https://quellbot.dev/blog Notes on AI code review, coding agents, and the verification half of the dev loop, from the people building quellbot. en https://quellbot.dev/blog/ai-teammate-github-issues-into-pull-requests https://quellbot.dev/blog/ai-teammate-github-issues-into-pull-requests Fri, 04 Sep 2026 09:00:00 GMT quellbot turns GitHub issues into pull requests: plan, build and test in a throwaway machine in your own Fly.io account, pass a policy gate. It never merges. https://quellbot.dev/blog/ai-code-review-pull-requests https://quellbot.dev/blog/ai-code-review-pull-requests Fri, 04 Sep 2026 08:00:00 GMT Most AI reviewers get muted within a week. A verdict, severity grouping, committable suggestions, and delta re-reviews are what keep one on the team. https://quellbot.dev/blog/why-an-ai-coding-agent-should-never-merge https://quellbot.dev/blog/why-an-ai-coding-agent-should-never-merge Fri, 04 Sep 2026 07:00:00 GMT Prompt injection, egress, and the policy gate: the security model behind an agent that has write access to your repositories and still cannot hurt you. https://quellbot.dev/blog/bring-your-own-key-ai-dev-teammate-cost https://quellbot.dev/blog/bring-your-own-key-ai-dev-teammate-cost Fri, 04 Sep 2026 06:00:00 GMT No per-seat pricing and no markup on model spend. A plain breakdown of what a run costs: your model credential, about a minute of Fly Machine time, and one flat plan for private repositories. https://quellbot.dev/blog/failed-github-actions-root-cause-triage https://quellbot.dev/blog/failed-github-actions-root-cause-triage Fri, 04 Sep 2026 05:00:00 GMT When a workflow fails, quellbot reads the failed job, step, and log tail together with the diff and posts the likely root cause and a next step, on the PR or on the deploy commit. Advisory only.