So nice to see this model grow! The term sync engine wasn't really in vogue back when I wrote Flushout, but decided to finally add it to the project description
github.com/saarw/flushout

Where speed matters, there are probably many systems that won't even put analytics in a DB and instead post the data from the app directly on some queue to the data warehouse

Question is how many interviewers have experience with this kind of system themselves to evaluate solutions

Database use looks very different in the 100M DAU consumer apps I've worked on vs enterprise apps etc

Are code AIs any help at all for a project like this?

For other languages I'd expect them to provide a helpful starting implementation, but guessing Claude and GPT-4o are worse at TLA+ than Python or TypeScript 😅

What became known as stream processing for applications (or "kappa architecture") was how you built low-latency trading platforms a decade earlier

Traditional databases were too slow so you needed to replicate events to hot-standby redundant in-memory systems (stateful stream processing nodes)

That looks useful for adding version handling to data in apps!

Yes, how Bluesky will monetize is an interesting question... but even so, with a federated protocol you're going to get communities outside of certain markets, just like websites linking to each other from different countries

Has any US entity without an EU presence been successfully targeted by the European regulatory regime?

National regulations are aimed at international companies with some local presence that the national regulator can go after, so I'd expect federated communities to help here?

Time to rewrite in Rust