Andy Pavlo joins ClickHouse to establish ClickHouse Labs

Most reactions were simple. People see this as a strong recruiting and prestige move for ClickHouse because Pavlo is one of the best-known public faces in database systems. The more useful conversation was about what kind of work a lab like this is for. Several people pushed back on the idea that a research lab is just a fancy title for engineering. In databases, the hard parts often live in the gap between known theory and production reality. Companies need room to explore indexing, caching, query execution, storage layout, and other bets that may take years to matter or may never ship directly. Another thread connected the hire to a broader shift in analytics infrastructure. ClickHouse, StarRocks, and similar systems are increasingly converging with Trino around decoupled storage and compute, plus direct querying of open table formats like Iceberg. That raises practical questions about what remains special about a database engine if data lives in object storage. The answer people kept coming back to was indexing, native formats, and engine-level optimization. But the line is getting blurrier. One commenter even claimed ClickHouse can now run some Iceberg workloads faster end to end than its own MergeTree format if you throw enough compute at the scan, which is a strong sign that raw scan performance is no longer the main moat. The academic angle landed in a more nuanced place than the post's marketing copy. A few readers lamented that industry labs are absorbing work universities used to do, especially while government funding is unstable. Others said database research in academia is still healthy, and that industry becoming a competitive home for researchers is good news rather than a warning sign. The takeaway was not that universities are finished. It was that frontier database work now needs both settings, and the commercial side has become much more attractive.