Four jobs are genuinely solved, two are getting there, and one is oversold. Knowing which is which saves a quarter.
These have a property in common: the output is verifiable in seconds, and being wrong costs you a retry rather than an incident.
Multi-step analytical reasoning — questions that need three queries and a judgement between them — works when each step is checkable and degrades when the chain gets long. Schema design suggestions are useful as a second opinion and poor as an authority, because the right normalisation depends on a workload the model cannot see.
Autonomous write operations. The pitch is an agent that applies the migration, adds the index and fixes the data. The problem is not capability, it is that a database write has no undo and the failure mode is unbounded — a bad SELECT wastes a minute, a bad UPDATE is an incident with a recovery time. Until verification is cheaper than the damage, this stays a bad trade, which is why SQLore has no write path at all rather than a write path with a confirmation dialog.
Database AI works where output is cheap to check and errors are cheap to absorb. That is not a limitation of the current generation of models — it is the shape of the problem, and it is why "show the query with every answer" is a design decision that outranks any amount of accuracy tuning.
More answers on the main FAQ, or ask us directly.
The category, explained
Evaluation
For analysis
Every page in this section covers a different question about the same product. See all solutions.
Start free, upgrade when you are asking enough questions for it to matter.
