.png){kind=logo}
Give AI the ability to shape how it sees.
Almost every sensor is fixed: one point of view, one modality, built for human eyes. Elio makes the optics programmable and lets AI reshape them in real time, so a single sensor can take on new sensing tasks and return what a model actually needs.
Every leap in AI has come from the same two places: more data, and bigger models. We have taught machines to read, to reason, to write code, to fold proteins. But look closely at the point where all that intelligence actually meets the world, and you find something that has barely changed in decades. The sensor.
A sensor is the eye of every AI system, and almost every one is fixed: a single point of view, a single modality, ground into the hardware in advance, and tuned to produce an image a person can look at. Everything downstream, all that intelligence, is working from a picture it was handed. Not the world itself, but a flattened, human-shaped copy of it, with most of the information already thrown away before the model ever gets a look.
That is the real ceiling. Not the model. The eye.
Elio builds sensing for AI: sensors that adapt, in real time, to what a model needs to know, instead of fixed instruments, designed by humans for human eyes, that capture a single view decided in advance.
What if the eye could think?
Elio started from a simple question: what if the sensor were not fixed? What if, instead of a lens shaped once for one purpose, the optics could be reshaped, continuously, in real time, by the intelligence using them? What if an AI could design its own way of seeing, tuned to each thing it needs to see?
That is what Elio builds, and it is a paradigm shift because of what it removes: the human standing between intelligence and reality, deciding in advance what a machine is allowed to perceive.
Take that person out of the loop, and sensing changes in kind, not degree. It stops being a fixed act of capture, one exposure, one modality, one spec, and becomes something closer to exploration: the sensor looks, learns from what it sees, reshapes itself, and looks again. The sensor stops being a passive component at the edge of an AI system and becomes part of the intelligence itself, the point where the model reaches into physical reality and chooses its own information.
