v1.2.1 — what a server-side SQL client actually costs

A new engineering post, Moving the SQL client off the laptop: four problems nobody warns you about.

Running a database client on a server next to the data removes two unwritten requirements — every laptop reaching the production network, and every laptop holding every credential. The post is the invoice for that trade: the trust boundary lands inside your own process, ambient browser authority makes CSRF real (and a reverse proxy can silently break the fix), your own rejection logging becomes a denial-of-service surface, and per-user state stops living in localStorage. Each section names the concrete shape the problem took in LibreDB Studio, including the two we shipped wrong first, and closes with what relocating the client does not buy you.

No code or page changes otherwise. 120 pages, 166 tests, accessibility at 100 on every audited route except the landing page at 97.