If you survive the ritualistic disembowlment, she may never let you do laundry again.

My wife washed my shirts with a load of towels once when we first started dating so I've been doing all the laundry for last 8 years. This could work out in your favor in the long run.

I'm kind of curious if plotly can do varying opacity to get a gradient fade effect based on a density function similar to {ggdist}'s stat_slab_interval(). plotnine and seaborn aren't capable of it, so I had to resort to matplotlib but in theory, D3 should be more powerful.

Not gonna lie, I've never seen a case that didn't involve spatial data where interactive graphs were actually necessary 🤷‍♂️

Would I have bothered doing that if I couldn't vibe code myself 90% of the way there such that I only had to spend one day getting to my desired outcome? Probably not.

The appeal of matplotlib is its a low-level API that can do just about anything if you're willing to write a few hundred lines of code. I had to port some parts of {ggdist} to python and used matplotlib only because the other plotting APIs shit themselves if you try varying opacity based on a PDF.

Part of the reason people develop a deep hatred of python (I was once one of them) is because the "traditional data science stack" of pandas + scikit-learn/statsmodels + matplotlib is a steaming pile of garbage. Polars + PyMC + Arviz + plotnine (optional) is much, much less painful.

GCP, on the other hand, is much, much more R-friendly and both are relatively easy to use in that ecosystem.

If we're talking about building systems that go into production, I've built systems in R and Python, and it really depends what ecosystem you're in. Doing anything with R in AWS is a massive pain in the ass compared to python (as I learned when I had to write a custom Lambda runtime in R).

So, this kind of depends what we mean by "data science." If we're talking about ad-hoc statistical analysis, R is clearly better but who really gives a damn what language you use?

I hate when it tells me that there's something wrong with my design matrix construction functions when my very thorough unit tests for those are all clear and it's suggested fix breaks things

I still have lots of misgivings about AI generated code because its so often bad, but using it for code review is nice because it can do in 30 seconds what it would take a human a few hours to do. Makes for a nice first pass before human code review.

Yes, but also do this for people who refuse to vaccinate their children.

Yup, traded in my 24' 4 cylinder on a 25' DH back in mid-October. I love it so much, but man the city mileage is bad 😬

The Dark Horse has a modified 5.0L Coyote V8 with 500 horsepower and 418 lb-ft of torque. Mine is a 10-speed automatic transmission

Stop lights and stop signs are basically my greatest enemy now 🙁