Thinking about this essay again this morning.

I tried to do a more complex coding task on Work Claude, and found the experience stupefying: On personal Claude things are so tuned I can almost one-shot a working iOS or macOS app and deploy it ad hoc or notarized from my phone. On Work Claude we’re back to “there’s no such endpoint in that API, but that’s just me guessing because I didn’t read the docs.”

Personal Claude is the beneficiary of a couple of years of steady, quiet, mostly unseen work to get it into the right place.

The answer, of course, is more Claude to get Personal Claude to help Work Claude understand how to Claude Like Mike Likes.

AI haters and resisters should read this essay and send it to the “ur doing it wrong” people in their lives. I’m not a hater or resister and still value it: When I’m working with resisters at work, it reminds me to slow down and go deep on the idea of context in its many forms.