Every term the meter uses, defined so each one stands alone. The vocabulary comes from real reports, starting with the audit we ran on our own product in 2026. For how they fit together, read the methodology.
A 0 to 100 score of the gap between what a product claims and what the evidence supports, where higher means more bullshit. It is not a viability score: it measures honesty of description, not quality of business. A weighted five-dimension rubric computes it in code.
A separate 0 to 100 score for whether there is enough real signal to keep building, across problem reality, product usefulness, differentiation, distribution, and monetization. It is deliberately never averaged with the Bullshit Index, because the interesting information is usually in the gap between the two.
The free tier of Bullshit Meter. It extracts every claim your pitch makes, shows the full interrogation plan (what would be checked and where), and names the one claim that smells worst. No scores, no verdict, no account required.
The paid audit ($29 at launch). It reads up to 25 real files from your repo, checks every claim against your code and the live market with counter-searches, and returns both scores, a verdict, a receipts table, three Monday actions, and kill criteria.
A re-run of the full audit on the same project after you have fixed things ($19, first one included with every report). It shows score deltas, claim status changes, and what actually moved, so improvement is measured instead of felt.
The list of every claim a pitch makes, stated or implied, that the audit extracts before checking anything. Each entry ends up verified, contradicted, or explicitly marked unverifiable; none are quietly dropped.
The evidence table attached to every finding: the source, the observation, how strong the evidence is, and a link you can click. A score you cannot audit is a vibe with a decimal point; receipts are how the meter avoids being one.
A market search phrased to prove a claim wrong rather than right, run as a mandatory part of external research. Searching only for confirmation is how founders end up sure; counter-searches are how audits end up honest.
A pipeline stage whose only job is to attack the audit's own findings and delete any that are unfair, unsupported, or overreaching, before the verdict is decided. It exists so the report earns its rudeness.
One of seven final calls: KEEP BUILDING, FIX THIS SHIT FIRST, NARROW IT, PIVOT AROUND THE GOOD PART, COOL PROJECT BAD BUSINESS, KILL IT, or NOT ENOUGH EVIDENCE. KILL IT has strict conditions; NOT ENOUGH EVIDENCE is an honest answer, not a failure.
High, Medium, or Low, based on how much real evidence the audit had to work with. It is about the evidence base, not the model's self-reported certainty.
Exactly three things to do Monday morning, each tied to a specific finding. Three because a list of twenty is a way of recommending nothing.
One to three observable events that would mean the thesis failed, chosen so the builder can actually notice them happening. They convert 'maybe it will work' into a testable bet.
The part of a project the evidence suggests is stronger than the part being pitched. When the audit finds one, it says what it noticed, why it may be stronger, and how to test it.