Turn GitHub issues into draft pull requests with long-running agents
The author used Dots for read-heavy long-horizon tasks like reading GitHub issues and making PRs, which were typically poor but gave a starting point.
Done withDotsCodexClaude Code
- What they did
- The commenter used Dots, a long-running agent tool, alongside their Codex usage for read-heavy, long-horizon tasks. The agent read GitHub issues and drafted pull requests from them. The source gives no further setup detail, and no step where a person intervened is described.
- How it went
- The pull requests were typically pretty bad, but they gave a starting point. The commenter got much better real-world output from a Claude Code style approach with a human in the loop.
- Worth knowing
- Running it against regular plan limits was a dealbreaker for the author, who also wanted more control over the context window when the agent drifts off task.
Try it yourself with Dots
Read the open issues in my GitHub repo [repo URL] and pick the [number] most well-defined ones labeled [label or topic]. For each, make a branch and prepare a draft pull request with a first-pass fix, tests if the repo has them, and a short description of what you did and what you're unsure about. Don't merge or push anything beyond draft PRs, and show me the list of PRs you plan to open before you open them.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: HN · Oct 21, 2026
Related use cases
Turn tickets into merged pull requests
Pulls tasks from GitHub Issues or Linear, then runs Claude Code or Codex in isolated Kubernetes pods to carry each one through to a PR.
Compose songs by writing synthesis code
The author ran one-shot experiments where Claude Code wrote Python to synthesize music, including a rock song with sung lyrics via macOS say; sampling a song failed.
Restyle a pull request to match repo house style
A developer asked Claude to update a PR to the repo's house style, but it only rewrote comments to be lowercase and less useful, making no code changes.
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Reverse engineer legacy systems without source code
Thoughtworks consultants used Gemini 2.5 Pro in a two-week pilot to produce validated blueprints of legacy systems' functionality.
Use a bot as a QA intern on a live app
A Grok Bot clicked through an app's core flows several times a day, reported problems and drafted fix PRs; it found a real bug.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.