$2M idea make car headlights that are always warm yellow

large large language model the size of a small large language model

There is going to be a big wave of "omg what if you didn't treat this as an answer machine" "what if we rediscover active learning" commentary from CS in about six months, especially the voices currently describing everything as "brain rot and cognitive decline," and I'm already a bit tired

This is a really neat idea but I'll never remember the commands, so I created a little `git-recon` bash script that runs them in series, complete with some ascii bar plots.
gist.github.com/gadenbuie/46...

First, install it (on Mac that's `brew install mergiraf --HEAD` to get the latest with R support). Then set it up as a merge driver mergiraf.org/usage.html

After that, you just use git the same way you're used to, but some things that would have been merge conflicts will just merge cleanly!

That’s how you know I’m human 😂

I am a human and I am stating my agreement with the premise of this post

If they had launched on March 31 this could have just slipped past everyone as an April fools joke

feat: R language support Hi! This PR adds support to mergiraf for the [R programming language](https://r-project.org) using the [tree-sitter-r](https://crates.io/crates/tree-sitter-r) crate. The result is syntax-aware mergin...

It's a very small contribution to this list, but also `mergiraf` also now uses `tree-sitter-r`! codeberg.org/mergiraf/mer...

Kermit the frog screaming with excitement

We have summer internships y'all! Come work at Posit on the PyData, tidymodels, shiny, or Connect teams: grnh.se/tigz810a3us. You will have an awesome time, learn a ton, and help advance our open source and pro tools 🧰 #rstats #pydata

We’re thrilled to welcome Sara Altman and Simon Couch of the Posit AI Core team to the #positconf 2026 keynote stage!
They’ll be sharing a practical, hype-free look at how AI is being thoughtfully integrated into the tools you use every day.
✨ See what’s next for AI in open source: pos.it/conf

Screenshot of Kagi's translation tool translating "I got a new job" into "LinkedIn speak," producing an overly enthusiastic corporate-style announcement with superlatives, gratitude, and a rocket emoji.

Not sure how we went from plain, straightforward words to LinkedIn speak but now you can use Kagi Translate to fit right into that crowd:

translate.kagi.com?from=en&to=L...

The flip side of this for enterprise Claude users is getting work done early east coast hours before the west coast wakes up

oh yeah, that’s a great idea

feat: add btw CLI via Rapp for command-line access to R tools by gadenbuie · Pull Request #176 · posit-dev/btw Summary Adds a btw CLI that wraps the package's tool groups — docs, pkg, info, and cran — so that both humans and LLMs can invoke them from the command line via Rapp. Key pieces: exec/btw.R — ...

In a PR still if you want to preview and kick the tires

github.com/posit-dev/bt...

Humans AI Agents
🤝
CLI tools

Coming soon to btw, an #RStats cli tool for agents and humans

Softerware Keep your Claude skills loose for better performance

Softerware (new post) share.google/T1AajhhxRprb...

Does anybody here use `gh` with 1password and have a good way to let Claude Code use `gh` without constantly needing my fingerprint?

at least one person (me): YES

my hand-wavy theory is that it's added deterministically by Claude Code so the model wasn't totally wrong

Claude Code settings - Claude Code Docs Configure Claude Code with global and project-level settings, and environment variables.

I think it's (still) configurable. You should be able to turn it off with this in your settings.json:
`"attribution": { "commit": "", "pr": "" }` code.claude.com/docs/en/sett...

You're doing awesome work and we want to hear your stories! Submit a talk for posit::conf(2026) - Today is the last call for talks!

Oh yeah they do tend to drift quickly! I wouldn’t say I trust Claude summaries any more than human summaries (those drift too) but they at least provide some hints about what the author thought they were doing.

As a detailed PR describer, I just really dislike empty descriptions

Haha I’m saying I’ll definitely change my mind when Claude starts adding “any update?” comments on years old issues

Maybe it’s a pick your poison thing, but I’m way happier with Claude’s description than the bare “Fixes #123” PR descriptions

Still, I agree I want to work with humans and even if we’re all using agents you can tell when you’re getting more robot than human

That said, it’s important to protect the human to human interactions. The issue/description is just a starting point, I still want to know there’s someone else on the other side of the computer

Until humans get better at writing issues and describing PRs, I’ll happily accept AI-assisted submissions

Sorry but I get shouty levels of salty every time I try reveal’s auto animate feature. It’s always so so close to what I want but it’s entirely impossible to close the gap and I end up giving up

AN AUTO ANIMATE THAT ACTUALLY WORKS

You Can't Make Me Vibecode On the historical parallels between vibecoding and open source

Reminds me of davekarpf.substack.com/p/you-cant-m...

The thing that stuck with me: most people don’t want to have the kind of relationship I have with computers