saved
Databricks & the Future of the Firm
Hraness wrote this summary from a saved copy of the source. Quotations are taken word for word from the source.
gist
Ben Horowitz and Michael McGuiness argue that Databricks is building the system of record that lets AI agents do real work inside firms. They trace its four S-curves from Spark to agents, then use Coase and Burnham to argue that agents will erode the professional managers' monopoly on coordination. Firms that own their data, ontology and evaluations stay sovereign when models change. a16z is a Databricks investor, and the company figures are self-reported.
ideas
- Four S-curves. Databricks moved from Spark to the Lakehouse, then Lakebase (operational Postgres on lakehouse storage), and now agents; the company reports more queries from agents than from people and over 17 million agent-launched databases a day.
- A system of record for agents. Agents need more than a model: they need shared state, an ontology of what the data means, permissions and budgets, and grading on the company's own tests, which Databricks sells as Lakebase, Genie Ontology, Unity Gateway and Agent Bricks.
- Coase, revisited. If agents cut the cost of routing information inside a company, a firm can grow larger or stay smaller than Coase's limit implied; the essay compares this with Jack Dorsey's idea of a company world model.
- The second managerial revolution. Burnham's professional managers gained power by routing information; the authors expect more people at the edges to specify, delegate and inspect work, ending that class's monopoly rather than management itself.
- The sovereign firm. Two banks renting the same model diverge when only one records facts, overrides and outcomes; owning that record lets a firm swap models without losing what it has learned.
quotes
“The second managerial revolution is not so much the death of management as it is the end of the professional managerial class’s monopoly on it.”
“for the firm to remain sovereign, it must own its model and everything the people at the edges teach it.”
“A year later, when a better model arrives, only one of them has built a better bank.”