Pilot

Six weeks, on your own questions.

A pilot answers one question with evidence: does Day Theory make Claude more accurate on your data? You get the measured answer, proven or not, before you deploy anything.

What happensSix weeks, start to report

Four stages, in this order.

  1. 01Baseline

    Measure Claude alone

    We measure Claude alone on questions your team has already answered, so the gain has something real to beat.

  2. 02Knowledge

    Draft and check

    Day Theory learns from half of your solved cases and checks every rule against your data.

  3. 03Held-out proof

    Prove it

    Claude answers the other half with and without the knowledge. A paired test decides.

  4. 04Report

    Get the decision

    Counts, rates, the test, the cost and the gate’s decision. If the gate refuses the knowledge, the report says so.

Your side, our side

What you bring, and what you keep.

You bring

To get started

One use caseQuestions your team has answered, with the answersA read-only role on your warehouseYour own Claude access

You keep

Whether or not it ships

A baseline for Claude aloneThe knowledge, as your fileA held-out test of the gainThe report and the gate’s decision
Who does what

Your team runs it. We sit beside you.

Who runs it
Your operator runs Day Theory on your machine. We coach over a screen share, so we never need your credentials.
What decides it
The release gate, set before the test: more right answers, under a 1 in 20 chance of luck, and no group of questions down more than 10 points.
What you see
The same kind of proof as on our Results page, on your own questions.

See how many wrong answers it removes on your own questions.

Bring questions your team has already answered. You’ll see the measured gain on your data, with the same proof shown here, before you deploy anything.