I've spent 3+ years talking to the DuckDB community, and at MotherDuck we talk to DuckDB users every day. One pattern shows up over and over: self-hosting DuckDB has a graduation path, and most of the pain comes from not knowing which rung you're actually on. Two things move you along it: - more readers push you UP - more writers push you RIGHT Bottom-left green corner: one person querying local files on a laptop. The best setup you'll ever have, compute sitting right next to storage. Top-right: running all of it, for tons of users. Congrats, you're now a database company (hi, that's us). Every level quietly adds a box you build AND maintain. Auth. An interface. Storage-class choices. A coordinator for fan-out. Multi-tenancy. Observability. Backups. You never take them on all at once, they just pile up. The full post walks it level by level, with the gotchas at each step (and a few benchmarks you should absolutely not trust, run your own). https://motherduck.com/blog/self-hosting-duckdb-road-to-production/ Take care of your ducks 🦆