FQ Labs

Experiments we run on our own work. Some of them stay experiments. The ones that hold up end up inside the platform.

FQ LABSINSTRUMENT 00 · SPHERE · UNDER TESTSCROLL

[00 / WHY WE BUILD]

We build our own tools first. What survives that becomes product.

The films on this site are rendered from code we wrote. The music under them was composed on a machine in the office. The page you are reading runs on an engine built for it. None of that started as a product · it started because the work needed it.

Six experiments are running right now. Three of them are already inside REVEAL, one is in build, and two are still open questions. Everything on this page is measured against the same bar as the platform: can somebody else check it.

01 · STUDIO

The film system

Every film on this site is drawn in dots and lines and rendered frame by frame from a script · not shot, not edited. Change a number in the source and the whole set re-renders overnight. It is how the films stayed consistent while the product moved under them.

LIVERENDERED FROM CODE

02 · ENGINE

The site engine

Dots, hairlines, a virtual scroll with its own physics, and a canvas that redraws whatever the page needs. Written for this site in TypeScript, with no framework underneath it. The panel beside this paragraph is the same code, running at a smaller size.

LIVETYPESCRIPT · NO FRAMEWORK

DOTS · LINESDRAWING NOW

03 · SOUND

The music engine

Three beds carry this site · surface, machine, finale. All three were written and rendered on our own hardware, then mastered to sit under narration without fighting it. Nothing licensed, nothing borrowed. The panel reads the file it plays, so the shape you see is the track.

RESEARCHCOMPOSED ON OUR HARDWARE

FQ BED · FINALE

04 · MODELS

The local model stack

Speech, reasoning and scoring models running on one machine, with nothing leaving the building. We put our own work through it first · transcripts, drafts, scoring passes, whole days of it. What survives that is what REVEAL Local ships, from 100 seats.

LIVERUNS ON YOUR HARDWARE

05 · DATA

The synthetic corpus

Quality systems are hard to test, because the failures worth catching are rare and real client data cannot leave the client. So we generate the test set instead · multichannel conversations, one firm per industry, every failure mode placed on purpose. It lets us prove a scorecard works before it reads anybody's calls.

RESEARCH100% SYNTHETIC · NO CLIENT DATA

06 · ACTION

Flows and Voice

A finding is worth nothing until it moves. Flows turns one into a task, a call, a message or an escalation, with the trigger and the channel written down. Voice is the same chain answering the phone. Both run against our own operation before they run against anyone else's.

IN BUILDTRIGGERS · CHANNELS · VOICE

[07 / WHAT COMES BACK]

Labs feeds the platform.

Nothing here is a side project. Each experiment started because something in REVEAL needed it, or because we wanted to know whether four people could do it at all.

When one holds up it moves into the platform and stops being an experiment · the film system, the engine and the local stack all made that move. When one doesn't, we write down why and leave it on this page, because a company that only shows its wins is not showing you much.

[08 / TOGETHER]

Let's work
together.

Bring us the part of your operation nobody can see today. We'll show you what a full month of it looks like read end to end, with a chain you can check yourself.