Introducing GEN-1.5, a one-shot learner. It can learn new tasks in a few seconds.
Before we start
The Breakfast Machineby Simone Giertz, on YouTube
Hand it the keys to everything I own — and trust it never to break a thing.
Introducing GEN-1.5, a one-shot learner. It can learn new tasks in a few seconds.
Introducing S1, our new foundation model that learns from one example.
GPT-6 Astra scored 95% on a robot control task, up from Fable 5.1’s 40%.

i gave astra a robot, a paint brush, and a camera then asked it to paint the golden gate bridge in real life!
Capability is sprinting. Trust is still tying its shoes.
So here’s my plan.
A robot that fails safely, gets better from its failures, and tells a restaurant owner what happened, why, and what to do — even at 4:55 AM.
Every day it ends up like this. Every day I clean it up.

Cheap and imperfect — so it fails often enough to learn from.
Right object, wrong angle. The gripper comes in rotated, and the grasp fails.
The wrist rolls. Its camera sees wall and ceiling — the table is gone.
Every one of these is a signal. Right now, they just get thrown away.
SmolVLA drives the arm from its wrist camera.
“Pick up the can.”A cheap check on every frame.
Gripper closed. Nothing in it.Pause before anything gets broken.
Arm paused. Holding still.A vision-language model puts it in words.
“I missed the can. It’s further left than I reached.”Help, skip, or let it carry on alone.
RetryNudge itSkipMost research builds one of these boxes. My work is the wiring — one loop, with a person in it, on a cheap arm.
Where I actually am
The plan is on paper. The arm is on my desk. I’ll build the rest in public — failures included.
Thank you.