'Natasha', the AI that built your app, was reportedly 700 engineers: the overclaim problem outlives the fraud

The tool known as Natasha, sold as an AI capable of building applications, was reportedly nothing of the sort: the work was done by hand by around 700 human engineers based in India and Ukraine over roughly eight years. We have only a short report to go on, so we will stay with those facts rather than speculate about who knew what.

The mechanics deserve a name, because the pattern recurs: pseudo-automation, where humans deliver the output while the interface implies a machine. It works for one reason — the customer buys the result, never the process. If the app ships, nobody audits whether a model or a night shift in another timezone produced it. That blind spot is the whole business model, and it does not require a single line of AI to exploit.

Our reading: the fraud is the smaller story. Eight years is long enough that the deception began when the capability genuinely did not exist, and it now ends in a market where a decent coding model really can do a meaningful share of that work. The incentive to fake automation outright is collapsing. The incentive to overclaim how much of it is autonomous is not — it is growing, because the claim is now plausible. Which is why the useful due-diligence questions are economic rather than technical: what is the marginal cost of the tenth app versus the first, what is the latency distribution, what fraction of deliveries touch a human. Genuine automation shows up as margins that improve with volume. Anything else is payroll wearing a product's clothes. The abundance we expect from this technology is real, but it only counts when the machine is actually doing the work — and telling those apart is the job.