Some useful tools only need to exist for the duration of a job. Imagine an equipment specialist assembling a one-off repair plan: parts already inspected, conflicting supplier information, and questions still waiting for an answer. A chat transcript can describe that work. A small application could retain its changing state so the specialist can resume, correct it, and hand it over.
Supabase’s announced acquisition of Turso points toward infrastructure for that smaller unit of software. Supabase says its Postgres offering and Turso’s SQLite work will continue, while it develops a more unified experience for agents creating applications. The announcement describes a direction; it does not deliver an automatic SQLite-to-Postgres migration system.
The underlying difference is that a database need not bring a dedicated, continuously running server with it. Turso’s account of database-per-tenant architecture explains the implications of SQLite databases being files, with work loaded on demand. Idle workspaces can retain their data without each reserving the resources of a separate database process. That changes one part of the cost of keeping many small tools around.
For the repair example, persistent state could mean a parts checklist, sources attached to disputed facts, and explicit unresolved items. The proposed benefit is continuity across sessions, not a smarter answer in one conversation. An agent could help create or update that workspace while the specialist determines which evidence to accept. This is a hypothetical product opportunity, not a demonstrated customer result.
Cheap storage does not make such a product cheap to operate in every respect. Authentication, recovery, and support remain work. Turso also identifies coordination across schema changes and reporting across separate databases as tradeoffs. A tool that grows into a shared business system may need different architecture; the acquisition announcement is not a promise that transition will be effortless.
The interesting test is whether lower provisioning overhead makes a neglected expert task worth serving. A temporary tool can have lasting value if its records remain usable after the job ends. Building another permanent platform is not always the best starting point; sometimes the product should be as small as the piece of work it helps someone finish.
