iiinova — where your digital self keeps evolving

Bruno Rodrigues @brodriguesco.bsky.social · 08/10/2026

that looks honestly pretty neat! I'll have to try with some actual data!

110

Bruno Rodrigues @brodriguesco.bsky.social · 08/10/2026

Sim, falo português (mas já falei melhor, só falo com os meus pais 😅). Manda-me o link do teu projeto, se puderes.

110

Bruno Rodrigues @brodriguesco.bsky.social · 08/10/2026

I was recently recommended a talk about the concept of “pits of success”, the idea that good systems make the right path the easiest path. I didn’t know about this concept, but it turns out I’ve been designing T around many of the same principles. New blog post: brodrigues.co/posts/posts/...

130

Bruno Rodrigues @brodriguesco.bsky.social · 04/10/2026

T is designed to be generated and maintained by AI agents rather than manually programmed by humans. The idea is simple: make the structure of reproducible statistical analysis explicit. You can compose these structures however you want, but you can't opt out of them. tstats-project.org 3/3

tstats-project.org

T Programming Language - index

010

Bruno Rodrigues @brodriguesco.bsky.social · 04/10/2026

Nodes run R, Python or Julia code seamlessly and input/output is handled automatically. Basically, I'm tired of asking "Please remember to set up your projects in a reproducible way" With T: "This isn't a pipeline, therefore it isn't a T program and won't run". 2/3

tstats-project.org

T Programming Language - index

110

Bruno Rodrigues @brodriguesco.bsky.social · 04/10/2026

T is a programming language for reproducible data analysis in which software-engineering best practices are enforced by construction. T projects are reproducible Nix environments, T programs are pipelines, and computational nodes are explicit, isolated and cacheable. 1/3

tstats-project.org

T Programming Language - index

151

Bruno Rodrigues @brodriguesco.bsky.social · 27/09/2026

T 0.55.3 is out 🚀 The latest release of the T programming language brings new data verbs, smarter tooling, and dozens of fixes that make reproducible polyglot data science pipelines more robust! tstats-project.org

tstats-project.org

T Programming Language - index

0111

Bruno Rodrigues @brodriguesco.bsky.social · 25/09/2026

Thankfully, it is AI now that will have to deal with nested cells that are also conditionally formatted and merged.

000

Reposted by Bruno Rodrigues

UGent Open Science @ugentopenscience.bsky.social · 08/09/2026

Want your R analyses to run the same way across computers, OS, and time? Join our webinar on reproducible computing in R with {rix}, a package that makes the power of Nix accessible to R users. tinyurl.com/3555uwza With @felipefv.bsky.social, co-author of a {rix} tutorial #rstats #Reproducibility

tinyurl.com

Microsoft Virtual Events Powered by Teams

0106

Reposted by Bruno Rodrigues

Felipe Fontana Vieira @felipefv.bsky.social · 11/09/2026

Next week I will give a free webinar about how I have been using Nix (by itself) but mainly through {rix} for my current articles. If anyone is intested, my plan for the structure of the webinar (in the comments):

11913

Bruno Rodrigues @brodriguesco.bsky.social · 02/09/2026

I know it's easier said than done, but ignore the haters. These are miserable people who can only dream to make a hundredth of the impact you made.

050

Bruno Rodrigues @brodriguesco.bsky.social · 01/09/2026

of course it rocks 🤣🤣🤣 (I'm not modest)

020

Reposted by Bruno Rodrigues

Isabella Velásquez @ivelasq3.bsky.social · 31/08/2026

Tired of "it worked on my computer" breaking your #RStats projects? Tomorrow on the Data Science Lab, @rpodcast.bsky.social is going live to demo {rix}, an @ropensci.org package that uses Nix to build reproducible R environments. I caught a sneak peek and it's WOW! Join us! https:// pos.it/dslab

34411

Bruno Rodrigues @brodriguesco.bsky.social · 31/08/2026

Pretty unreal to see people using and enjoying rix 😂

1100

Bruno Rodrigues @brodriguesco.bsky.social · 29/08/2026

The cost of not running capable agents continuously is already very high and compounding.

010

Reposted by Bruno Rodrigues

Bruno Rodrigues @brodriguesco.bsky.social · 21/08/2026

Friends, if you want to support my open source contributions, you can now buy a WIP version of my upcoming book, "Reproducible Polyglot Data Science A Unified Approach with Nix and T" on leanpub! #R #Python #Julia #nix #DataScience leanpub.com/reproducible...

162

Bruno Rodrigues @brodriguesco.bsky.social · 21/08/2026

162

Bruno Rodrigues @brodriguesco.bsky.social · 19/08/2026

NixOS is not "just another linux distro" it’s a thing of its own, which is its greatest strength but also why getting into it is so hard. But thanks to LLMs, it’s never been easier! 2/2

000

Bruno Rodrigues @brodriguesco.bsky.social · 19/08/2026

Some would certainly disagree with this take, but NixOS is (in my opinion) not even a linux distribution. Sure it uses the Linux kernel, but other than that, it’s so different that I’d argue the jump from Ubuntu to NixOS is wider than the jump from Windows to Ubuntu. 1/2

120

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

I’d bet actual money this was done by a human, not by an LLM

000

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

All of this is now fixed in 0.55.0, alongside a long list of correctness improvements to joins, `drop_na`, `distinct`, `get`, pipeline dependency inference, and more. Also `t fix`, which can now mechanically apply some of the suggested fixes from `t check`'s diagnostics. 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

None of these were caught by hand-written unit tests. They surfaced because property-based testing explores the input space far more thoroughly than a human ever will: and then hands you back a minimal, reproducible failing case instead of a wall of noise. 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

🔍 Case mapping was ASCII-only, silently failing on accented and special characters 🔍 Nested verb calls inside NSE verbs (e.g. `mutate(select(df, $x), .)`) were being misparsed and rejected 🔍 Factor-grouped aggregations could group by the wrong levels entirely when level order wasn't the default 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

🔍 String and regex functions were operating on raw bytes instead of Unicode code points: multi-byte characters could get split, mis-indexed, or corrupted in `str_substring`, `index_of`, `str_split`, and every PCRE2-backed regex function. 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

I then turned `propcraft` on T's own standard library (dogfooding, as it should be), generating thousands of randomized Ints, Floats, Strings, DataFrames, Dicts, dates, and factors, and it surfaced real, previously-hidden bugs: 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

`propcraft` introduces property-based testing. Instead of writing tests one case at a time, state an invariant: "serializing and deserializing a value always returns the same value" and it gets checked against 100s of generated inputs. 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

110

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

As controversial as this might sound to some, I believe that AI-only programming is here to stay. So we need to provide tools that allow LLMs to self-correct as they program. This release (and the previous 3) of T were focusing on just that. This release introduces a new package, `propcraft`. 👇

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

110

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

🚀 T v0.55.0 is out, and this release is a big one for correctness! T does not replace Python or R, but instead, offers a reproducibility harness around them!

230

Bruno Rodrigues @brodriguesco.bsky.social · 12/08/2026

Hi, maybe you’d like to take a look at another of my project, called T (tstats-project.org)

tstats-project.org

T Programming Language - index

010

Bruno Rodrigues @brodriguesco.bsky.social · 04/08/2026

Both these use-cases should be covered by T! Thanks for testing, looking forward to your impressions :)

010

Bruno Rodrigues @brodriguesco.bsky.social · 03/08/2026

New blog post: The Future of Programming Looks Like Its Past or Why I build T brodrigues.co/posts/2026-0...

brodrigues.co

The Future of Programming Looks Like Its Past or Why I build T – Econometrics and Free Software

000

Reposted by Bruno Rodrigues

Bruno Rodrigues @brodriguesco.bsky.social · 01/08/2026

Willing to help me test T, a new domain-specific language for building reproducible pipelines (in #RStats, #Python or #Julia)? I'd like to see how easy it is for people to use agents with T to build reproducible pipelines. If you're interested in trying it out, please follow these instructions.

tstats-project.org

T Programming Language - nix-installation

243

Bruno Rodrigues @brodriguesco.bsky.social · 01/08/2026

Any feedback, whether from the agent's experience or your own, is greatly appreciated. Github repo: github.com/b-rodrigues/...

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

000

Bruno Rodrigues @brodriguesco.bsky.social · 01/08/2026

Can your AI agent successfully build the pipeline using only the onboarding md files included in the project? Is the docs clear enough for humans to follow without additional guidance? If you run into any confusing or missing docs, please let me know where you got stuck or what could be improved.

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 01/08/2026

In short, the steps are: Install Nix. Install T. Create a new pipeline project with:t init --project my-demo-project Add some R or Python code that an AI agent could use as the basis for building a pipeline to the projet template's src/ folder. What I'm most interested in learning is:

github.com

GitHub - b-rodrigues/tlang

Contribute to b-rodrigues/tlang development by creating an account on GitHub.

100

Bruno Rodrigues @brodriguesco.bsky.social · 01/08/2026

First, you'll need to have Nix installed. If you don't already have it and are willing to install it for this test (it's easy to uninstall afterward), follow the installation guide here: tstats-project.org/nix-installa... Then continue with: tstats-project.org/getting-star...

tstats-project.org

T Programming Language - nix-installation

100

Bruno Rodrigues @brodriguesco.bsky.social · 01/08/2026

tstats-project.org

T Programming Language - nix-installation

243

Bruno Rodrigues @brodriguesco.bsky.social · 20/07/2026

🚀 Version 0.54.1 is here! It introduces a host of new features that make it even easier for AI agents to build pipelines, so you can focus on your R, Python, or Julia analysis code! tstats-project.org

tstats-project.org

T Programming Language - index

050

Reposted by Bruno Rodrigues

Bruno Rodrigues @brodriguesco.bsky.social · 09/07/2026

🚀 T v0.54.0 is out! 🚀 Maximum reproducibility for data science pipelines: 📦 Auto-resolves R packages from renv.lock 🐍 Support for Python uv.lock via uv2nix virtual envs #Nix #Python #RStats #DataScience 1/2

1102

Bruno Rodrigues @brodriguesco.bsky.social · 09/07/2026

time to replay Doom for the nth time

000

Reposted by Bruno Rodrigues

coolbutuseless @coolbutuseless.bsky.social · 08/07/2026

Introducing {rdoom} - Play Doom in #RStats (in time for #UseR2026 !) github.com/coolbutusele... This is a demonstration that {tigerfb} + {nara} can do realtime interactivity in R! Includes: keyboard & mouse control, audio, cheat modes!

410132

Bruno Rodrigues @brodriguesco.bsky.social · 09/07/2026

❄️ Strict per-node Nix flake environment isolation 🔗 Reproducible hash-pinned fetchurl() to download assets 2/2 tstats-project.org

000

Bruno Rodrigues @brodriguesco.bsky.social · 09/07/2026

1102

Bruno Rodrigues @brodriguesco.bsky.social · 04/07/2026

In return, Luxembourg gave America René Arend, the McDonald's executive chef who developed the Chicken McNuggets (introduced in the early 1980s) and the McRib. Happy 4th of July to my American friends!

000