tinypkgr just hit CRAN!
The whole tinyverse development toolchain (tinyrox, tinypkgr, rformat, and saber) is now one install.packages() away.
cornball.ai/posts/tinyve...

mx.api 0.1.0 is on CRAN! 🎉
An R client for the Matrix Client-Server API
Chat between R and @matrix.org
Login, rooms, messages, history, media. E2EE next.
#rstats
cran.r-project.org/package=mx.api

After 'helping' @dfalbel.bsky.social land the torch 2.8 bump, I vibe-coded a Linux-only libtorch wrapper.

Also, the 3rd use of ASTs I teased: compilers!
TorchScript is on the way out in favor of Python-only TorchDynamo; this is a path for R torch users

#rstats
cornball.ai/posts/torcha...

Added Kimi support after seeing this:
bsky.app/profile/royf...

llm.api is on CRAN!

Chat with the usual LLMs (OpenAI, Anthropic, and Ollama, plus Kimi!)
Agent loop with tool use and an MCP client for R, too!
Just curl and jsonlite under the hood, API design derived from ellmer.

install.packages("llm.api")
#rstats *got it this time*

litedown would be the more natural partner for that job. Trying it now.
pkgdown pulls in 53 non-base recursive hard deps, litedown pulls in 2 (commonmark, xfun). This is the whole point of the tinyverse philosophy.
Thanks for building it @yihui.org!
github.com/yihui/litedown

𝚝𝚒𝚗𝚢𝚝𝚊𝚋𝚕𝚎 0.6.1 for #Rstats is out!

It's an ultra simple, super flexible, and 0-dependency package to draw beautiful tables in HTML, LaTeX, Typst, Word, PDF, and PNG.

And for those who ❤️ documentation, 𝚝𝚒𝚗𝚢𝚝𝚊𝚋𝚕𝚎 ships with a billion pages of tutorials:

vincentarelbundock.github.io/tinytable/

Definitely slowing it down for me!
I find it easy and worthwhile to vibe code minimal dependency R packages.
bsky.app/profile/troy...

#rstats!
because I’m good at Bluesky

tinyrox is on CRAN!

Minimal roxygen2 alternative. Zero non-base R deps, generates your .Rd files and NAMESPACE

`install.packages("tinyrox")`

Part of the tinyverse toolchain:
cornball.ai/posts/tinyve...

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/04/12#littler_0.3.23 describing release 0.3.23 of CRAN package littler.

littler 0.3.23 on CRAN: Mostly Internal Fixes
Everybody's favourite way to rock R at the command-line and in scripts
dirk.eddelbuettel.com/blog/2026/04...
#rstats

inst/ ?

Cool to see @tylermw.com arrive at the same conclusion: AST-based analysis of R code beats token parsing… for AIs and developers!

tarborist does this for {targets} pipelines in IDEs. saber does it across all your packages for agents like #ClaudeCode and #Codex

🌲 #rstats
cornball.ai/posts/tokens...

bsky.app/profile/alex...

If they’re doing install_github, sure. But the advantage of CRAN over npm is precisely that it’s well curated

First one-shot submission!

To directly answer your question:
astgrepr does the second one. saber combines the AST with R’s well-defined DESCRIPTION files to do the third

saber uses the AST (getParseData in this case, but could be treesitter too) to accomplish the last two goals.

You intuitively understand these relationships. Your AI does not!
3/3

The AST of a single function will help a formatter format it.
The AST of a single package can help us understand the structure of the package.
The AST of your entire code base can help us understand how different packages and their functions relate to each other.
2/3

My brain’s the same!
The AST, whether it’s from treesitter (fast, all the languages) or utils::getParseData (batteries included base R), is a tool that gets you the structure of the code. “This is a function. That’s a constant. This function calls that one”, etc.
1/2

The LLM profvis work discussed in the post:
bsky.app/profile/emil...

The LLM skills/copilot work discussed in the post:
bsky.app/profile/jont...

github.com/cornball-ai/...

New R package: saber
Claude Code keeps forgetting. `saber` builds the graph it can't hold in its head.
LLMs work on tokens. Code lives in trees.
#rstats
cornball.ai/posts/tokens...

Balasubramanian Narasimhan & Anqi Fu — Development Plans for CVXR YouTube video by CVXPY

I very much appreciated this talk by R Foundation member Balasubramanian Narasimhan. Not just because the content and slides are great, but because it helps to legitimize the use of LLM CLI coding tools in the R community:
www.youtube.com/watch?v=Zvou...

#rstats whisper Release
~18x speedup by fixing things that shouldn’t have been broken in the first place: a pipeline object, SDPA, hard coded constants.
Plus multilingual support and per-word/token timestamps now.

rformat 0.1.0 is on CRAN. Base R-inspired code formatter built on utils::getParseData(), R's built-in AST. One Import: Rcpp, so it's lightning fast.

install.packages("rformat")
rformat::rformat_file("my_script.R")

#tinyverse #rstats

years of base #rstats meant I had no problem reading this

I just hate tibbles so much.
(don't @ me)

Feb. update to the LLM+R guide 💪

8 new packages including:
code review, predictive modeling, speech-to-text, text-to-speech, HuggingFace integration, Gemini CLI companion, a CLI coding agent written in R🤯 ,and more!

available in English 🇱🇷 and Spanish 🇲🇽
luisdva.github.io/llmsr-book/
#rstats