A product page, learned first
Name a kind of product. Neutrino learns it into a knowledge store, compiles a programme bound to what it learned, and that programme runs the page and its order flow. No AI is called after the learning.
This site shows a recorded intake: a real learning run (two AI models through the Neutrino gateway, over one public source) whose result is replayed here. A live intake needs the gateway, which is not open on this site yet. The order flow is a demo: no payment, nothing is shipped, nothing is kept.
1 · Learn
Recorded on this site: skateboard.
2 · Bind and compile
Each knowledge slot of the shop programme is filled from the store’s rows, then the translator compiles it in your browser. Nothing else goes into the programme.
Checking the compiler against its pin…
…
Under the hood
the programme’s source, bound and compiled here
3 · The page, run by the programme
Behind the page
How it is checked
Every change to this page is tested in CI: the generated interface, no control without a question from the programme, the component boundary, the page built only from the store’s rows, no network call while the programme runs, and the built page mounted in a real browser.
Not yet proven: the claim that the programme, not React, owns the flow (two intents, one unchanged component package) and the live/recorded tags; both wait on the project definition’s list bound.
Stand-in until C’s browser knowledge seam: the binder copies the store’s rows into the definition as literals before compiling; lists travel as one value each until the list bound lands.