For people who actually build things.

An investor, a customer, or a competitor is going to poke a hole in your story.Poke it first.

The private dry run you run on yourself before any of them do. Give it a public GitHub repo or paste your pitch. It reads the actual artifact, checks every claim against your code and the live market, and hands you the receipts while the stakes are still zero.

Run the free Sniff Test See a sample report Pricing

The sample is a real audit of a real idea. That is the point.

Self-test recordUnit under test: bullshitmeter.devResult: printed as measured

We ran it on ourselves. Here is the card.

41/100

Bullshit Index

0255075100

72/100

Build Signal

Verdict

Fix this shit first

"The product is real. The category story is the bullshit."

That is our own product, scored by our own tool, printed exactly as it came out. If it will say that about the thing we are trying to sell you, you already know what it will say about the thing you have not shipped yet.

Test 01

Two scores, on purpose. We never average them.

The Bullshit Index (0 to 100) measures the gap between what you claim and what the evidence supports. The Build Signal (0 to 100) is a separate score for whether there is enough real signal to keep going. It is not a probability of success, and it does not move with the Index. We never average the two into one comfortable number, because your code can be strong while your story is bullshit, and your story can be great while there is nothing under it. When the two disagree, that gap is the most useful thing in the report.

0100

Fig. T-01: Index 41 against Signal 72, printed from our own card above.

Test 02

No, this is not ChatGPT told to be harsh.

Idea validators grade your pitch. Code tools stop at the code. Roast tools are entertainment. This one opens the files: up to 25 real ones from your repo. It keeps a ledger of every claim you make out loud or by implication, then checks each against your code and the live market with mandatory counter-searches built to prove you wrong. An adversarial critic hunts down any finding that is unfair or unsupported and kills it before the verdict lands. The score is a weighted rubric computed in code, powered by Claude, not a number the model felt like giving you. A score you cannot audit is a vibe with a decimal point, so every finding ships with a receipt: the source, the observation, how strong it is, and a link you can click.

Test 03

Everyone around you is too kind to be useful.

You are too close to the product to see it straight. Your friends round up because they like you. Your AI assistant, the one that helped you build the whole thing, agrees with every word you feed it. Enthusiasm is not evidence. So you keep shipping the feature everyone politely nodded at, until someone finally admits it was the least interesting part of your own product. By then it is already built.

Test 04

What actually lands in the report.

A Bullshit Index and a Build Signal, both 0 to 100. A Confidence rating (High, Medium, Low) based on how much real evidence it had to work with, not model confidence theater. One of seven verdicts, from KEEP BUILDING to KILL IT. A receipts table, exactly three things to do Monday morning, one to three ways to know you were wrong, and an honest list of what it could not verify. It takes a few minutes, not seconds, because gathering real external evidence and running the checks is actual work. Anything promising you an instant verdict is skipping the part that matters.

Poke it first.

Check my bullshit See a sample report

The Sniff Test is free, no account needed. The full report is $29 while we are in a good mood. Private by default: your report lives at an unguessable link, nothing is published unless you share it, and secrets are redacted before any model call. Evidence-backed. Occasionally rude to your positioning. Never to you.