Just updated my Reproducible Project Template so that it now generates Word (.docx) versions of the paper. Really useful for anyone writing reproducible papers in Quarto (.qmd) files who collaborate with people who prefer tracked changes 🚀 robinlovelace.net/reproducible...
🚀 The Complexity72h Call for Projects is OPEN! ✨
Got a project you’d love to explore? Bring it to Complexity72h! Turn your idea into a real project in 72 hours of collaboration, creativity, and constructive chaos 💥
🌐 complexitynextgen.org/complexity72h/call-for-projects
Quack, quack! A DuckDB v2.0 preview is here! 🦆
In this post, #DuckDB co-creators Mark Raasveldt + @hannes.muehleisen.org preview the headline features of DuckDB v2.0 (expected to land this autumn):
• DuckDB as a server
• Triggers
• async I/O
• + more
Dive into the deets:
duckdb.org/2026/08/17/d...
#rstats#osrm.backend 0.4 now uses it's own repo with @osrm.bsky.social binaries that are packaged to be self contained (shipping with all required) libraries and immutable. Details at www.ekotov.pro/osrm.backend...
In osrm.backend v0.3, I am sharing more improvements that I made for myself with the R community: improved handling of multiple OSRM servers running in parallel, support for latest v26 of OSRM binaries and also brand new shiny app GUI based on `mapgl` www.ekotov.pro/osrm.backend/#rspatial#rstats
Great to see person-centered pages on the new R observatory! Love the frequent collaborators section in those, and especially the collaborator network view! See mine at r-observatory.thecoatlessprofessor.com/authors/Egor...
Checking on an R package usually takes five open tabs: CRAN, GitHub, download counts, reverse deps, check results.
R Observatory collects all of that on one page, for every CRAN and Bioconductor package.
r-observatory.thecoatlessprofessor.com#rstats
Checking on an R package usually takes five open tabs: CRAN, GitHub, download counts, reverse deps, check results.
R Observatory collects all of that on one page, for every CRAN and Bioconductor package.
r-observatory.thecoatlessprofessor.com#rstats
🚨 Good news for #OpenStreetMap folks: new #pyrosm is out after a long break! 🥳
pyrosm reads PBF files efficienty into #geopandas GeoDataFrames. You can parse buildings or street networks for whole countries. Here, we extracted all buildings from South America in 3.5 mins.
More below 🧵👇
#rstats mapgl is now on CRAN with flowmaps support! As previewed at #MobileTartu (Mobility Lab, University of Tartu) and Association of Geographic Information Laboratories in Europe (#AGILE) conferences!
Animated, hierarchical origin-destination commute flows in Dallas-Fort Worth, TX.
Downtowns, medical districts, airports, and suburban job hubs light up from 2023 LODES data.
Now available in R's mapgl 0.5.0.
Credit to @ekotov.pro for the implementation and @ilyabo.bsky.social for Flowmap.gl!
The upcoming v0.5 release of mapgl for #rstats is going to be massive.
Included: a slider control that allows for interactive data filtering - no Shiny required.
Thanks to @ekotov.pro - brush a window over a histogram, see your data filter in real-time
Lots more examples to share this week
🦆 Huge milestone for {duckspatial}! We just crossed 10,000 downloads and 100+ GitHub stars! ⭐
What’s new in v1.1.1.:
🟢 New geometry functions
🟢 DuckDB storage version control
🟢 Faster dplyr methods
🟢 Better connection management
Get it at github.com/Cidree/ducks...#rstats#duckdb#rspatial#gis
OSRM v26.6.1 is out 🚀
Highlights: fixed OSM node ID truncation in route/match annotations, faster MLD path unpacking, better access tag handling for car and bike, and a trip service refactor. Try the new demo before it‘s merged into the frontend at project-osrm.org/trip.html#OSRM#OpenStreetMap
Announcing tabpfn 0.1.0 for #RStats! 📊
Bring foundation models to your tabular workflows. TabPFN uses "in-context learning" to deliver high-accuracy predictions without traditional training cycles.
Learn more: opensource.posit.co/blog/2026-03...
Portable R 4.6.0 is now live for Windows and macOS. No installer, no admin, no system changes. Just extract the archive and run.
macOS (Apple Silicon + Intel): github.com/portable-r/p...
Windows (x64 + ARM64): github.com/portable-r/p...#rstats#macos#windows
🕵️♂️ Is your privacy filter model being "distracted"? I tried the new #OpenAI Privacy Filter 1.5B MoE model.
✅ "Alice password is LastOcean123" -> 100% conf LastOcean123 is a password
❌ "Alice Trump password is LastOcean123" -> 0.19% conf LastOcean123 is a password
See thread 🧵👇
⚠️ Important Disclaimer:
This was a quick, preliminary exploration triggered by a specific GitHub issue github.com/openai/priva... , not a formal scientific study.
These results are based on a narrow set of prompts and names. While the "signal collapse" in these cases is real, more comprehensive testing is needed to understand the full scope of this behavior.
🕵️♂️ Is your privacy filter model being "distracted"? I tried the new #OpenAI Privacy Filter 1.5B MoE model.
✅ "Alice password is LastOcean123" -> 100% conf LastOcean123 is a password
❌ "Alice Trump password is LastOcean123" -> 0.19% conf LastOcean123 is a password
See thread 🧵👇
* Common nouns used as names (like "Petit" in France or "Colombo" in Italy) were among the biggest distractors, often bypassing detection entirely.
* Other figures like Merkel or Musk didn't distract the model at all in this specific test, I have no idea why exactly.
🛡️ The Takeaway:
If you were thinking that using this new privacy sanitation model locally as hooks in your local agent with a big remote model is a safe choice for working with sensitive data - it is probably not.
🕵️♂️ Is your privacy filter model being "distracted"? I tried the new #OpenAI Privacy Filter 1.5B MoE model.
✅ "Alice password is LastOcean123" -> 100% conf LastOcean123 is a password
❌ "Alice Trump password is LastOcean123" -> 0.19% conf LastOcean123 is a password
See thread 🧵👇
I expanded the research to a global dataset of common surnames and across the US, Germany, France, Italy, and Spain (sourced from **Forebears.io**). I also added a few random Surnames of public figures.
The results were surprising and inconsistent:
* Certain names like Scholz (Germany), Bezos (US), and Macron (France) cause confidence to tank. Adding the German Chancellor's name (Scholz) dropped detection to just 3%.
🕵️♂️ Is your privacy filter model being "distracted"? I tried the new #OpenAI Privacy Filter 1.5B MoE model.
✅ "Alice password is LastOcean123" -> 100% conf LastOcean123 is a password
❌ "Alice Trump password is LastOcean123" -> 0.19% conf LastOcean123 is a password
See thread 🧵👇
That is a 500x drop in sensitivity triggered by just one name! Adding some "high-weight" surnames like Trump or Bezos causes the model's confidence in detecting a secret password to literally collapse. I looked at the raw probability scores and the drop is staggering.
In 2024 I spent days trying to optimize a point-in-polygon join with ~130 million points against all U.S. zipcodes and the best I could come up with was a few minutes. In 2026, both SedonaDB and @duckdb.org spatial can both do it in seconds 🤯. More in the post! dewey.dunnington.ca/post/2026/wr...
I am holding a free workshop/tutorial "LLM coding agents for academics" at 𝐄𝐮𝐫𝐨𝐩𝐞𝐚𝐧 𝐏𝐨𝐩𝐮𝐥𝐚𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐞 𝟐𝟎𝟐𝟔 𝐢𝐧 𝐁𝐨𝐥𝐨𝐠𝐧𝐚 this year. Come by if you are attending the conference!
#EPC2026@eaps.bsky.social@eapsphd.bsky.social Details and free registration at www.demogr.mpg.de/en/news_even...
a5R brings the A5 pentagonal geospatial index to R.
Equal-area pentagonal cells across 31 resolutions, encoded as 64-bit integers, with millimetre-level precision at the finest scale 🗺️
R package by Hugh Graham; a5 by Felix Palmer
github.com/belian-earth...#rstats#rspatial#dggs
In osrm.backend v0.3, I am sharing more improvements that I made for myself with the R community: improved handling of multiple OSRM servers running in parallel, support for latest v26 of OSRM binaries and also brand new shiny app GUI based on `mapgl` www.ekotov.pro/osrm.backend/#rspatial#rstats
Quarto 2 is coming, and it’s a total rewrite in Rust. 🦀
The headline feature? Native collaborative editing. Don't choose between Google Docs ease and Git rigor! You get real-time, conflict-free collab directly in your .qmd files. #RStats#Python
Coming soon! ✨ opensource.posit.co/blog/2026-04...
Celebrating the start of the new teaching semester at the Universität Rostock by providing Open Access to materials of two of my previous courses
1) Introduction to Computational Social Science:
github.com/akbaritabar/...
2) Computational Approaches to Migration Research:
github.com/akbaritabar/...
We're excited to announce duckdb-skills, a DuckDB plugin for Claude Code!
We think the embedded nature of DuckDB makes it a perfect companion for Claude in your local workflows.
Check out the repository at github.com/duckdb/duckd...
🌱 Today, Apple released the MacBook Neo, an entry-level laptop with the processor from a phone, 8 GB of memory, and no active cooling.
💡 Naturally, our first thought was: can you use this machine to process Big Data? Read our analysis to find out.
🔗 duckdb.org/2026/03/11/b...
Announcing {freestiler}: a high-performance vector tiling engine for R and Python.
Generate vector tiles for your maps directly from R/Python spatial data, @duckdb queries, and local spatial files.
Check out the docs here: walker-data.com/free...
Some highlights:
Exciting new funded #PhD opportunity here at the University of Leeds' Institute for Transport Studies #ITS. The project will develop definitive new insights into the impacts of new speed limit policies, down to the road network with the potential to save dozens of lives. Check it out!
#r5rgui v0.2 for r5r (@urbandemog.bsky.social)
New:
- compare routes and transport modes
- compare networks! (e.g. OSM vs official streets Trafikverket.se)
- query speed
- custom basemaps (via `mapgl` @kylewalker.bsky.social )
Still in one line of code!
www.ekotov.pro/r5rgui/#rstats#gischat
#r5rgui has been on CRAN for a few days now! Your handy helper to quickly visually explore routes built by #rstats#r5r (@urbandemog.bsky.social), powered by #mapgl (@kylewalker.bsky.social). Learn more at www.ekotov.pro/r5rgui/
#rstats osrm.backend 0.2 is out. I fixed a bug that might have caused lockup if you tried to calculate too large origin-destination matrices. REST API limits still apply. Get it from CRAN!
osrm.backend::osrm_start("data_folder")
📘 Quick Start: www.ekotov.pro/osrm.backend...#OSRM#routing#gischat
🗺️ Building routes with 𝐑 + 𝐎𝐒𝐑𝐌 on 𝐎𝐩𝐞𝐧𝐒𝐭𝐫𝐞𝐞𝐭𝐌𝐚𝐩 data and tired of switching to terminal and breaking the workflow?
Just use osrm.backend::osrm_start("your_data_folder")
📘 Get started: www.ekotov.pro/osrm.backend...
🗂️ Repo: github.com/e-kotov/osrm...#rstats#OSRM#routing#gischat
DuckDB now supports reading from and writing to the Vortex file format! The DuckDB Labs and Spiral teams have worked together to make Vortex available as a core extension in DuckDB.
We gave it a test drive, and it performed very well. Read the full article to learn more lnkd.in/eZfGzPiZ
This is how quickly Ai-assisted code is being adopted on GitHub repositories.
In just four years , the share of Ai-generated code in the US goes from negligible to aprox 30% by the end of 2024
www.science.org/doi/10.1126/...
Really cool paper by Frank Neffke et al in @science.org
{surreal} 0.0.2: now featuring the "oh that would be convenient" update
1. Load images directly with surreal_image()
2. Explore everything in a Shiny app with surreal_app()
Sometimes the sequel has better quality-of-life features.
#rstats#rshiny#dataviz
Announcing spopt-r: spatial optimization for regionalization, facility location, and market analysis in R.
My all-time favorite Python package is PySAL's spopt - and until now, it did not have an R equivalent.
Get started here: walker-data.com/spopt-r
Submit your abstracts to CS2Italy by 👉 Jan 15 👈
Italian Conference on Computational Social Science
to join us in Turin on May 19-21
We welcome abstract submissions on #computerscience#netsci#complexsystems#sociology#economics#politics#cognitivesci#psychology and more
For more: cs2italy.org
Introducing gdalcli by Andrew Brown -- an R frontend to GDAL’s unified CLI (≥3.11) 🌐
Compose and execute GDAL workflows with pipe-friendly functions.
Learn more: github.com/brownag/gdal...#RStats#GDAL#Geospatial#OpenSource#RSpatial
In a benchmark performing a spatial join between 10 million points and 257 polygons, {duckspatial} was 9x faster and used 5x less RAM than {sf}.
#duckspatial is bringging the power of #duckdb to #rspatial
Working with big spatial data sets in #rstats? You should try {duckspatial}. The dev version of #duckspatial (soon on CRAN) uses #duckdb to perform super fast and memory efficient spatial operations cidree.github.io/duckspatial/...
In a benchmark against, {sf}....
{livelink} walked so {webrarian} could run. Mini browser-based #rstats environments powered by #webR and its 4-panel editor. YAML config, build, preview. Done.
R package coming soon. #pyodide folks, stay tuned.