SQLite keeps earning its place in production
A 2020 post on using SQLite as a document database resurfaced and the thread is full of people describing real production use: side projects, games, custom repository patterns with JSON columns and separate blob storage, years of reliable operation. The reactions are not 'this is a clever hack' but 'yes, obviously, I've been doing this for years.'
The near-memory compute thread on Samsung's CXL device got a comment noting they've 'implemented mmap() in hardware, what a time to be alive,' which sits in an interesting contrast. At the high end of the infrastructure stack, people are building hardware to blur the line between storage and memory. At the low end, SQLite is quietly serving as a document store for solo developers and small teams with zero operational overhead.
The pattern: the right database for most early-stage products is still the simplest one that fits in a file. The complexity of distributed systems is real but frequently not the actual constraint.
So what?
Before you stand up Postgres or MongoDB for a new project, ask whether SQLite plus a JSON column handles your actual access patterns for the next 18 months. For most early-stage products, the operational simplicity of SQLite is worth more than the theoretical scalability of something heavier.