
Joshua Poddoku
Signal over noise for Open Source ecosystems
I read what your community is saying, and what arrives in your repo, and turn it into decisions.
Seven years inside pre-seed to Series A startups.
- Signals
- The read
- Decisions
Pull requests and issues now arrive faster than maintainers can read them. I use AI on the other side of the same problem: contributor dashboards, onboarding, and reviews that teach rather than reject. A person still decides.
The dashboards are the part I do in public, as an Apache committer on Apache DevLake.
Projects rarely lose people over the code. They lose them because there is no clear way in: nothing in the docs saying how to start, nobody triaging what a newcomer opens, no reply when someone tries. People pick a tool for what it does, and stay for how it is run.
With a business on top of the project, that gets sharper. Velocity, funding and investor expectations land on the same maintainers, and unless the volunteer work and the commercial work are told apart, the community absorbs the cost of all three.
Worth reading right now
- ResearchStudy on developers attention to AI code securityPeer research on whether developers verify AI-generated code security.Bluesky
- PractitionersDefensible vibe coding runs generated code through CIConcrete pipeline for reviewing AI code safely.Bluesky
- PractitionersLegal risk of AI code in decompilation projectsExplains cleanroom and training-data provenance concerns.Bluesky