6 Comments
User's avatar
Daniel Vos's avatar

The toll versus build distinction is the sharpest way I have seen anyone cut this, and "a build you never weed is a slower leak" is going to stick with me. The trap I keep watching people fall into, myself very much included, is treating every session as production when the first frustrating one is really reconnaissance. The stranger wearing your name is not a failure, it is the cheapest diagnostic you will get all week, telling you exactly which file to fix. And decide, then delegate is the whole discipline: the moment I let the tool draft before I had made the actual call, I stopped getting a faster decision and started getting no decision plus a repair bill.

Kate Peabody's avatar

This comment is its own article, Daniel 🙏. Thank you for reading so closely and giving it back this way and "no decision + a repair bill" is exactly the tax nobody's naming. The reconnaissance point is the one people skip past fastest and it's the one that actually saves them.

Dr. Michael Meneghini's avatar

More time doesn’t always mean more capacity

Dennis Berry's avatar

Sometimes the real constraint is cognitive load, not time, and the body usually notices before the calendar does.

Cory Blumenfeld's avatar

sometimes the tool isn’t the problem it’s the looping around the tool that gets expensive.

Steve Tornello's avatar

"Human intelligence first. Always. The tool comes after."

You just described my entire newsletter better than I have.

It's so true. Without putting the work in first, all you're doing is prompt rolling until you get something you like you think. All it takes is a vision or a brief, something to give you a north star for the AI to work towards. And I'm blown away how little of that I'm seeing. Ideation abdicated for generation.