I'll be at posit::conf(2026)! if you see me, ask for a sticker! I have Prior Labs stickers for all the tabular foundation model geeks 🤘🏻
Ambassador Frank Hull
@frankiethull.bsky.social
ambassador for prior labs 🗣️ enjoyer of the tibble ⌨️ ai/ml in production 🤖 https://github.com/frankiethull
i kind of remember testing this out from Positron IDE, also kind of don't remember because it was so easy to publish frankiethull-tabicl-studio.share.connect.posit.cloud
TabICL Studio
frankiethull-tabicl-studio.share.connect.posit.cloud
It’s the end of an era: we are sunsetting rpubs.com, quartopub.com, and shinyapps.io in favour of connect.posit.cloud. These services are important, but we need to focus our efforts on a single platform (and it’s overall less confusing to have one place to publish everything you do in R or Python).
I'm happy to share `ir`, a new command-line tool for running portable R scripts and Quarto documents. It's inspired by my two favorite parts of `uv`: self-describing scripts and tools you can run without installing first. opensource.posit.co/blog/2026-07...
ir 0.1.0: self-describing R scripts and Quarto documents
`ir` is a new command-line tool for running portable R scripts and rendering Quarto documents whose package requirements, and optional R selection, live inside the file itself.
opensource.posit.co
after months of thought & given how the space has expanded i think it should be called {wtf} Wrappers for Tabular Foundation models
great, now we need a unified tabular foundation model interface in R i wonder what the library name would be 🤔
🚨🚨🚨🚨🚨 RELEASE UPDATE As part of the 250-year anniversary of US Independence - GoatBoost will no longer have a public repo release. Instead, GoatBoost will be rolled out case-by-case, customer-by-customer, on a government-approved rollout plan.
btw, this is a real benchmark from my newest ML engine called GoatBoost
EBMs are really slick. whatever happened in 2020 to divert attention from this must've been bonkers interpret.ml/docs/ebm.html
Explainable Boosting Machine — InterpretML documentation
interpret.ml
btw, this is a real benchmark from my newest ML engine called GoatBoost
something new in daisugi, support for XGBoost, LightGBM, & CatBoost but not in the way you know them frankiethull.github.io/daisugi/arti...
gaiden
frankiethull.github.io
Working on a few new Tabular Foundation Model bindings for R
teammate made this for me
Explainable Boosting Machine is all you need
you can now use chronos 2 with R via brulee in tidymodels 👀 brulee.tidymodels.org/reference/br...
Chronos-2 pretrained forecasting model — brulee_chronos
brulee_chronos() loads a pretrained Chronos-2 time series forecasting quantile regression model from HuggingFace and ingests historical ("context") data so that the returned object is ready to forecas...
brulee.tidymodels.org
#RStats challenge and potential community project: * Create a minimal package 'dumpsterfire' that triggers as many 'R CMD check' issues as possible! Fun, but could also become a excellent resource for illustrating and understanding package check issues Here's a start: github.com/HenrikBengts...
here's the glossary of terms, models, etc. frankiethull.github.io/daisugi/arti... my favorite part of the project might be the TLDR model explanations. i wrote a haiku for each machine 🌳
glossary
frankiethull.github.io
daisugi v0.0.7 building out the not-so-well-known booster package. feel free to check it out, let me know what you would like to see. i also started tidymodels bindings for EBMs & Yggdrasil in 0.0.7
daisugi v0.0.7 building out the not-so-well-known booster package. feel free to check it out, let me know what you would like to see. i also started tidymodels bindings for EBMs & Yggdrasil in 0.0.7
our two goats, Theon y Arya! they are nigerian dwarf goats, which are known for being great pets. they are loving our backyard, browsing for weeds, playing on tables, slides, etc. they kind of remind me of a curious cat but playful like a dog.
@t-kalinowski.bsky.social saved me all the headaches mentioning py_require a few weeks ago. It saves so much time & works great 💯
reticulate::py_require is all you need
this made me wonder what other weirdo booster methods existed & decided to start curating for R frankiethull.github.io/daisugi/
Novel Tree Machines for Classification and Regression
daisugi provides an interface for not-so-well-known tree machines. particularly focused on novel tree and boosting algorithms for machine learning tasks. those tasks being tabular regression and class...
frankiethull.github.io
for context, stumbled across this unique parameter so that "each leaf is linear instead of constant" lightgbm.readthedocs.io/en/latest/Pa...
this has been a really fun spring in the backyard. - doubled our veggie garden - planted six fruit trees (4 bare roots, 2 wild plum transplants) - adopted two goats the goats are the newest & most interesting. they follow me around like puppies & eat the weeds i toss 🐐🌱
I'm officially a nigerian dwarf goat farmer 🐐
I'm officially a nigerian dwarf goat farmer 🐐
parsnip::set_engine("lightgbm", counts = FALSE, linear_tree = TRUE) is all you need