Product, engineering, trust & safety, design, and the broader phone-automation category. Written when we have something to say; not on a content schedule.
The phone is the source of truth, not the server. Here's how we queue workflow runs locally, replay on reconnect, and the one architectural mistake that took six months to fully unwind.
Per-grab pricing was the obvious model and the wrong one. The four-month walk we took to a flat $9 tier, the three experiments along the way that failed, and the customers we deliberately price out.
Twelve months of triaging reports to abuse@. The categories we expected, the ones we didn't, the false-positive rate, the four denylist additions that came directly from this channel, and one report that changed how we think about review SLAs.
Phone automation has a trust problem and a design problem. We're trying to fix both, starting with a product that says no when no is the right answer.
Defense-in-depth for the categories of automation we refuse to enable: server validation, marketplace review, and a compiled-in on-device gate.
The TypeScript builder simulator and the Kotlin on-device engine must agree to the byte on every fixture. Here's the test rig.
Google's policy for Accessibility apps tightened in 2023. Here's how we think about it, what we ship in the disclosure flow, and why we always plan for Play Store removal.
Why we adopted a Wix-flavored visual system, how pill buttons changed our marketing copy, and the tokens we kept vs. dropped.
