HUMAN SIMULATING END TO END APP TESTING

Your coding agent should be doing human simulating end-to-end app testing. Having your agents run unit or integration test on the code that they've written for your app does not test your app. Instead, you should think about number one, what surface will a human or agent be using your app on? Some surfaces include a web app, an iPhone app, a desktop app. Coding agents now have tools that allow them to manipulate all of those surfaces. Use something like the DisplayPort CLI for browser testing. Use something like native computer use tools for testing desktop apps and use something like the code MCP for testing iPhone apps. Number two, within your app surface, what is the branding tree of potential user flows? Each decision that the user makes in your app, clicking a different button, navigating to a different page, branches a new branch off of that user flow tree. Step three is to combine these two ideas together to enable human simulating end-to-end app testing. Now, instead of your coding agent just running unit or integration test that it probably wrote to overfit and pass those tests and it probably didn't count for all the edge cases. It has a full map out journey of all the potential branching user flows through that surface of your app and it has the tools to itself navigate through all of those user flow branches simulating human use of your app. This isn't to say that unit and integration testing is not important. It can really help your coding agents during development and can help you consistently prevent regressions as you add more code to your code base. If you're building a car, you don't just put it on a big threadmill and connect a computer to the electric steering and repeat the same three steering movements over and over, you actually let a human drive it through different condition and let that human introduce edge cases and bugs that you wouldn't have caught. The same way you can have your agent do that with your app. As always, link in bio for more sauce.