79.4%
of agent-authored PRs are open concurrently with another agent PR
Exact temporal overlap (k = 0) across the AIDev-pop corpus.
Manifesto
Software will be built by fleets of agents.
That future is already here in the numbers. 90% of developers regularly use AI coding tools at work. 74% have already adopted specialized AI coding assistants, editors, or agents. 84% of developers use or plan to use AI tools in their workflow. The bottleneck is no longer whether agents can write code — it is whether dozens of them can share one codebase without colliding.
Long term, Managent is the shared control plane for that world: claim work before it conflicts, make waiting visible, verify merges are conflict-free, and keep every agent — laptop, CI, cloud — aware of the same territory. Not another autocomplete. The coordination layer agent-heavy engineering orgs will need when one developer runs many writers at once.
- JetBrains AI Pulse, Apr 2026 · 90% / 74%
- Stack Overflow Developer Survey 2025 · 84%
The evidence
We measured the problem first.
Our founders published the first empirical study of concurrent AI-agent pull requests on GitHub. 33,596 agent PRs across 2,807 repos. 747 merges replayed with git merge-tree. Published on arXiv.
19.8%
textual merge conflict rate among same-agent co-active pairs
601 same-agent pairs among 716 evaluable, from a 747-pair git merge-tree replay.
41.7%
textual merge conflict rate among cross-agent co-active pairs
Roughly double the intra-agent rate; 95% Wilson CIs do not overlap.
84.4%
of conflicted files are source code, not lockfiles
42% of conflicts are structural (modify/delete or add/add).
From the paper
Figure 1: Concurrent agentic PRs are the norm. 79.4% of PRs co-active at k = 0, rising to 95% within ±7 days. Cross-vendor pairs are rare (0.5%).
Full methodology in the PDF.
Managent is the coordination layer the data calls for.
Exclusive leases, a visible queue, and merge-replay verification. Watch the demos →