Move Fast and Don't Break Things: Scaling Databases for the AI Era — PlanetScale
- Source
- youtube.com
- Author
- AI Engineer
- Date
Context
Ben Dicken of PlanetScale argues in an AI Engineer talk that databases can keep up with the sudden traffic AI agents create without teams giving up speed. PlanetScale sells managed databases, so the talk draws on its own tools. He starts with general reliability ideas, isolation and redundancy, then covers sharding, which splits data across servers, along with back pressure and decoupling.
The -specific part, as the description lays it out, is that an agent can design a sharding plan for Vitess from a simple JSON file, that a feature called Traffic Control sets resource budgets so a surge of agent traffic slows down gracefully instead of crashing the database, and that branching, deploy requests and one-click reverts let agents change database structure with an undo path. His conclusion is that good developer experience turns out to be good building blocks for AI. The summary offers no measurements, so these are described capabilities and his argument, not tested results.
- PlanetScale's Ben Dicken grounds reliability in isolation and redundancy, and says failures become weekly events at scale.
- For scaling he covers sharding, back pressure and decoupling, the same approach he says PlanetScale uses for fast-growing AI customers such as Cursor.
- An agent can design a Vitess sharding plan by writing a VSchema JSON file, and Traffic Control resource budgets let an agent traffic surge degrade gracefully instead of crashing the database.
- Database branching, deploy requests and one-click reverts without data loss let agents change schemas safely. His conclusion: good developer experience doubles as good primitives for AI.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Checking sign-in…
Loading comments…






