Research, engineering and design thinking from the team building sharm.
Today's large models predict tokens. To act in the world, a model has to remember it. Here's why we built memory into the architecture, not around it.
Reading a scene as 'positive' or 'negative' throws away most of the signal. We model threat, valence and arousal as separate axes — and they behave like the brain's.
The most important design decision in sharm is what never leaves your device. A look at the on-device redaction moat.
