Method

How we check the books of an AI business

Plenty of people post revenue. Very little of it is verifiable. This page sets out the method we use every episode, so you can check us, and use it on any business.

Three questions

Every episode answers the same three: Are the numbers real? How did it grow? Can you do it? Conclusion first, evidence after.

Evidence levels

Verified Payment-account data (e.g. Stripe or RevenueCat read by TrustMRR), company filings, official price pages.

Self-reported What founders or companies say publicly: press releases, blogs, podcasts. Recorded, never treated as fact.

Third-party Media, data platforms (e.g. Similarweb), industry surveys. We name the source, date and basis.

Our estimate Calculated from public prices and stated assumptions — per-user compute costs, one-year scenarios. Assumptions are published so you can redo the math.

Base rates against survivorship bias

Each episode includes a base rate: how far most comparable projects get. If we can't find a reliable one, we say so.

Sourcing rules

  • Every number comes from a page we opened that day, logged with URL, date and the original figure.
  • Numbers we can't trace to an original are dropped, however widely repeated.
  • Figures on different bases are not added or compared without saying so.

Cross-checking

Before a deep report ships, its key numbers and reasoning are checked independently by a second AI model, then re-verified line by line against the original pages.

Listed companies

We analyse the business, never the stock: no buy/sell calls, no price targets, no share-price predictions. Figures come only from filings and exchange documents.

What we don't do

  • No investment advice, no "sure things".
  • No mocking founders. We show the evidence and let you judge.
  • No payment from the companies we analyse for favourable coverage. Any future partnership will be disclosed on the page concerned.

About Jimmy

Former Fortune 500 manager, e-commerce operator and serial founder. Full story →

Corrections

Spotted a wrong number or source? Email jimmy_yu2020@outlook.com. Confirmed errors are fixed on the page with a correction date.