TIL about the Gemini CLI. I’ve been having Claude code hack at a thorny problem recently, and keep running into walls, so I figured I’d try something else
On first pass, it performed worse, but also revealed a shortcoming in my own harness. I have a lot of work/learning I need to do to get good at giving the agents the tools they need to test their own success.
In this case, that means setting up an easy expo build command for checking if the native code is building correctly