Pick the screen for what I am typing
Pick, as a person types, the screen that already does what their words ask from the screens they are offered, the interfaces made for them and the screens they kept, with no model, and keep the shown screen calm
Pick, as a person types, the screen that already does what their words ask from the screens they are offered, the interfaces made for them and the screens they kept, with no model, and keep the shown screen calm
Give words (what is typed so far) and the catalog: rows (offered screens, each { id, label, description?, asked?, source? }), made (interfaces, each { interface_id, title, intent }) and asks (kept screens, each { id, words, screen, input? }). memory is the last call's memory (omit it the first time). Answers picks (ranked, each with its entry and score) and memory: shown is the screen to show, which a challenger replaces only after winning twice in a row or with a certain score. Weights, thresholds and bounds are src/instant-screens.json; it names no screen.
Nothing: your AI starts it with your words.
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to pick, as a person types, the screen that already does what their words ask from the screens they are offered, the interfaces made for them and the screens they kept, with no model, and keep the shown screen calm”