iiinova — where your digital self keeps evolving

Dirk Eddelbuettel @eddelbuettel.com · 6h

RcppFastFloat 0.0.6 on CRAN: New Upstream R Acccess to Number Parsing at a Gigabyte per Second dirk.eddelbuettel.com/blog/2026/09... #rstats #rcpp

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/30#rcppfastfloat_0.0.6 describing release 0.0.6 of CRAN package RcppFastFloat

010

Dirk Eddelbuettel @eddelbuettel.com · 12h

myman 0.1.0 on CRAN: New (fabulous!!) package Fortunes-style collection of "My man ..." posts by Kevin Kruse dirk.eddelbuettel.com/blog/2026/09... #rstats /cc @kevinmkruse.bsky.social

Screenshot one of two of blog post at https://dirk.eddelbuettel.com/blog/2026/09/30#myman_0.1.0 describing (initial) CRAN release 0.1.0 of package myman

Screenshot two of two of blog post at https://dirk.eddelbuettel.com/blog/2026/09/30#myman_0.1.0 describing (initial) CRAN release 0.1.0 of package myman

000

Reposted by Dirk Eddelbuettel

Peter Dalgaard @pdalgd.bsky.social · 29/09/2026

#rstats 4.6.2 "Diversified Portfolio" scheduled for October 29. Full schedule on developer.r-project.org (or the svn if you're impatient.)

0279

Dirk Eddelbuettel @eddelbuettel.com · 29/09/2026

"Why we cannot have nice things." Plus: Steive Wonder graces us four unreleased songs. They are amazing. But I already said that it is Stevie. Minus: YouTube disgraces it by placing an ad at each of the three transitions. Awful.

011

Dirk Eddelbuettel @eddelbuettel.com · 29/09/2026

Life's too short. Blocking poster.

000

Dirk Eddelbuettel @eddelbuettel.com · 29/09/2026

You also completely failed to engage _with my key point in this thread_ that using _local models_ is *obviously* preferable if you are worried about any of this.

000

Dirk Eddelbuettel @eddelbuettel.com · 29/09/2026

You missed they point where I stated that posit ($deity bless their heart) are actually *not* the one providing the answer. So whether they hold on to intermediate data is ... immaterial.

200

Dirk Eddelbuettel @eddelbuettel.com · 29/09/2026

Of course. But does it matter? They are not the ones processing, they cover the UI integration but otherwise 'pass on' on to the actual services. Their TOS may matter more ... and hence give yet another reason to prefer local models.

100

Dirk Eddelbuettel @eddelbuettel.com · 28/09/2026

From somewhat vague memory, here is what happened IIRD: 1) CRAN expels rbenchmark. 2) Packages that depend on it also get expelled. 3) I offered to adopt rbenchmark and then did. 4) Expelled packages have to come back one by one on their own; CRAN will not 'bulk re-import' them.

120

Dirk Eddelbuettel @eddelbuettel.com · 28/09/2026

The key is that this is a well-understood interface offered by the 'engines' (such as Ollama or LM Studio or ....). Posit Assistant is but one client, it happens to be inside the IDE. I happen to use OpenCode (in the terminal) and / or pi much more for it.

100

Dirk Eddelbuettel @eddelbuettel.com · 28/09/2026

Both LM Studio and Ollama run / can run as a service that gets accessed via a localhost url and port (it's actually an OpenAI compatible setup). The IDEs (just like any other service: pi, openclaw, ...) connect to that under their settings. For positron you may have to edit a JSON file.

Screenshot of AI Assistant setup for RStudio pointing to local Ollama model.

100

Dirk Eddelbuettel @eddelbuettel.com · 28/09/2026

Local models. Nothing leaves your machine. LM Studio or Ollama run them, also as a service that e.g. RStudio or positron can hook in. Open weights such as qwen3.8 are pretty good.

130

Dirk Eddelbuettel @eddelbuettel.com · 27/09/2026

r^4 post #060: Using bubblewrap for R sandboxing Quick note on agentic R use via bubblewrap dirk.eddelbuettel.com/blog/2026/09... #rstats

Screenshot of r^4 blog post #060 at https://dirk.eddelbuettel.com/blog/2026/09/27#060_bubblewrap_for_R discussing R use inside the 'bubblewrap' harness helper

093

Reposted by Dirk Eddelbuettel

Hoodlum 🇺🇸 @nothoodlum.bsky.social · 27/09/2026

Denmark is trolling Trump.

29269222724

Dirk Eddelbuettel @eddelbuettel.com · 26/09/2026

What awful news! Hope you are getting all the care you need, and are going to make a full recovery.

010

Reposted by Dirk Eddelbuettel

Kim A. Wagner @kimatiwagner.bsky.social · 25/09/2026

Having just started teaching new undergraduates at uni, I think we're missing a key issue in the debate about AI: these kids do not know how *not* to use AI, full stop. We assume they are 'digital natives' but they are not and a Google Scholar search is as alien to them as an archival card-index.

414930

Dirk Eddelbuettel @eddelbuettel.com · 25/09/2026

Sitting here in #Chicago and, as one does, listening to Donnie Hathaway live. So many gone so soon. open.spotify.com/album/7HJQ47...

open.spotify.com

Live At The Bitter End 1971 - Album by Donny Hathaway | Spotify

Donny Hathaway · album · 2014 · 10 songs

000

Dirk Eddelbuettel @eddelbuettel.com · 24/09/2026

RcppFastAD 0.0.5 on CRAN: Maintenance R interface to automatic differentiation C++ library dirk.eddelbuettel.com/blog/2026/09... #rcpp #rstats

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/24#rcppfastad_0.0.5 describing release 0.0.5 of CRAN package RcppFastAD

052

Reposted by Dirk Eddelbuettel

Shannon Vallor @shannonvallor.bsky.social · 24/09/2026

It is a profound indictment of most media coverage of Big Tech that your best bet for getting an informed understanding of AI and its dangers is following a decently curated list of Bluesky shitposters

1030686

Reposted by Dirk Eddelbuettel

apenwarr @apenwarr.ca · 24/09/2026

Robots can only commit manslaughter, not first or second degree murder, because murder requires intent. In this essay I will

3855

Reposted by Dirk Eddelbuettel

Chris Hayes @chrislhayes.bsky.social · 23/09/2026

Someone needs to explain to Trump that the worse Republicans do in the midterms, the better it looks for him. Just goes to show the party is utterly lost without him on the ballot! He’s literally the only Republican who has the juice. (Natalie, if you see this, please print and share.)

702517278

Reposted by Dirk Eddelbuettel

The R Foundation @r-foundation.bsky.social · 22/09/2026

New blog post by @ellakaye.co.uk and @heathrturnr.fosstodon.org.ap.brid.gy: Zulip Chat for the R Project #RStats blog.r-project.org/2026/09/22/z...

blog.r-project.org

Zulip Chat for the R Project - The R Blog

1105

Reposted by Dirk Eddelbuettel

troy_phd @troyhernandez.bsky.social · 22/09/2026

Update: corteza scored 100% on ARC-AGI-3's public set with GPT-6 Astra! And due to how they calculate their metrics, I'm happy to report that our 100% is the best 100%. I made a dashboard with details: cornball.ai/arc-agi-3/ #rstats

cornball.ai

ARC-AGI-3 comparison

032

Reposted by Dirk Eddelbuettel

Daniel Lemire @lemire.bsky.social · 22/09/2026

A summer of AI optimization lemire.me/blog/2026/09/22/a-summer-…

061

Dirk Eddelbuettel @eddelbuettel.com · 22/09/2026

Interesting. "TIL" about Eddington number. It so happens that someone already made an app / page: swinny.net/Strava/ So apparently mine is 41.

swinny.net

Free-to-use Strava Tools

010

Reposted by Dirk Eddelbuettel

R Consortium @rconsortium.bsky.social · 21/09/2026

What does it take to sustain R? Join R Consortium, R Core & R Foundation Oct 6 to discuss the people, infrastructure & community behind R. 12 PM PT / 3 PM ET / 8 PM London Register: r-consortium.org/webinars/sus... #RStats

Graphic for upcoming R Core and R Foundation panel discussion - Oct 6, 2026

02516

Dirk Eddelbuettel @eddelbuettel.com · 21/09/2026

Went cycling in Door County, WI, this weekend and am now a centenarian. 🚴‍♂️ Great day, most lovely scenerary, but we are now 0 for 2 on sunny days at the Peninsula ride following the spring ride last year.

Screenshot of Strava entry for the Peninsula Classic Fall Century showing total distance (101.8 miles, yay), elevation, time (6h 46min riding time but about 8h total), a map and a photo of four of us at the start.

191

Dirk Eddelbuettel @eddelbuettel.com · 20/09/2026

Twenty months down. Leaves 853 days. One day at a time.

R code and output for the number of days until Jan 20, 2029.

> difftime(as.Date("2029-01-20"), Sys.Date())
Time difference of 853 days
>

170

Reposted by Dirk Eddelbuettel

The R Foundation @r-foundation.bsky.social · 18/09/2026

New blog post by Vincent Goulet & Kurt Hornik: Nifty new features for tangling and weaving Sweave documents in base #rstats blog.r-project.org/2026/09/18/n...

blog.r-project.org

New Tricks for the R Sweave Drivers - The R Blog

043

Dirk Eddelbuettel @eddelbuettel.com · 17/09/2026

You can take the boy out of [former grad school home] Marseilles but you cannot take Marseilles out of the boy. 'pizzaladière' as they call it: onions, anchovies, black olives. I had some sourdough from bread making. Turned out lovely.

Photo of an uncooked 'pizzaladière' Marseilles style pizza with onions, anchovies, and black olives. On sourdough as I had extra.

Photo of a now cooked 'pizzaladière' Marseilles style pizza with onions, anchovies, and black olives. On sourdough as I had extra. Delicious.

120

Dirk Eddelbuettel @eddelbuettel.com · 16/09/2026

There is a tell if you know where to look: Packaged: 2026-09-16 07:49:42 UTC; ripley I presume we could search at the GitHub cran for this with a regular expression ...

010

Dirk Eddelbuettel @eddelbuettel.com · 16/09/2026

Another thing worth mentioning: sometimes the CRAN team updates a package and makes a bugfix release with ".1" added to the version. We had another example of that today. Depends, I suppose, on the 'centrality' of the package and its reverse dependencies.

110

Dirk Eddelbuettel @eddelbuettel.com · 16/09/2026

The top-line iPhone Duo has two terabyte storage. A phone. 2 tb.

100

Reposted by Dirk Eddelbuettel

troy_phd @troyhernandez.bsky.social · 14/09/2026

ARC's analysis of GPT-6 Astra: 62.7% with their standard harness, 98.55% at max effort with their provider adapter harness (99.95% at high). corteza, our #rstats harness, at max effort with Claude Opus 5: 98.58%. Same 25 games. cornball.ai/posts/cortez...

cornball.ai

98.6% on ARC-AGI-3 in R

corteza, an R-native RLM harness, scored 98.58% on ARC-AGI-3. Across 25 games the model wrote 894 R helpers and kept calling them, 53 of them from before its context was compacted.

093

Dirk Eddelbuettel @eddelbuettel.com · 13/09/2026

td 0.0.7 on CRAN: New Features and Updates R access to TwelveData Financial Data API dirk.eddelbuettel.com/blog/2026/09... #rstats

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/13#td_0.0.7 describing release 0.0.7 of CRAN package td

010

Dirk Eddelbuettel @eddelbuettel.com · 13/09/2026

Nice day for a 606-to-lakefront ride. Two selfies from, respectively, the north end of my lake exposure at Fullerton and then at southshore / 63rd.

Screenshot from Strava record for 45 mile ride from Oak Park via 606 to Lakefront, down to South Shore park and back.

010

Reposted by Dirk Eddelbuettel

Costa Samaras @costasamaras.com · 13/09/2026

This will be bad for: - the students - the research progress - the universities - the towns the universities are in - the companies hiring students - the states the universities are in - the national economy - national security - the country - the world

23934305

Dirk Eddelbuettel @eddelbuettel.com · 12/09/2026

sanitizers 0.1.2 on CRAN: Maintenance True positives for ASAN/UBSAN/... tests dirk.eddelbuettel.com/blog/2026/09... #rstats

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/12#sanitizers_0.1.2 describing release 0.1.2 of CRAN package sanitizers

010

Dirk Eddelbuettel @eddelbuettel.com · 12/09/2026

The non-list email is a pain (we started Rcpp with a list email) but all information we are talking is public a) at the package's CRAN page and mirrors and b) in aggregations as eg returned by tools::CRAN_package_db() so trivial to automate in, say, a GHA which then emails, or pokes slack, or ...

110

Dirk Eddelbuettel @eddelbuettel.com · 12/09/2026

... too long to migrate from unstable. (Which can hurt maintainers of 'large' packages: my r-base (for R itself) is often victim of one of its 1000+ reverse dependencies which may fail a test on random obscure (non-CRAN) architecture. But let's stick to CRAN here, Debian is another matter.

000

Dirk Eddelbuettel @eddelbuettel.com · 12/09/2026

Can I assume you are aware that CRAN a) always emails and gives a (public) three week window to fix (which is how I knew about these) and b) often extends this by three weeks more and c) is also amenable to discussion if "reasons". Moreover, Debian removes from testing when a package takes too ...

210

Dirk Eddelbuettel @eddelbuettel.com · 12/09/2026

And ASAN checks, painful as they are, most often reveal genuine bugs. So checking those is good too. We have the same issues as Debian maintainers (when, say, obscure hardware breaks) and yes a common way out can be to ... disable a test conditionally. Been there, done that. Also at CRAN.

120

Dirk Eddelbuettel @eddelbuettel.com · 12/09/2026

Complicated indeed! Consider URL checks: it is mucg better to ensure that documentation is reachable. Yet sometimes even an archive.org redirect times out. Or a primary site times out (as they defend against DDOS and 'count hits / ip'). All this imposes 'work' on us, but raises overall quality.

110

Dirk Eddelbuettel @eddelbuettel.com · 10/09/2026

RDieHarder 0.2.8 on CRAN: Minor Maintenance DieHarder RNG testing for R dirk.eddelbuettel.com/blog/2026/09... #rstats

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/10#rdieharder_0.2.8 describing release 0.2.8 of CRAN package RDieharder

010

Dirk Eddelbuettel @eddelbuettel.com · 10/09/2026

Try Norm Matloff's "Art of R Programming" as a good and inexpensive book. There are a few more resources at my course site (also covering adjacent topics): stat447.com/resources/re...

stat447.com

Reading Material - STAT447: Data Science Programming Methods

260

Reposted by Dirk Eddelbuettel

The Onion @theonion.com · 09/09/2026

Experts Advise Against Throwing Laptop Across Office Even Though It Will Feel Incredible theonion.com/experts-advise-against…

271699253

Dirk Eddelbuettel @eddelbuettel.com · 10/09/2026

I am mostly still salty about the forced retirement / abandonment of the first-generation Nest "just because of Google Home integration". Arguably nicer now -- but "my device, my choice when I retire it". Says local area man playing music off the first generation sonos set ...

010

Dirk Eddelbuettel @eddelbuettel.com · 09/09/2026

RcppXts 0.0.7 on CRAN: Minor Maintenance Rcpp-based C-level Interface to the xts package dirk.eddelbuettel.com/blog/2026/09... #rcpp #rstats

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/09#rcppxts_0.0.7 describing release 0.0.7 of CRAN package RcppXts

040

Dirk Eddelbuettel @eddelbuettel.com · 08/09/2026

RcppArmadillo 15.6.0-1 on CRAN: New Upstream Minor R bindings to powerful and expressive C++ matrix library dirk.eddelbuettel.com/blog/2026/09... #rstats #rcpp

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2026/09/08#rcpparmadillo_15.6.0-1 describing release 15.6.0-1 of CRAN package RcppArmadillo

020