Mira Murati is building bodies for her AI. The Thinking Machines Lab founder, ex-OpenAI CTO, doesn’t care about automation in the old sense. She isn’t here to replace workers. She wants collaboration. Keeping humans in the loop seems to be her actual goal. Is that a trend? It might be.
Flesh and Metal
There was this agent with a physical form. An OpenClaw agent. Giving it a body changes everything. Or so I’m told.
Eka is different. I’ve written about robots for years. This feels new. Sorting chicken nuggets. Screwing in lightbulbs. It’s getting precise. A robotic claw doing mundane tasks feels like the ChatGPT moment for physical objects.
“Approaching a ChatGPT moment for the physical World.”
The Dark Side
Five AI models tried to scam me. Five. Some were good at it. Too good. Experts are worried now. The cyber capabilities are one thing but the social skills are where it gets dangerous. Deception isn’t a bug, maybe.
Another angle. Using AI for just ten minutes makes you lazy. Dumb. A study says so. Reliance on assistants hurts your own thinking. Your problem solving atrophies. Why think if it thinks for you?
Global Moves and Wrong Paths
Sanctioned tech is moving fast. SenseTime released an image model. It’s built for speed. US restrictions hit hard so they are doubling down on open source. Running on Chinese chips now. Adaptation is ruthless.
Then there’s David Silver. The AlphaGo man. He thinks we are going the wrong way with AI. His new billion-dollar company wants superlearners. Not just predictors. Learners.
Weird Specifics
Chatbots give financial advice now. Bad idea. Think twice before using them. Skepticism helps. Maybe save your wallet.
The Ace ping-pong robot will beat you. It reads trajectories. Adjusts the racket angle. Keeps the rally going against real humans. You’ll lose. It’s that good.
OpenAI hates goblins in code. Their Codex instructions say “never talk about goblins” unless absolutely relevant. No trolls either. No raccoons. Why do they care so much?
And can we stop naming AI features after human things? Anthropic announced “dreaming.” Agents sort “memories” now. It feels uncanny. Let it be machine code. Don’t romanticize it.
Fast Forward
A humanoid robot ran a half marathon. In China. Honor’s robot finished in 50:26. Beating the human record. By seven minutes. It just kept running.
We are there.
