This talk explores how ideas from decades of computer science - semantic triples from 1970s AI research, Datalog from logic programming, Probabilistic Search Trees from modern distributed systems - combined in novel ways address existing technical challenges in local-first software.

The presentation delivers a crash course in each technology before demonstrating how their synthesis enables: query-driven partial replication, cross-application cooperation through schema-on-read semantics, and fully reactive UIs built with declarative Datalog rules. All syncing through commodity blob stores (S3, R2, IPFS) without coordination servers.

Drawing from Dialog DB, the presentation covers what works, what doesn't, and why this specific combination matters. Attendees will leave with both theoretical understanding and practical knowledge to apply these techniques in their own local-first projects.