AI coding agents on real mobile apps

Notes from running AI coding agents against shipping Android and iOS apps: multi-agent setups, review gates, and proof that a change works.

Working With AI AgentsDon't Trust an Agent That Says the Tests PassedAn agent's report is a claim. The setup I run only accepts evidence a machine or a second agent can check, and the cases where that caught something are on the page.October 21, 2026 · 9 min read
Working With AI AgentsWhen Your Coding Agent Hits a Rate Limit: A Fallback Ladder That Keeps WorkingA rate limit, a stalled run that already finished the work, and a pull request that still fails review need three different responses. Here is the ladder I use to tell them apart.October 21, 2026 · 8 min read
Working With AI AgentsTwelve Cores, Seven Devices, One DiskGit worktrees keep parallel agents out of each other's files. They do nothing about the CPU, the test phones, the disk or the run that hung seven hours ago. Those are the parts that actually broke.October 21, 2026 · 8 min read
Working With AI AgentsWhat Keeps Agent-Written Code Clean When Nobody Reads ItOne boundary rule lived in a review prompt and quietly broke in 25 files. The ones that held were unit tests that fail the build, and a second model that reviewed the plan by reading the repo.October 20, 2026 · 8 min read
The App-Rescue PlaybookThe Agent Taps the Phone. Nothing Touches the Mac.Writing code stopped being the expensive half; verifying it started. On iOS the last mile is a real device — and unlike Android, the platform does not simply let you in. Here is what it takes, and what it catches.September 9, 2026 · 9 min read
The App-Rescue PlaybookHow I Review AI Changes in a Production Mobile AppThe coder doesn't get to grade its own homework. A working method for putting AI on a production codebase — one agent that plans and reviews, one that writes, and durable state that lives in files instead of the chat.July 13, 2026 · 7 min read
The App-Rescue PlaybookHow to Know an Android App Audit Actually Found NothingA clean bug report is only good news if the process behind it was built to draw blood. Three bugs, one phantom, and the four questions that separate a real audit from theater.July 10, 2026 · 7 min read