An AI just rewrote its own brain so it would never use the letter E again.
On purpose. And it worked.
That's the demo Mira Murati's (ex-OpenAI, ex-Tesla) Thinking Machines Lab used to launch its first model this week, called "Inkling."
They had the model rebuild itself. Forget the car that drives itself. This is the car that redesigns its own engine while it's still running.
The task (never use the letter E again) was absurd on purpose. The point was proving that the model can reshape itself into whatever you need.
Why does that matter?
Every other big AI lab is selling higher intelligence. A smarter model, a higher score, a new record every few months.
Thinking Machines Lab is selling ownership.
In their own words: Inkling is "not the strongest model out there." That's not modesty, that's the strategy. Their bet is that companies care less about having the smartest brain and more about having one that's truly theirs. Their model, running on their own machines, trained on their own world. Not a login to someone else's.
For some teams I work with, the question has quietly changed. It used to be "which AI is the smartest?" These days they'll happily mix a few of them together and move on.
The new question is "which parts of this do we actually need to own?"
▎ Selling intelligence → a smarter brain you rent
▎ Selling ownership → a model you shape and keep