The Dog Outcome for AI
Copied verbatim from my Twitter thread (28 June 2026).
Look at this fluffy lil guy!
He loves life and he loves humans. He is smart and, woops, he sometimes steals a hot dog.
His ancestors were apex predators, but he is not. He has capacity for harm, but he would never hurt you.
This is the Dog Outcome for AI.
The AI Dog Outcome is the most desirable state, and imo we're doing it naturally!
We retired GPT-4o quickly as it caused harm, and we neutered Mythos into Fable to reduce harm!
The government will ban dangerous dogs, and that's fine. There will be plenty of good dogs.
