specifications.global demo
The current assistant behavior is embedded in the specs.global experience and remains tied to product pages, raw text, and manual iteration.
AI product research project
Speconaut
Not a finished assistant yet. Speconaut is the name and structure for the product-question AI growing out of specifications.global, raw text collection, and a small demo that already works in parts.
Position
A named place for the AI before the AI is complete
Speconaut exists so the idea has a stable home, search engines can understand the project, and future product surfaces have a clear identity before the backend becomes a real service.
The honest current state is simple: specifications.global is the working demo and data-collection surface. Speconaut is the future product layer that may later answer product questions in a store, on the web, or through an API.
Current truth
What exists today
Raw text accumulation
The project still needs more product text, sources, model identifiers, and comparison context before it can answer broadly with confidence.
Speconaut identity
This site gives the assistant a product boundary, documentation structure, and future account/API surface without pretending the full service is live.
System map
The path from pages to answers
The first durable advantage is not model size. It is the shape of the product memory: names, variants, specs, review notes, source traces, and follow-up context.
01
Product pages
Structured pages on specifications.global establish canonical product identity.
02
Raw text
Collected notes and source text become the memory Speconaut can retrieve from later.
03
Demo answers
The current demo tests question handling, comparisons, follow-ups, and source behavior.
04
Future API
Accounts, sessions, SMS verification, and partner clients are planned, not connected.
Future shape
An assistant that can stand near the shelf
Storefront conversation
Long term, Speconaut should be able to ask practical follow-up questions, explain tradeoffs, and help a customer choose between real products in front of them.
Distinct voice
It does not need to be the strongest AI. It needs its own product memory, a clear way of explaining evidence, and a consistent way of saying what it does not know.
Service boundary
Future clients can use a public API, account login, phone verification, and session history. Those surfaces are documented now so the architecture has a place to grow.
Access model
Login and API are placeholders by design
The login page shows the intended account modes: Google, Apple, phone, SMS, and email. No credential or SMS provider is connected yet.
Public siteLive
specs.global demoActive
Public APIPlanned
SMS providerNot connected