@dzhng: This thing has gotten 1.8k github stars in 2 days (cool!), and I've gotten a ton of people making PRs and filing issues…

This thing has gotten 1.8k github stars in 2 days (cool!), and I’ve gotten a ton of people making PRs and filing issues (awesome!)

It’s clear that being an OSS maintainer in 2026 is very different even compared to 2025. Basically 100% of the things filed are AI generated by other people’s coding agents

First - I don’t mind that at all. There were a lot of issue and PR spam, but there was also a good amount of useful ones too.

BUT, I do wish there’s some standards established & enforced, so that I know exactly what models & harnesses were used to create those PRs, and whether those PRs were created using the software factory stack that is already defined in the repo.

I don’t really read code anymore, and I definitely don’t want to read random PR’s code, but if I know those PRs were made by good models following my existing software stack and it went through the entire testing & verification loop, I’d have no problem merging it.

I feel like with the right process guardrails (maybe uploading traces to verify process is followed?), we can actually make autonomous self updating oss projects a reality. I love the idea of the community contributing their tokens to an OSS project and have the project essentially grow organically in a decentralized way.

Feels like there’s a layer missing on top of code hosting here. Essentially some sort of autonomous coding agent collaboration layer.

David (@dzhng): Introducing jevgrep - a research agent CLI powered by jev from @typesafeai that reduces your coding agent cost by 40% (verified on SWE-bench)

Make sure to use the built in skill so your coding agent knows to use jg for context collection