This is one of the best reviews I've read on the film, very nice 🎬
23.02.2026 23:14
👍 0
🔁 0
💬 0
📌 0
This is one of the best reviews I've read on the film, very nice 🎬
I do find that, for some reason, LLMs and Claude specially runs crazy with pipes, sometimes printing 100+ lines of code in a single pipe. Some crazy code like: read_csv() |> mutate(...) |> ... |> ggplot() + ...
pipes are incredible, just keep it reasonable: chain 2-4 functions at a time. While in some cases they can make a code harder to debug, I rarely find that it makes it harder to read. Also, not even sure if some packages (like leaflet) can exist without pipes.
I'm late on both of the R quarrels.
tibbles are great, just set options(pillar.width = Inf) and be happy. Nested tibbles are underrated. Only thing that bothers once in a while is the rounding of numbers.
This is the correct way to calculate the MAD of a vector in R. #rstats
something like this
Excel is home to some of the worst data practices. Just received a sheet that's supposed to be a "heatmap". Cells are colored manually, columns are days and rows are "regions". The actual information is stored as comments inside each individual cell.
just to be clear, .direction = 'right' doesn't exist
{ggview} can print plots to Viewer (Rstudio or Positron) with true scaling. Adjust plot settings and see the final proportions immediately.
Save the plot once 👌
e.g.:
ggplot(...) +
ggview::canvas(width = 220, height = 220*2/3, units = "mm", dpi = 300)
per @nrennie.bsky.social #datavis #rstats
When I say that Claude is very good at coding in R, there are caveats.
As G10 countries succumb to simplistic populism, reckless fiscal policy and fiscal dominance of their central banks, the line between the G10 and EM is getting blurred. Real long-term interest rates in the G10 have begun to rise towards EM levels...
robinjbrooks.substack.com/p/the-global...
Closeread Prize winners are here! 🏆
Want to create engaging, scroll-based stories with your data? See stunning scrollytelling examples made with Quarto: posit.co/blog/closere...
Learn more about Closeread and its power to create dynamic narratives. Congrats winners! #Quarto #RStats #Python
Claude can be savage at times
What happens when you get lazy and don't follow your own advice...
Happy to know I'm not the only one who creates a ggplot2 theme package for every company I work with #positconf.
Not only does it standardize communication, but it also makes a great impression. theme_quintoandar() took about a day to make and led to me reporting directly to the C-level.
One strategy I'm using is to ask ClaudeCode to write a small summary of what it's done + add context for future work when it's almost done with a task. I then use /clear, ask Claude to read the document, and give it instructions on the next task.
4. Avoid auto-compacting
The last point is directly tied to this one. ClaudeCode has a context limit (the UI warns when approaching this limit) and it auto-compacts by default. The quality of compacting can sometimes be unreliable and I found it best to just workaround it.
You can ask for ClaudeCode to take notes of useful information and store it in separate md files to help build this knowledge.
A great example of useful context to provide Claude with (from @sarahjohnson.io) gist.github.com/sj-io/3828d6...
3. Knowledge
Anthropic recommends that you create a CLAUDE.md file to store important context. Currently, I'm experimenting with an entire claude/ directory stashed with useful content. Before starting a new task I ask Claude to read the relevant content.
Ex: I asked ClaudeCode to download an Excel file, import it in R, and make it tidy. The download, however, failed due to network issues and Claude decided to invent a dataset with fake values instead of stopping and pointing this error.
2. Manually approve
While ClaudeCode seems magical when it "just works" it often doesn't. Auto-accept can lead to errors that hard to detect since Claude can sometimes omit them and say that everything worked according to plan.
1. Make plans
Always start a new task in plan mode and invest some time making sure that Claude has a solid plan. A neat tip from Anthropic is to use "think more...", "think hard about potential issues...", "think step by step ... ".
"think" < "think hard" < "think harder" < "ultrathink."
Been working with ClaudeCode for the past ~2 months (mainly R and Quarto). My main tips:
1. Plan a lot.
2. Manually approve (auto-accept only for very simple tasks).
3. Create knowledge that can be referenced.
4. Avoid /compact and auto-compacting.
I've been using Air since switching to Positron and have been loving it. My only criticism is the styling of tibbles (it just breaks them). Apparently that has been solved
# fmt: tabular
coming soon
#PositConf
The fastest way to start using & exploring the new 2024 1-year ACS data?
R with tidycensus + mapgl.
With just a few lines of code, you can pull down any of the thousands of variables available in the new data and explore on an interactive MapLibre map.
Jair Bolsonaro was sentenced to more than 27 years in a Brazil prison for attempting a coup after losing the 2022 election.
I was loving Claude Code... until I tried it with #rstats. Constant errors, wouldn't use the tidyverse even when asked, "optimized" functions were slower.
Frustrated, I started a session just to teach R to Claude and summarize what it learned into a CLAUDE.md file gist.github.com/sj-io/3828d6...
Saw a post from an influencer telling followers to "stop using R for anything -- use Python like a normal person"
As a 10+ year R & Python user, the irony is that 2025 is the best time EVER to be using R.
Here's why:
