That's just a small, fast non-reasoning model.

It's almost here! My latest @oreilly.bsky.social Report is a collaboration with Red Hat to bring you the latest best practices in open-source AI platform infrastructure development for large foundational models using Kubernetes.

Also @chiphuyen.bsky.social’s new book looks great in person!

Which one of you #rstats posters is this?

Just switched to it and am really enjoying it so far!

BUT! This is terrible and largely unnecessary, if fun (for certain definitions of fun). If you're going to use regular expressions, set up your dict so that the keys are extractable, unique pieces of your string (in the example, "customers") and extract that from your input string for the lookup:

In this snippet, we set up a dict that points to Model objects using keys that are valid regular expressions representing dynamic table names. To get the right model, you use dict.items() and check each key for a match. You can build a list or just take the first match.

It’s old (I left the field a decade ago), less blocky, and focused on the brainstem (the lab I was in focused on the brainstem and periphery), but I was always partial to this style because they show rough structure of each organ:

I mean,

Bins (for evergreen notes, fleeting notes, recipes), Daily Notes, Sources (for highlights and things from books, videos, etc.), Spaces (for work projects, career projects, hobbies, etc.), System (for templates), and "The Nexus" for index notes that serve as topical hubs to evergreen notes.

I refuse to join Bluesky. I don't want to become part of a Coal Chamber.

Bluesky accounts are looking good on GitHub even if they don’t have the butterfly logo yet