• LLMs are the missing link to shipping side projects

  I’ve been procrastinating on shipping a small UI improvement to my blog, and an LLM made it happen.

• Python 3.14 will change the way you parallelise code

  Free-threading in Python is finally here. Here’s why it matters and how I plan to use it.

• On good interfaces

  Didn’t think a tech book on fundamentals could strike me so much.

• Run Python functions on K8s

  Inventing once more. This time, a decorator to run Python functions on K8s, straight from notebooks!

• Should we replace Spark with DuckDB?

  As we’ll see, only for some queries (but we can use ML to predict for which ones!).

• FOSDEM 2025 has exceeded my expectations

  Commercial conferences don’t hold a chance against it. Arrow, WASM, Graph DBs, Distributed Systems, CNCF, and more.

• I tried using wheels as a model package format

  Inventing a model package format. Spoiler alert: MLFlow is here to stay.

• Uv has a killer feature you should know about

  uv run - install, activate, and run in one command…

• Async I/O Is Not Enough

  Gonna run some benchmarks on CPU-heavy workloads to illustrate why achieving true parallelism in Python is so hard, why async I/O alone has little to offer there, and how No-GIL changes the game - or just using Go instead.

• Let's build a Bloom filter. Part 1 - A Space-Efficient Hash Set

  Implementing a Bloom filter from scratch (in Rust 🦀) to understand what makes them tick. Apparently, they’re very similar to hash maps.

• A Feature Store, or Just a Warehouse Table?

  Is a feature store really necessary when you already have a SQL analytical database?

• Walrus operator

  Exploring the new Walrus operator in Python. Neat syntax sugar!

• No Time to Plan

  Reflecting on the art of acing without a thought-out plan.