About Ilayer

Built inside your systems, and still running

Most of what we build runs inside somebody else's systems, and we stay on call for it after launch. 200+ agents shipped, counting the tooling we build for ourselves.

A violet glass beam balance seen head on, its arm engraved with a fine ruler scale and a bright needle at the centre pivot. One pan is empty; the other holds a single calibration weight, and the beam is level.

A reading that wanders, and then stops on the line. What gets published here is where it stopped.

Why there are no names here

Most of our work runs inside other companies' systems. The agreements do not let us name them. So we publish the measurements instead: every number on this site, its scope, and what it does not cover.

Where your data goes

The bar is simple. A system that hands you a confident wrong answer is worse than one that admits it does not know. We build the second kind, and we publish what it scored.

Most of the work is the part other teams skip. The tests. The honest no. The trail from an answer back to the document behind it. That is the difference between a demo and something your business can run on.

The engineering does not change with the industry. What changes is whose records the answers come from, and whose systems they have to run inside. We have done that in finance, in security research and in operations.

The record

Three numbers, three different scopes

Two from one production system we still operate; the third from our own research pipeline, and it says so where it is printed.

1.000Correct every time it held backIt scored 1.000 for holding back correctly: when the answer was not there, it said so.One live system, measured at every release. It measures whether holding back was right, not whether it held back often enough.

29M+Holds at real data volumeMore than 1.3 million documents, broken into more than 29 million searchable pieces, every one of them citable.Indexed on the deployment above. Scale shows the method holds; it says nothing about answer quality.

7-gateFindings filtered before a human sees oneSeven checks every finding clears before a person sees it, on our own audit system.The security research system we built for ourselves, not a client's. The habit is the one we reuse everywhere: attack your own findings before anyone else gets to.

Two dark mirrors angled toward each other on a machined base. One violet beam enters from the lower left, crosses between them several times as a set of straight lines, and leaves to the right. — We point the same checks at our own claims that we point at a client's system. What survives both is what gets published.

Each one is a single measured record with its limit attached, not an average across clients. It is the same standard we build the systems to.

One of these numbers went down. We used to publish a more precise figure, until our own claims check re-read the source and found about 72,000 fewer searchable pieces. Every page carrying the old number was corrected. Nobody outside would have caught it.

The four ways to buy it

Start a project

See this method on your own workflow

Name the workflow, and you get a straight read on whether it earns the method, and what the first fixed-scope piece would cost.

Start a project