Skip to content
Perceptual

Intelligence thatrewrites itself.

Billions of
AI models.

Models that map their own neurons and rewire themselves to improve.

We’re making intelligence as cheap to shape as software.
It’ll be built by everyone, for everything.

Join waitlist.

Get early access to what we’re building. Leave your email and we’ll let you know when it’s ready.

Our team comes from

  • Apple
  • Berkeley Artificial Intelligence Research
  • Harvard University
  • Y Combinator

Perceptual’s thesis

Why intelligence should be engineered, not just trained

Intelligence lives in the weights. Yet the only way to change them is gradient training: expensive, slow, and imprecise.We’re building the frontier of intelligence engineering: models that can read, repair, and extend their own neurons.

  1. Precise

    Change one behavior or add one skill by editing the neurons responsible, without the model forgetting everything else.

  2. Efficient

    No new dataset and no training run for every change. Edits are written in language, not paid for in data and compute.

  3. Self-improving

    A model that understands its own weights can improve them, with every edit traced and auditable.

What weight editing unlocks

Understand a model. Then change it.

Model debugging

Trace any output back to the neurons and circuits that produced it, so you can see why a model behaves the way it does before you ship it.

Behavior editing

Capability transplant

Safety and alignment