What it finds
Everything on this site rests on two demonstrations, and neither of them is a client. The first is my own company, because I built one for myself before selling one to anyone. The second is a listed company read entirely from the outside, using nothing it had not already published itself.
What did it find in my own company in seventy-two hours?
I built the first of these for myself. Inside three days, before it was doing anything clever, the structure surfaced three things that months of ordinary work had not.
- A wrong telephone number, published. Not in a drawer: live in Google and in the answers AI assistants were giving about the company, structured data included. It had been wrong long enough to be the version the internet believed.
- Sixteen inquiries the pipeline never saw. A contact form had quietly rejected sixteen live inquiries in one week while accepting two. Nobody noticed, because a form that rejects silently produces no evidence of itself.
- A gap of $2,800 between the bank statements and the journal. Found by reconciliation rather than by an accountant, which is the difference between a rule that runs every day and a person who runs once a quarter.
None of these was hidden. Each lived in a different place from the fact that contradicted it, and nothing in an ordinary week gave anyone a reason to hold the two side by side.
What did it find in a listed company, from public documents alone?
In September 2026 I put the same method on a public company: a listed group with billions in revenue, dozens of subsidiaries and a market that watches it. Eight research agents in two waves, working only from what the company had already published. No access, no interviews, no non-disclosure agreement and no engagement of any kind. It has never been a client, and nothing below came from anywhere except its own documents and its own live pages.
Places where a fact disagreed with another statement of the same fact, was absent where it was needed, or was stated where it could not be supported.
Rated critical or high: touching an audited estimate, a regulated disclosure, a material exposure, or a quantified recurring cost.
Each finding carries an identifier and an evidence label. Confirmed means observed directly in a primary document or on a live page and recorded with its location. Nothing was asserted that could not be pointed at.
The same fact, two values
The cheapest kind of gap to find and the most expensive to leave, because each one is a place where nobody owns the number.
- A credit-loss allowance identical to the dollar in two consecutive years, in a year when the receivable book behind it moved sharply.
- Two notes in one audited document giving different values for the same prior-year figure.
- Six live answers, across the company own pages, to the question of how many countries it operates in.
- A published brand count that disagreed with the published list of brands directly beneath it.
A number that degraded in one hop
Facts do not usually get corrupted inside a company. They get corrupted on the way out, and come back as the version everyone quotes.
- Audited net profit republished in the press as a materially higher figure, because the number that travelled was profit before tax.
- A headline revenue figure still on the corporate profile seven months after a larger audited result had been published.
- A shareholding stated on one language version of the investor page that is arithmetically impossible against the other.
Money, said plainly
None of these required access to a bank statement. All of them were derived from the company own filings.
- Financial expenses consuming roughly a third of operating profit, with a quarter of that being fees rather than interest.
- Receivables pledged as security rising ninety-one percent in a year, far faster than the balance sheet behind them.
- A large improvement in operating cash that decomposes, line by line, into supplier credit rather than earnings.
- Realised write-offs rising twenty-three fold in the same year that the assumed default rate on the riskiest bucket was cut.
The systems nobody owns
A record cannot be truer than the software that writes into it, and software leaves its age in public.
- The platform carrying the majority of revenue generated in 2016, still serving, with no responsive behaviour possible by construction.
- Roughly forty percent of orders arriving by telephone and electronic mail, stated by the company itself.
- Eighty-five open vacancies across twenty-eight countries containing no engineering role of any kind.
- No member of the board holding a technology brief, in a business whose principal asset is a trading platform.
- Artificial intelligence appearing once in a full annual report, describing a consumer appliance.
The public face
Everything in this group is visible to anyone who looks, which is precisely why it is worth finding first.
- The link to the current investor presentation resolving to an error, on the regulated disclosure channel of a listed issuer.
- The founder forename spelled two ways across the corporate site, the investor site and customer-facing pages.
- Twelve of thirty-five listed markets with no site behind them.
- The single largest social asset, six hundred thousand followers, silent for over three years.
What is not said
Absences are harder to see than contradictions and usually cost more.
- A contingencies note declaring no commitments, no litigation and no regulatory matters of any kind.
- Insurance cover described by reference to a policy that had expired four years before the report was published.
- Guidance missed twice and then discontinued, without an announcement that it had been discontinued.
- Own brands built over a decade with no disclosed revenue or margin, in a document that quantifies every distribution line to two decimal places.
Why does it matter that these were public documents?
Because every one of them was already visible. An analyst, a journalist, a competitor, an acquirer or a counterparty running ordinary diligence would meet the same contradictions, and would draw a conclusion about a company whose own account of itself does not agree with itself. None of it required privileged access.
Which leads to the only question that matters, and it is not a question about that company. If this much is visible from outside, from published documents alone, what is the state of the record inside?
Why do gaps like these exist at all?
Not through carelessness. Each exists for the same structural reason: the same fact lives in more than one place, each place is maintained by different people on different cycles, and no single version is the one the others are derived from. A figure is approved in a board pack, restated in a filing, summarised on a website and quoted to the press, and by the fourth hop nobody can say which was the original.
That is what a second brain is for. One record, one owner for each figure, every fact naming the document it came from, and agents that read what arrives and say so when two things stop agreeing. Contradictions do not become impossible. They become visible the day they appear, instead of the day somebody outside finds them.
What would it find in mine?
I do not know, and neither does anyone who tells you otherwise before looking. What I can say is that in both demonstrations above the first pass found things the people closest to the company did not know were there, and that in neither case was anyone being careless. Ten questions, about four minutes, and your scope, your timeline and your price come back within twenty-four hours.