01 · the premise
Understanding is the new interface.
LLMs are trained to finish the task, not to work out what you want. xevall is a world model of human intent, predicting what you want a machine to do next, for better human-AI interaction.
- processing · on this device
- media transmitted · 0
02 · natural inputs
Interfaces are losing their keyboards.
What replaces typing is you: speech, tone, gaze, gesture. Intent comes from those signals and from connected context that learns over time: your calendar, your history, your preferences. A machine with both knows what you mean as you say it, and sometimes before.
- gaze
- blink
- speech
- tone
- gesture
03 · system misfires
Nobody measures the understanding.
When an AI system misunderstands a person, the product feels broken and the reason stays hidden. xevall grades the system against the meaning a human reviewer established, and hands back the moment it went wrong.
| session | human meaning | system action | grade |
|---|---|---|---|
| kiosk 04:12 | a pause, not a yes | took it as a yes | misfire · consent |
| agent 00:31 | still mid-thought | jumped in | misfire · timing |
| telehealth 11:47 | said fine, meant not fine | noticed, and asked | understood · 0.92 |
- intended
- inferred
05 · the improvement loop
Every interaction creates traces to improve your product.
You already eval the words. Nobody is evaluating the coherence between the modalities: the tone, the distance, the body language. xevall grades whole interactions against human judgement, marks where the understanding broke, and re-scores the same sessions once you have fixed it. The grading exists to make products more human, not to mint a number.
- agreement 0..1
the Human Input Benchmark
The Human Input Benchmark goes public this autumn.
Building a product around voice, timing, gaze or gesture? Early access is open for teams testing natural interfaces.
xevall ·a world model of human intent, predicting what you want a machine to do next, for better human‑AI interaction.