The feedback loop you get when using @marimo_io is insane! I recently had the deepest feeling of being in the zone when trying to solve a data analysis problem. Thank you @akshaykagrawal. I consider this to be a *MAJOR* development in the @PyData stack. (@DVCorg last mjr dev)

best @PyData tooling progress i've seen in ~24 months, @marimo_io , @quarto_pub , @mitsuhiko (rye), @astral_sh (ruff), @prefix_dev (pixi).

I just did something like this with minikanren "facts". The query was just a conjunction of edges (pattern matching). Seems to work.

are we witnessing the "great rewrite of the scientific programming stack" in #rustlang ? #pydata

can't help but use @mrocklin 's voice for my internal vocalization when reading @dask_dev docs! :D

recently solved a rather complicated data engineering problem at work (but i identify as a data /scientist/), points: * solving it was harder than some of the math modeling that i've done (time and thinking) * there is a big gap b/w data reality and fancy models

If I understand you correctly, the answer that you're looking for is in the status cmd dvc.org/doc/command-referenc…

blame that on employers that take the easy path of qualification by tech compliance instead of assessing problem solving ability.

Shout out to pyprojroot @chendaniely !

Two separate issues there: the source and the delivery mechanism. As far as delivery mechanism, I want my RSS feeds back!

I like it too but it's got a hard, complicated tech dep: K8/docker. There was nothing like it when it came out. Nowadays, the DAG idea has been beaten to death. You should be able to define (data) pipelines abstractly and have pipeline executors. My eyes are on @dagsterio.

Agree but that's a separate activity. I was talking about the generic reproducible pipeline problem. Each stage in the pipeline is an arbitrary file including complicated binary models. The versioning here is on the file level.

Interesting but it's too specific. Initially you're working with files and you want the full reproducible pipeline where the content of the input-output files is arbitrary. Dvc solves that data science concern with one tool. covers the basics generally. Then get the specific tool

Working on this. Assume multilingual. 1. Create directories for each exploration/task/module/package (no distinction between them) 2. Automate setup of each dir 3. Evolve each dir with source control Specifics: conda, git, and dvc solves basic data science issues . @thos_jones

... The situation has de-volved into centralized, commercial services. Only email has survived this trend. Back in the mid 2000s I was hoping that xmpp would be an example of a federated messaging application (like email).

If I had spent that much I would be tempted to just play games on it with max details.

Marking 10 years since I first used #Python which was motivated by scientific computing. Thank you @sagemath for the introduction! Self-reflection: Ctrl-C (aborting)

i hate hardware that has 'smarts' builtin esp. when the h/w is usable w/o it. * can i please have a 'dumb' tv? i can plug in the 'smarts' via hdmi thankyouverymuch. * can i please buy a car without an infotainment system that's trying too hard to be like a tablet computer?

@marskar i think this is what i've been looking for #streamlit

"Many articles about functional programming are written by nerdy mathematician assholes. Reading them without preliminary training is dangerous: categories and morphisms can blow your mind in exchange for nothing." hackernoon.com/two-years-of-…