croft
The "one DuckDB file" bit is what got me. For a scrappy startup, the operational overhead of data pipelines is often the silent killer. You're not just writing transformations, you're babysitting Airflow, Kafka, and a warehouse that costs more than your AWS bill. Croft is essentially saying, what if the whole thing was just a file you could copy, version, and ship? That's a compelling pitch for a team of three trying to move fast.
I think the real signal here is the continued march of DuckDB into places people never expected a database to go. It's already the go-to for local analytics, and now it's creeping into the pipeline layer. The TypeScript angle is smart too, it meets the massive Node.js ecosystem where it lives, instead of forcing people into Python or Scala. I'd want to see how it handles genuinely massive datasets or complex streaming scenarios, but for a huge swath of startup use cases, "good enough" with a fraction of the complexity is a winning trade. This feels like the kind of tool that quietly becomes the boring, reliable backbone of a hundred products.