FEDIDEVS

Pair programmed with my mum for 3 hours on some of the notebooks for #nbpy vectors talk tomorrow.

Nice to have an excuse to geek out with her! Our programming interests usually don't overlap so much.

Learning about marimo, an open-source reactive notebook.
Looks really nice for visualizations.

https://marimo.app/?slug=55kclj

Plus, notebook changes results in small diffs and the notebooks themselves are executable as Python scripts!

#nbpy

My #nbpy talk on a "visual explanation of vectors" was entirely in the form of Jupyter notebooks! Check them out here:
https://github.com/pamelafox/visual-exploration-vectors

Try them with your fav embedding model!

TIL: the hypothesis package has a "ghostwriter" functionality that'll write your tests for you:
https://hypothesis.readthedocs.io/en/latest/ghostwriter.html

It includes autodetection of "roundtrip" functions, like compress/decompress:
https://zhd.dev/ghostwriter/?q=gzip.compress

Here's the awesome regex for roundtrip funcs:
https://github.com/HypothesisWorks/hypothesis/blob/d8c17832141f587fec9ef89895ed01a4e9c1650d/hypothesis-python/src/hypothesis/extra/ghostwriter.py#L1112

#nbpy

Observation: in my talk, I mentioned reading the highly readable pgvector code, and in @lukas's talk on optimizing postgres, he suggests reading the highly readable postgres code.

So maybe postgres is the DB with the most readable code?

#nbpy