Check the sentence against its citation before borrowing the argument
A linked source can support the revenue figure while using a different customer universe. Verify each claim separately, including the corrections you intend to make.
The correction I expected to make needed correcting. An essay’s $29.3 billion figure for Cursor’s Series D is supported by the article it links to. The same article also discusses a later proposed $50 billion valuation. Those are different events, not necessarily contradictory numbers.
That small check is the argument: a citation needs to support the sentence attached to it, and criticism of that citation needs the same discipline.
The Product Folks’ essay on AI interfaces links several growth claims to a TNW report on Cursor. The report contains the cited annualised-revenue progression and the historical Series D figure. Its customer statistic concerns nearly 70% of the Fortune 1,000, while the essay says more than half the Fortune 500. The former does not establish the latter.
A shared link can hide several different checks
A paragraph with four claims and one citation looks sourced. It may still contain four separate verification jobs: the number, the date, the population and the status of the event. A proposed financing is different from a completed round. Monthly users are different from weekly users. A customer universe can change while the company name remains the same.
I would split the paragraph into those claims before carrying it into a business case. Mark what the source directly states, what can be calculated from it and what needs another source. If the support is missing, narrow the sentence or omit the number.
That is more useful than giving the whole paragraph a green citation check. It also avoids the opposite error of declaring the entire passage wrong because one figure does not follow. Verification should preserve the accurate parts as carefully as it challenges the unsupported ones.
Separate an interface argument from its growth decoration
The essay’s underlying design question deserves consideration: how much work does a blank chat box push onto a user who does not know what the product can do? That question can be investigated through the interface and a task test. A company’s valuation cannot answer it.
For a hypothetical document tool, I would compare an open chat entry with a specific action such as “extract the renewal terms.” Give users the same document and objective, then observe whether they understand the available action and obtain a usable result. Leave room for open conversation when the task genuinely requires exploration.
That proposed test would not prove one interface is best for every user. It would provide evidence for the specific product decision. Borrowing a fast-growing company’s financial numbers to validate the same choice skips the causal question entirely.
I have not run that interface comparison, and I have not independently audited the business figures in the reporting. Here I checked whether the essay’s claims match what its cited article says. Source consistency is one layer of verification, not a substitute for primary financial evidence.
The practical habit is to make the smallest defensible correction. Preserve the historically accurate valuation. Flag the unsupported change in customer universe. Keep the design hypothesis available for testing on its own merits.
An attractive argument does not deserve weak evidence in its support. It deserves evidence that still says the same thing after someone opens the link.
Verify each number, date and denominator against its source—and make your correction pass the same test.
