How to think about it
Chess has a right move. Poker — play it perfectly and still lose a hand. But over a season, the pro cleans up. Judge it over the long run.
And the edge is jagged
It’ll draft a flawless job spec, then miscount the words in it. Genius and nonsense sit right next to each other — and you can’t always tell which you’re getting.
What it really is
Fluent translation between you and the machine — however you want to talk to it.
The old split just broke
Slow, sloppy — but brilliant.
Fast, exact — but stupid.
Fast & fluent — sometimes brilliant, sometimes nonsense.
Best of both. Worst of both. Your job is knowing which one you’re getting — and staying in the driver’s seat.
The one shift that changes everything
Augmentation > Automation. Don’t hand over the thinking — use it to think better.
…but not always · diagnostic accuracy
So point it at the right thing
Aim it at the drudgery — the admin, the formatting, the first draft — so you get the time back for the judgment, the relationships, the work only you can do.
Which tasks do you hand over?
Keep it. That’s your work.
Hand it to AI.
Don’t end up folding laundry while it paints.
The 80 / 20
The landscape
Nearly every new tool is built on one of these. The tools churn — the engines stay.
Hare, or tortoise?
Everywhere else, be the tortoise — what a new tool does today, the big models embed in ~6 months.
The other edge