybe tag the existing open issues (they have about 609) with these new dimensions and see whether it actually changes what people choose to work on.
What do you think… How should we create pathways for ai assistent non-devs to contribute? and are there existing best practices for this? What is different than what we’re doing? What type of code bases most benefit?
We could even have a domain expert do a dry run on a codebase and get feedback on what worked and what didn’t. For us, I could see someone Juliet do this for Surveystack (as an example of a domain expert and user who’s not a full stack developer on this tech stack per se.). Maybe you have someone like that in your code base…