LUMEN exists to surface what’s worth reading before it’s everywhere. Two things crossed my desk this year that, read together, say more than either does alone.

The first is about the body.

World Labs, founded by Fei-Fei Li, is building what she calls spatial intelligence — AI that understands three-dimensional space the way a physical object does, not the way a sentence describes one. Her argument is blunt: language models compress text but have no native model of geometry, persistence, or gravity. For anything that has to act in the physical world — a robot, a vehicle, a surgical tool — that’s disqualifying. Spatial intelligence, she says, is “as critical as, and complementary to, language intelligence.” The company’s own research now frames this explicitly as robotics infrastructure: simulated environments real enough to train a robot’s behavior before it ever touches the real world.

The second is about the loop.

Thinking Machines Lab, founded by Mira Murati, works on something adjacent but different — not embodiment, but interaction. Their stated position: “Instead of focusing solely on making fully autonomous AI systems, we are excited to build multimodal systems that work with people collaboratively.” That’s not a philosophy borrowed from OB1x. It’s a frontier lab arriving at the same conclusion independently, from a different direction entirely — real-time, multimodal, built to stay in conversation with a person rather than run ahead of one.

Read together, that’s the whole ORBIT thesis, stated twice, by people with no reason to agree with each other. One is teaching machines to understand a body. The other is teaching machines to stay in the loop with a mind. Neither company is affiliated with OB1x, and neither should be read as endorsing anything here — these are signals worth watching, not partners to name-drop. But when two well-resourced, independently-arrived-at research programs both land on ground it, and keep the human in the loop, that’s not a trend. That’s a discipline forming.

Augmentation, not automation, isn’t just a phrase. It’s a bet other people are placing too.