STUDY / THE QUESTION
How does an AI company show usefulness instead of science fiction — and instead of promising?
THE DECISIONS
01A console in place of the hero
The visitor types a request and watches the agent's plan unfold: tool, action and verification at each step. It is scripted and deterministic, calling no model at all — what is demonstrated is the cadence of a system that shows its intent before it acts.
02Cyan only on active states
Near-black background, violet for the tools, cyan only on what is happening right now. No background gradient.
03Security as a table
Where the data sits, for how long, who has access — in rows, not in a paragraph. Benchmarks with axes, a scale and the word "illustrative" beside them.
WHAT CAN BE MEASURED
- Network calls in the console
- 0 — three scripts embedded in the page
- Animations
- 2 — the hero particle field and the unfolding of the plan; the rest is static
- Photographs
- 1 — the server room, beside the data table
WHAT WE LEARNED
The only demo in the showroom where the visitor acts. Leaving having "used" the product — even knowing it is a demonstration — is worth more than any list of capabilities.