CAIN-42 is in Public Beta. Expect changes, and do not rely on it as your only control for anything critical. What beta means
CAIN-42 CAIN Studio

FALSEWORK · Prove it's real

FALSEWORK Reality Check

Your team or a vendor says the feature is “AI-powered”. Is it? In five business days you get a written report that grades each part of the system REAL, PARTIAL or STUB, with the file, the line and the evidence behind every finding. Fixed price. 100% money-back guarantee. No integration: we need read access to the code and the addresses of the endpoints.

$2,500 fixed price · founding price for the first five customers
50% deposit ($1,250) to start, 50% when the report is delivered. Report in five business days from access.

Pay the $1,250 deposit by card Order by email Ask a question first Read a sample report

Card payment is taken by Stripe on its own page; we never see your card number. After you pay we email you to schedule the scope call.

100% money-back guarantee. A full refund for any reason before delivery, if we are late, or if the report is not useful to you. The terms.

What you get

  • A grade for every component of the code base, and one overall grade.
  • Every finding with its evidence: file, line, the code itself and what it means.
  • A live test of up to 20 endpoints. Each is called twice with different inputs; an endpoint that gives the same answer both times is not computing anything from what it is sent.
  • A person's reading of every finding. Each one is confirmed, set aside with a written reason, or returned to you as a question. Nothing the engine reports is hidden.
  • A machine-readable verdict (verdict.json) with digests, so the report can be checked against the code you supplied.
  • A 30-minute walkthrough with the person who did the audit.

What it finds

  • Random numbers returned as scores, confidences or measurements.
  • Routes that call code which does not exist.
  • Missing dependencies that are silently ignored, so a feature quietly does nothing.
  • Placeholder and unfinished work shipped as finished.
  • Success answers returned without doing the work.
  • Endpoints that return a canned answer whatever they are sent.

The grades

REALNo finding that counts against it. The code computes what it returns.
PARTIALWorks in part. Some of what it presents is incomplete or is not computed.
STUBAt least one result it presents is not computed at all, or a route calls code that does not exist.
UNKNOWNCould not be determined from what was supplied.

How the five days go

  1. Day 0. A 20-minute call: what to audit, which endpoints, who gets the report. You grant read access.
  2. Day 1. The engine scans the code on our own machine. Your code is not uploaded to any third party.
  3. Days 2–3. The auditor reads every finding and runs the live test against your endpoints.
  4. Day 4. Open questions come back to you, so nothing is graded on a guess.
  5. Day 5. Report, verdict.json and the walkthrough.

Who buys it

  • Teams about to sign with an AI vendor who want to know what they are buying.
  • Agencies delivering AI features who want third-party evidence for their client.
  • Founders before a funding round or sale, who want to find the gaps before the other side does.
  • Investors looking at an AI company.

What it is not

100% money-back guarantee

If you are not satisfied, you get your money back. All of it.

Why believe the engine

We ran it on ourselves first. On 2026-10-13 the engine and the tests around it were pointed at our own four premier products and found six real defects, including two checks that had never worked and a security gap. All were fixed and re-tested the same day (the evidence). You can also run the engine yourself with a free trial key before you spend anything, and read a sample report produced by the same tool on a demonstration project.

Pay the $1,250 deposit by card Order by email Ask a question first

FALSEWORK is a service of FIRME AI, built on the CAINWitness engine. Same-project tests only; no third-party audit of the engine yet.