{"slug": "my-ai-coding-workflow-revisited", "title": "My AI Coding Workflow (Revisited)", "summary": "A developer's AI coding workflow now regularly merges 170+ pull requests a day, up from the 100-PR milestone reached after a February 2025 workflow snapshot, with roughly 60% of merged PRs now automerged versus about 25% at the 100-PR mark. The workflow runs around twenty concurrent tasks but caps open PRs to create backpressure, uses a dashboard bot to gather CI/CD status and comments, and has a separate agent rate each PR's risk to decide whether it gets \"Merge When Green\" or human review. The author reports that zero-touch changes such as typos, help documentation, visual quirks and dependency bumps rose from about a quarter of October merges to closer to 60%.", "body_md": "I wrote a snapshot of my [AI coding workflow](https://jonathannen.com/my-current-ai-workflow/) back in February, as part of the push to [100 PRs a day](https://jonathannen.com/building-towards-100-prs-a-day/). It was explicitly point-in-time, so nine months on, here's the next one. Since then I've [hit the hundred](https://jonathannen.com/one-hundred-and-fifty/), and 200 in a day is now a normal day.\n\nThe core aim hasn't moved: reduce \"touches\", the number of times I have to pick up a change before it ships.\n\n- **Zero** touches for simple changes: typos, help documentation, visual quirks, dependency bumps. These were about a quarter of what merged in October; they're closer to 60% now.\n- **One** touch for a medium feature. I jump in, review, maybe tweak, and away it goes.\n- **More** touches for security changes or things the whole team needs to be across. A separate agent rates each PR's risk, from \"a robot can approve this\" up to \"three brains\"<sup>[\\[1\\]](#fn1)</sup> , and that rating decides which bucket a PR lands in rather than me.\n\nWhat I didn't cover in February was how the day itself is structured, and that has changed more than the flow has.\n\n**The day.** Half of it goes to delivering and half to the harness: prompts, skills, lint rules, CI, the tooling around the agents. That half lives on the [Axe Branch](https://jonathannen.com/sharpening-the-axe/), and it is where the touch count actually comes down. Overnight is for long-running tasks that can burn quota without my attention.\n\n**Concurrency, with a cap.** I run a lot of tasks at once, around twenty, but only five to ten [epics](https://jonathannen.com/more-attention-not-less/), so most context switches land somewhere I've just been<sup>[\\[2\\]](#fn2)</sup>. The new part is a hard limit on open PRs. Once the queue is full the agents wait for me rather than the other way around, which is the [backpressure](https://jonathannen.com/the-importance-of-backpressure/) I didn't have in February. When I find myself re-reading a PR, the rule is now [\"never pick it up twice\"](https://jonathannen.com/never-pick-it-up-twice/): something upstream needs fixing, so stop and fix it.\n\n**Self-sufficient agents.** A growing number of processes run without me at all. Support documentation is one, patching is another, and it now stretches to fixing customer-reported bugs. That is a big topic, so I'll only touch on it here and write it up properly soon.\n\nThe flow itself, right now:\n\n1. \nStart every task in Claude Code/Web/Desktop, with images and annotated screenshots. No plan mode, no swarms; the \"axe\" work goes into making the first prompt land <sup>[\\[3\\]](#fn3)</sup> . Smaller PRs are the other unlock, since the hard part of many streams at once is managing the context.\n2. \nA dashboard \"bot\" gathers everything about each PR: CI/CD status, whether there are comments and which are resolved, and the types and directories the change touches.\n3. \nI've pushed the coding agent hard to produce PRs that are either straight-through or very easy to review. Screenshots, a short video of the feature end-to-end, a testing script, whatever is needed is added. All of it is one click away in the dashboard.\n4. \nAnother agent assesses each new PR. Simple ones are marked \"Merge When Green\" and go live when CI/CD passes. The rest get a review from one or two separate agents.\n5. \nClaude or Codex responds to that feedback and resolves the comments it can. Anything it can't stays open and I'm notified.\n6. \nOnce every comment is resolved, the assessment runs again: can it be \"Merge When Green\" now?\n7. \nIf all else fails, I review. For a visual change that might be a glance in the dashboard. Sometimes I read the diff. For the complex cases I pull it down and review locally.\n\nWhere are we at? I regularly merge 170+ PRs a day. When I hit 100, the automerge PRs were about 25%, now it's closer to 60%.", "url": "https://wpnews.pro/news/my-ai-coding-workflow-revisited", "canonical_source": "https://jonathannen.com/my-current-ai-workflow-revisited/", "published_at": "2026-11-09 00:00:00+00:00", "updated_at": "2026-11-18 21:41:43.092132+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools"], "entities": ["Claude Code", "Claude", "Codex"], "also_reported_by": [], "takeaway": "Capping open PRs so agents wait on the developer, plus a risk-rating agent that marks simple PRs \"Merge When Green,\" lifted automerged PRs from about 25% at 100 PRs a day to roughly 60% at 170+ a day.", "recommendation": {"advice": "Cap the number of open agent PRs so the queue applies backpressure, and route each PR through a risk-rating agent that automerges low-risk changes when CI passes.", "playbook": "coding-agents", "playbook_url": "https://wpnews.pro/playbooks/coding-agents"}, "alternates": {"html": "https://wpnews.pro/news/my-ai-coding-workflow-revisited", "markdown": "https://wpnews.pro/news/my-ai-coding-workflow-revisited.md", "text": "https://wpnews.pro/news/my-ai-coding-workflow-revisited.txt", "jsonld": "https://wpnews.pro/news/my-ai-coding-workflow-revisited.jsonld"}}