A map of AI where every fact carries the date it was last verified.
iDiscoverAI connects models, tools, agents, companies, and concepts in one graph. Every entry shows when a person last verified it, and anything that goes stale drops out of the index on its own.
- Product
- iDiscoverAI
- Type
- Our own product
- Platform
- Next.js on Vercel
- Status
- Live
How the product is built, in numbers
What we set out to fix
Information about AI goes out of date faster than any reference can keep up with. Context windows change, prices halve, and models are retired, while directories keep serving the old numbers with no way to tell.
- Models, products, and ideas are blurred into one list, so the same thing appears three times
- Nothing shows when a fact was last checked, so stale specs look as trustworthy as new ones
- Scraped listings look complete but carry no original information
- Paid placement and affiliate ordering make every ranking suspect
What we built and why
We treated trust as a data-modelling problem. Freshness, sourcing, and independence are enforced by the system, so they cannot erode when an editor is busy.
One canonical entity, one URL
Ten entity types with no cross-filing: a model is a model, the product wrapped around it is a tool, the idea underneath it is a concept. Slugs are unique across every type, and comparisons, collections, and learning tracks reference entities instead of restating them, so a correction lands everywhere at once.
Verification stored as data, not a badge
Each entry records when a person last verified it, separately from when it was last edited, so fixing a typo never resets the clock. Trust windows differ by type, from 90 days for models to three years for papers. At 70% of the window an entry is marked as ageing; past 100% the reader sees a warning and the page leaves the index automatically.
A quality gate before anything is indexed
Every entry is scored on eight signals: a one-sentence answer, depth of original writing, filled key facts, graph relations, at least two primary sources, real reader questions, stated limitations, and original insight. Below 60 out of 100 it is not indexed. A scraped row with a name, a logo, and a borrowed description scores about 14.
Independence enforced by the product
There is no paid inclusion, no sponsored entry, no affiliate ordering, and no star rating anywhere. Ordering is labelled as editorial opinion, and every entry must state its limitations alongside its strengths.
Static, fast, and locked down
Pages are prerendered with the Next.js App Router and served from Vercel's edge cache. A strict Content Security Policy keeps resources to the site's own origin and blocks framing, HSTS is enforced, and a command palette puts the whole graph one keystroke away.
Everything on this page is live. Open iDiscoverAI and check it for yourself.
The practices this product drew on.
Let us look at your architecture.
Two ways to start, both of them short. Bring the problem, not a specification — the first useful thing we do is tell you what we would build and roughly what it costs.
Book an architecture call
Fifteen minutes, no deck. We map your problem to an approach and tell you what a realistic scope and budget look like.
- A specific technical recommendation
- A budget band you can plan against
- An honest answer if we are the wrong fit
Send a written brief
Prefer to write it down? Send us the shape of the problem and we will reply with a first take, usually under 12 hours.
- Goes straight to an engineer, not a sales inbox
- We reply with an approach, not a brochure
- Links to repos, docs, and screenshots welcome
or email hello@folrik.com