What Happens When Billions of AI Agents Hit Your Database? (Andy Pavlo)

<p>AI agents are becoming the biggest users of databases: creating them, querying them, and sometimes deleting them. Andy Pavlo, Carnegie Mellon's database professor and now VP of Database Research at ClickHouse, explains what changes when billions of agents hit the data layer, and why most of the AI world's mental model of databases stopped at vector databases and RAG.</p><p><br /></p><p>Andy's CMU courses taught a generation of engineers (and, it turns out, the AI models); he writes the year-in-review on databases the whole industry reads; and this summer he joined ClickHouse to found ClickHouse Labs. We cover why vector databases were "just an index," what it means for an agent to create a database, why agents could mean 10 to 100x more queries, whether files or databases should hold an agent's memory, how text-to-SQL went from 60% to 99.5% accuracy, why 60% of open-source databases now have commits from coding agents, and what ten years of self-driving database research taught him. Plus ClickHouse, Postgres, the vector/graph/GPU database verdicts, Larry Ellison, and the Wu-Tang Clan.</p><p><br /></p><p>Disclosure: FirstMark, where Matt is a General Partner, is an investor in ClickHouse.</p><p><br /></p><p>(00:00) Cold open &amp; Intro</p><p>(01:21) From vector databases and RAG to the age of agents</p><p>(04:09) Neon's stat: agents create 80% of databases?</p><p>(07:25) Why agents keep deleting production databases</p><p>(08:19) Guardrails: the toddler-and-stairs rule</p><p>(10:51) The four eras of database volume: 10–100x more queries</p><p>(15:10) Agent memory: files vs. databases ("everything is a database")</p><p>(18:24) Which database do AI models recommend? The new SEO</p><p>(19:25) "It cites me back to myself"</p><p>(23:14) MCP for databases, and text-to-SQL from 60% to 99.5%</p><p>(27:09) Trust an agent the way you'd trust a junior developer</p><p>(27:47) Can AI build an entire database? Opus 4 and the CMU projects</p><p>(29:57) 60% of open-source databases now have AI commits</p><p>(30:43) Ten years of self-driving databases, Peloton to today</p><p>(33:52) What LLMs changed: 85% of the tuning in 15 minutes</p><p>(35:42) Should anyone still study databases?</p><p>(38:27) Free CMU courses, the DJ, and the Wu-Tang final exam</p><p>(42:09) Why start a research lab inside ClickHouse?</p><p>(45:53) Why ClickHouse looked like vaporware in 2016</p><p>(47:46) What makes ClickHouse fast: columns, vectors, Snowflake's lineage</p><p>(52:22) Why Databricks, Snowflake and ClickHouse all added Postgres</p><p>(58:15) Vector, graph and GPU databases: thumbs up or down?</p><p>(1:08:34) Is the database market stagnant? "A cheetah on cocaine in a Ferrari"</p><p>(1:10:14) The relational model is arithmetic; SQL as the new assembly</p><p>(1:11:59) Larry Ellison, Linux, and why databases still matter</p>