Maximilian Scholz

Infrastructure (and pokemon) for Bayesian workflows, simulating everything
Music, cooking, exercise enthusiast.

[Maximilian Scholz @scholzmx]

What would finishing your project look like if it was an act of self-love instead of self-violation?

[Maximilian Scholz @scholzmx]

@maria_rcks Handing them.out like Scooby snacks

[Jacob Miller @pwnies]

A bunch of people were asking how I made the effect for the "renderaissance" style page I made previously.

Here's a full walkthrough of creating that depth / dynamic lighting effect for the web.

Interactive version: https://t.co/CdEXq46TXS

[Maximilian Scholz @scholzmx]

@maria_rcks

[Maximilian Scholz @scholzmx]

The type data behind ry now lives in its own repo at https://t.co/XLtCNviZBu

Currently holds JSON stubs for ~30 packages. Partly generated from installed packages and their docs and the CI checks them against the real namespaces.

[Maximilian Scholz @scholzmx]

ry 0.4 is out: https://t.co/uGGSjZI4Ta
Should understand tidyverse NSE and reduced false positives on the top 300 R packages by about 75%

Also added --write-baseline to snapshot existing findings so only new ones fail CI and --min-confidence to get less noise.

#rstats

[Maximilian Scholz @scholzmx]

@MomaBera The anime-edgelord-mc phase? In 2026? A classic.

[Maximilian Scholz @scholzmx]

@samgoodwin89 Tell your agent to adapt this to cc and cursor: https://t.co/k2XdQC0pP8

[Maximilian Scholz @scholzmx]

@MomaBera *Me after submitting 10 PRs to R packges my new typechecker found real bugs in with a newspaper in his hands*
I should buy some evening cuddles...

[Maximilian Scholz @scholzmx]

I had an agent run ry on the top 300 most downloaded R packages and besides a huge amount of false-positives, there were some real bugs in the results. The first 10 PRs with bugfixes are out. 0.4 is also on its way with way fewer false positives. #rstats

[Maximilian Scholz @scholzmx]

ry 0.3.0 is out. Reduced the false-positives a bunch (and also found a minor test-bug in brms, yeay).

We got S3 dispatch for operators, better data-frame schema tracking, typeshed stubs for testthat, better NSE handling and more. #rstats

https://t.co/LZqEloMsc4

[Maximilian Scholz @scholzmx]

@jrosell @sama So far I am only deriving information that already exists but I've pondered type hints as well. Currently working on a big 0.4 rework that should massively reduce the number of false positives as I've had an agent run ry on the top 300 R packages and collect the results :D

[Maximilian Scholz @scholzmx]

@jrosell @sama I have a growing list of real bugs in R packages that I found with it :D kinda cool.

[Maximilian Scholz @scholzmx]

@clairemcwhite or @AchimZeileis I found a bug in colorspace with my static type checker. PR's here if you want to take a look: https://t.co/inMzdyMTlX

[Maximilian Scholz @scholzmx]

@MomaBera Having experienced this dynamic in the opposite direction, I think this is vastly compounded by socialisation and highly individual. Experiencing a well attuned partner for the first time was eye-opening. Only took 34 years though.

[Maximilian Scholz @scholzmx]

@theo Mind sharing those skills?

[Maximilian Scholz @scholzmx]

@zekramu Nah, not gonna play semantics with you. It is cool that it exists nonetheless. But it's not built in a principled way.

[Maximilian Scholz @scholzmx]

@portrays @theo @kyle_mccleary But... omp IS slop. There's no denying it. Slop can be useful and cool. But it is very obviously slop.

[Maximilian Scholz @scholzmx]

@Thomasdelvasto_ Maybe put down the phone and talk to a professional instead.
Until then, some chill vibes
https://t.co/djqlvxzeBn

https://t.co/IOC45X1u22

[Maximilian Scholz @scholzmx]

@sheherenow_ How do you get these haptic UIs. Mind sharing a prompt?