May 29, 2026
Small file, big feelings
SQLite is all you need for durable workflows
Why pay for a giant system when a tiny file might do the job?
TLDR: The article says many automated apps can safely store their progress in a simple local SQLite file instead of relying on a big database service. Commenters were split between “this is brilliantly simple” and “that’s not real durability if recent work can still disappear,” turning a boring storage debate into a spicy reality check.
The big claim here is deliciously simple: maybe you don’t need a huge always-running setup to keep automated tasks safe and recoverable. Maybe, for lots of projects, a humble SQLite file plus backups is enough. The article argues that instead of building a sprawling machine just to remember what a workflow was doing, you can keep that memory in one small local database file and copy it to cloud storage with tools like Litestream. Cheap, simple, inspectable — very appealing, especially for today’s wave of AI agents and experimental tools.
But the comments? That’s where the fireworks are. One camp instantly turned the whole thing into a meme, with the deadpan zinger, “Files is all you need,” complete with an xkcd link, basically saying: congratulations, we’ve reinvented “just save it to a file.” Another camp was much sharper, calling out what they saw as the article’s biggest loophole: if your latest changes can vanish before the backup catches up, is this really “durable” in the everyday sense of the word? One commenter practically translated the pitch into, “SQLite is all you need… unless you actually care about durability.” Ouch.
Still, supporters were upbeat. Some said they were already doing this for AI agents and loved how easy it made retries and tweaks. Others chimed in with “this already exists” energy, pointing to tools like Temporal and DBOS itself. So the vibe was classic internet tech drama: is this a clever simplification, or just old ideas with a shiny new AI wrapper?
Key Points
- •The article argues that many durable workflow systems can use SQLite instead of a separate orchestration tier or network database.
- •Workflow durability is described as primarily depending on preserving workflow state, while compute can remain cheap and disposable.
- •Obelisk is presented as a system where workflow progress is kept in an execution log, workflows replay from persisted history, and activities can be retried.
- •Litestream is proposed to asynchronously replicate SQLite changes to S3-compatible object storage for backup, migration, and inspection.
- •The article says Postgres is still the better option when higher availability, broader shared scalability, or stronger durability properties are needed.