No fluff, just future: if you care about what’s actually happening at the edge of AI and data infrastructure — not the hype, but the tech — then Data Council 2025 in Oakland was the place to be. Three days of densely packed technical insight and legit networking with people actually building the future — CTOs, AI researchers, data engineers, VCs, and the folks writing the code and laying the pipes for tomorrow’s intelligent systems. Some highlights from our Mike Perlov:
✔️ Watching Hamilton Ulmer present MotherDuck’s Instant SQL in action was wild. It lets you write SQL with live-updating results — no need to hit “run.” Built on DuckDB’s local-first engine, it speeds up data exploration by eliminating the usual write-run-wait loop. Users can now inspect CTEs in real time, preview external sources, and debug complex queries effortlessly. Powered by smart caching, AST parsing, and cursor mapping, Instant SQL makes query-building faster, more intuitive, and responsive.
✔️ DuckDB Labs shared deep insights into their in-process analytics engine - compact but powerful alternative to traditional warehouses. Hannes Mühleisen’s deep dive into its high-performance architecture was hands-down one of the most technically satisfying sessions. The sheer efficiency — from vectorized query execution to morsel-driven parallelism — makes you rethink what’s possible on a single node.
But the bigger theme? Convergence. Data infra and AI are merging — fast. Semantic layers, vector embeddings, agent memory, knowledge graphs: they’re moving out of research and into production. A prime example was the Pixeltable’s multimodal AI framework, presented by Marcel Kornacker, which introduced their unified multimodal AI infrastructure. This impressive framework transforms working with unstructured data into a familiar data engineering experience. So if you’re still thinking in rows and columns, you’re already behind
On a related note, we recently launched PondPilot at T1A: it’s a browser-based, local-first SQL tool powered by DuckDB-Wasm. No installs, no cloud uploads — just open the page and start querying local data instantly. It’s our small bet on local-first tech in a world that’s increasingly hybrid. And while much of the industry is racing toward bigger, cloud-based AI systems, Data Council 2025 also reinforced a simple truth: not all data needs a cluster. Sometimes, local is all you need 😉
