Anyone can build the search box

Claude, Codex and Grok are here, and a founder can stand up a CRUD app over a weekend. The question that follows is whether SaaS is dead, and it is the wrong question. Building got cheap. Owning the workflow did not.

Montenegro publishes every law it passes as a set of page images. Pulling those back into text is the part that looks hard and is not. The Official Gazette renders its own digital typesetting rather than scanning paper, so at three times the size and through a Croatian OCR model the text comes back with the diacritics intact. The register holds 78,308 acts, 38,696 of which now have full text. That is a pipeline and a subscription, and a competent person could rebuild it.

The expensive part arrives afterwards. The 2026 VAT law raises the registration threshold to 30,000 euros and does not apply until the country enters the EU, so the 18,000 euro threshold from the older law is the number that governs today. A search box that returns the newest document returns the wrong answer, confidently, to a small business deciding whether it has to register. Nothing in the text says so. Somebody has to know, and keep knowing, every time the law moves.

That is the whole distinction. The tool is a form over a database and always was. What anyone pays for is the promise that the answer is current: that repealed articles never reach the assistant, that a consolidated text from 2021 is labelled as a consolidated text from 2021, that an act with a missing page is published as incomplete rather than quietly broken. None of that is code. All of it is the thing that makes the code worth using.

Cloning a form into a dashboard is not a product company. Code is abundant now, and taste, distribution, operations and systems that stay calibrated are not. The thing worth owning is a decision surface somebody will pay to keep correct. Code was never the moat. The 18,000 was.