When changing outputs, I consider how people use the results. If outputs are intermediates used with your other (updated) functions, most user code may not even break. I went through this with fable's forecast output, most users used autoplot() or hilo() with the output, so very little code broke.

I'll still be teaching classes in RStudio for now, and probably for at least a year longer. While I strongly prefer using the code editor and personally have trouble with inline outputs, I do hope that they add them to (at least) the visual editor. I find my students like inline outputs too 😊

From >10 years of RStudio for R, and the occasional vscode or basic text editor for other languages that RStudio was too frustrating to work with.

Have you tried it? I'm also mostly desktop use with some cloud, and I've found both experiences to be better in Positron so far.

Yes! I know some people love the git pane in RStudio, but I sure don't. I'm using Sublime Merge at the moment, but I'll be trying out the built-in git client 😊

The shiny app is run on my own server using the rocker/shiny docker image.

It's also of course open-source, see how it was made here: github.com/mitchellohar...

Many thanks to Sam Parmar (github.com/parmsam) for making a similar app for posit::conf(2025) that I could adapt and extend for JSM2025.

It's my first time experimenting with RAGs for LLMs, and it was surprisingly easy to do in #rstats with ragnar, ellmer, and shinychat 🤩