I’ve gone all in on the clickhouse stack this year and I continue to be extremely impressed. Our product does physical simulation and generates a lot of time series rows per run (a million or so) so we needed something that could do fast time series aggregations. In clickhouse we’ve got one core db that supports that easily, while also being runnable locally in docker compose for dev. They’ve also made smart acquisitions in the otel observability stack, so our o11y is running on another clickhouse db with hyperdx/click stack included. And they acquired langfuse which mostly just works on top of clickhouse as well, so we’re self hosting that for evals etc.

The team has been easy to work with as well. Maybe they’ll get enshittified, I guess that’s inevitable, but it’s good now.

I think they’re a model of a co that will do well through the LLM transition - strong core tech plus useful bits on top that are OSS but can be hosted by them if you want.