It was a really great talk, you covered a lot in 15mins! I am keen to work on an OSS library for embeddings wrapping Langchain4j so may ping you for opinions/reviews if you have time?

Luckily we have 2 people on the Strimzi team who literally wrote the book on Kafka Connect and Strimiz could do something (TBD) to help.

Mostly about how everyone wants DR, try MM2, bounce off it and then end up with some very unique setups and trade offs. TL;DR MM2 (and connect more generally) is hard to get started with and has some sharp edges.

"Process" not "Polymorphic" 🤦‍♂️

So nice to get a deep technical deep dive on a subject. They did a great job of explaining the migration process. If you want more detail we have blog posts:

Ok so the main thrust is actually about removing Kafka from the mix & what you need to do in Iceberg to emulate topic compaction. A really interesting custom solution due to missing features in the Flink Iceberg connector. The headline, it works, it doesn't need Kafka but it's not as fast #Current25

Wooop Open Source Datastack FTW! Pinterest moving from an in-house to OSS based architecture! #Current25

Great to hear that OpenAI have identified and fixed several issues with PyFlink and intend to push those fixes upstream 👏👏👏

It's PyFlink for all the things and OpenAI uses the upstream Flink Kubernetes Operator

This was a very interesting talk full of intriguing ideas. However, there was a lot of "you _could_ do <complicated thing> and then do <even more complicated thing> based on that". I would love to see some real life demonstrations of all these agents linked together doing what we expect.