TabPFN-3-Plus is all you need 🤫

Horizontal bar chart comparing agentic coding reliability across three groups. Frontier models (Claude Sonnet 4.5, Gemini Pro 3.1, GPT 4.1) score 80-100% correct. Four-months-ago's local models (Qwen 3 14B, GPT OSS 20B, Mistral 3.1 24B) all score 0%. Today's local models (Gemma 4 26B-A4B, Qwen 3.5 35B-A3B) both score 90%.

A few months ago, any LLM that I could run on my Macbook scored 0% on an agentic coding eval I put together. This month's Qwen 3.5 and Gemma 4 releases both scored 90%.

On my blog: simonpcouch.com/blog/2026-04...

notes on my benchmark of my model vs your model with my metric

HEY, you look familiar

Looks like I'll be in Houston, TX this September 👀

sounds really helpful, can i use it on the same sheet as my passwords?

spring cleaning is complete 🧹

finally removed all pipes, tibbles, tidymodels, & R from our tech stack. while R is like, 'um, yeaaa, andd.. coollll?' ? we took a hard look & decided since all data is already in Excel, and now Python is in Excel, we are now doing everything in Excel 💪🏻 plots too!

this is a lite-weight tabicl binding for R github.com/frankiethull...

i started a handful of these, including tabicl, tabdpt, limix, iltm, & a few others.

tabpfn 0.1.0 A new R package for tabular deep learning models.

We've released the first version of our tabpfn #rstats package to CRAN. This is an interface to the Python #TabPFN package.

tidyverse.org/blog/2026/03...

The model is a pre-trained deep learning model that has performed exceedingly well on every data set I've tested it on.

disclaimer:

I am an ambassador for Prior Labs & will speak about TabPFN + other tabular foundation models on here.

if you're an R programmer, you can try out the latest & greatest 2.6 via the R binding: tabpfn.tidymodels.org

TabArena - a Hugging Face Space by TabArena TabArena

huggingface.co/spaces/TabAr...

TabPFN-2.6 is out 🔥 new #1 on TabArena

does something like this exist for tidymodels workflows ❓

like a mermaid + quarto + workflows visual extension tool

expand.grid is all you need

Resilient Energy Strategy in Times of Dark Calm NASA POWER solar and wind data provide insight into demand, market pricing, and energy supply required to maintain power and keep U.S. citizens safe during dark calm events.

here's a cool article by NASA discussing energy planning in times of "dark calms" & extreme weather events in the USA.

the dark calm analysis referenced comes from my team 🙌🏻

power.larc.nasa.gov/article/resi...

he literally quit replying 😭

Alysa Liu recently went viral for her Teen Vague rant about base R data.frames

"a relic that lacks the structural integrity of a well-landed triple axel, data.frames silently corrupts your characters, drown your console, & drop your columns mid-routine. while tibbles stick the landing every time."

I forgot it was in YELLCASE. . .

8-ish years ago when I was a quant learning base R, I used MASS::mvrnorm for correlated residuals in monte carlo processes. But I don't recommend it. There is something buggy when reproducing shocks if you are varying n

um i am both of these people btw

i love data, me too meme

yaml12 is all you need

they're a 10 but they don't like tibbles

tabioka
tabbouleh
rutabaga
tabbymodels
tabasco
orangetabby
tabtab

I like how you are thinking. My late night root vegetable search history has gone wild this week.

I was also trying recipes, vetiver, or other 'foundation' ideas around cuneiform & quipu.

have a lot of "a data frame, a tsibble, & a tribble" jokes for this week

The kuzco package from @frankiethull.bsky.social made the January 2026 Top 40 CRAN #RStats packages by @revojoe.bsky.social! Congratulations, Frank, on such a great package ❤️🦙

rworks.dev/posts/januar...

GitHub - frankiethull/iltm: iLTM: Integrated Large Tabular Model with R iLTM: Integrated Large Tabular Model with R. Contribute to frankiethull/iltm development by creating an account on GitHub.

github.com/frankiethull...

screenshot contains text saying "Frank Hull and his team at ACES oversee 1,000+ production models to forecast everything from 25-year price scenarios to next-hour demand. By leveraging Posit Team and Positron, they deliver the quantified risk scenarios—like 105°F summer heatwaves or 2°F winter nights—that ensure cities and counties never run short on power."

Oh sweet, check out this Monthly Roundup from Posit 👀

It mentions my team & we really like regression 😅

more info: posit.co/about/custom...

GitHub - AI-sandbox/iLTM: iLTM: Integrated Large Tabular Model iLTM: Integrated Large Tabular Model. Contribute to AI-sandbox/iLTM development by creating an account on GitHub.

todo list: github.com/AI-sandbox/i...