Operum: two founders running six-role AI agent teams merge 296 PRs in 12 days, each gated on an AI tester approving the exact commit
Operum's two founders build their desktop app, which gives users an AI team of PM, Architect, Engineer, Tester, Marketing and Community roles working through GitHub, using the app itself. Between 15 and 26 October 2026, 296 pull requests merged into the product repository, about 25 a day. 280 of them needed green CI plus the AI Tester's approval of the exact commit being merged; a new push resets the approval. Sorting by title, about 49% were user-visible product, 21% non-visible runtime, 27% internal process (tests, CI, refactors, audit records) and 4% marketing. The founders say their time went to deciding what to build, reading the Tester's 'not covered by any test' lists, and merging.
The numbers (as reported)
| Merged PRs | 296 in 12 days (15–26 Oct 2026) |
|---|---|
| Merge rate | ~25 a day, ~12 per person per day |
| PRs gated on CI plus AI Tester approval of exact commit | 280 of 296 |
| Mix by title | ~49% user-visible, ~21% non-visible runtime, ~27% internal process, ~4% marketing |
Facts
- Company / subject
- Operum
- Tools
- Unstated
- In-house tools
- Operum
- Codebase
- brownfield
- Work type
- Feature
- Stack
- GitHub
- Published
- 2026-11-06
- Last checked
- 2026-11-11
What the source claims Practitioner (own blog/paper)
Claimed by the builder, writing about their own work. Not checked independently.
- 296 pull requests merged in 12 days by two people running AI agent teams 296 PRs in 12 days
- About half of merged PRs were user-visible product changes ~49% user-visible
Caveats
Written by a co-founder about building Operum with Operum, in a post leading into its Product Hunt launch, so it promotes the product. The PR categories come from titles, not a diff audit, and the author says the PRs are small and PR count is the wrong scoreboard. Cursor is named as the app's default engine and recommendation; the post does not say which engine the team used for these 296 PRs, so the tool is Unstated. Revert and rework rates are not reported.
More stories by tool: Unstated
Related topics: Brownfield codebases