I lead engineering teams building internal apps with applied AI at Parexel AI Labs. I write here about the things I keep coming back to: using LLMs where they actually help, treating internal tools as real software, and leading engineering teams without losing touch with the code.
If you're wondering whether to write something down, that's your answer.
Every context switch carries a hidden ramp-up tax.
Smaller pieces of work create momentum and surface problems earlier.
The new pull request dashboard organizes PRs by what you actually need to do with them.
Make `make` your default build task and stop switching to the terminal.