Posted this on my LinkedIn; I think it can be helpful to you all:
If you really want to be at the vanguard of all this s**t, you need to think about the changes an AI-FIRST environment requires:
- How are you communicating the architectural decisions you want AI to follow? Why?
- Are you asking AI to document the process so the next agent doesn’t need to read the entire codebase?
- Are you shaping skills and commands so the next engineer can ask their AI to do something without a lengthy context handoff?
- Are you asking AI to create tests for its own code so it doesn’t accidentally introduce regressions?
Just as we don’t expect people to change binaries directly, we are rapidly evolving to a world where engineers will NOT touch code directly anymore. Getting ahead of that will help your agents to produce better code, better tests, better documentation, and, obviously, better software.
