A GitHub agent repo has more open PRs than issues

letta-ai/letta-code has 3,257 stars and splits its 353 open items into 219 pull requests and 134 issues.

That split matters for anyone deciding whether to depend on the project. A backlog of unmerged pull requests usually means reviewers are behind, not that the project is thriving.

Most agent repos run the opposite way, with issues outnumbering PRs. gentle-ai has 240 pull requests open against 679 issues, traycer 29 against 167, voicebox 172 against 513. letta-code's 219 PRs against 134 issues flips that pattern.

There are two plausible readings. Contributions could be outrunning the maintainers' capacity to review them, which happens on repos that get sudden attention. Or a bot is filing many of those 219 PRs. Automated dependency bumps or codegen submissions would inflate the count without reflecting real review demand. The repo's public numbers don't say which.

Star count won't settle it, but merge latency will. A repo where PRs stack up because reviewers are backlogged behaves differently than one where a bot files and something auto-merges within hours. Both would produce this same 219-to-134 split.