Bruno Rodrigues

I expect that #Nix will become more popular now that LLMs are being used so much for programming. Nix makes working with LLMs safer by making the entire environment declarative. Instead of guessing what tools exist on a system, the LLM can read a single Nix expression that precisely defines dependencies, binaries, and versions. If something’s missing, ask the LLM to add it to the flake. In contrast, letting the LLM running imperative commands that change the system is insane.

Reproducibility tools tell you how to be careful. A reproducibility-by-design language makes carelessness structurally hard, just the way a speed bump makes you slow down, not a sign asking nicely. Immutable bindings, mandatory pipeline DAGs, sandboxed nodes, pinned Nix environments: T is my attempt to build that language for #DataScience

https://tstats-project.org/

RE: @rpodcast@podcastindex.social

and the video is up now: https://www.youtube.com/watch?v=wkjLVhaazRI

🚀 T-Lang v0.52.0 "Kaméhaméha" just released! 💥

First-class, Nix-reproducible Julia support!

⚡ `jln()` nodes with automatic environment & package provisioning
🧠 Native PMML & ONNX model scoring inside Julia nodes
🌀 Sandbox-safe world-age resilience (via `Base.invokelatest`)
📦 New lightweight `tlang` companion packages for R, Python, and Julia

Orchestrate the big 3 data science languages seamlessly using T!

Learn more: https://tstats-project.org/

#JuliaLang enjoyers, what's the most commonly used plotting library?

RE: @brodriguesco@fosstodon.org

The next version of T will add support for #Julia

NixOS is technically a Linux distribution but it's different enough that it could be its own, separate operating system

RE: @brodriguesco@fosstodon.org

Treat Agent Output Like Compiler Output

https://skiplabs.io/blog/codegen_as_compiler

I want to talk to scientists that use Python for statistics, do I know anyone that organises talks at their labs or institutions where I could give a talk ?

https://brodrigues.co/talks.html

🛠️ And More:

• Auto-Quoting: Even cleaner data-wrangling syntax.
• Structural Integrity: A new StructuralError category for ironclad DAG validation.
• Official Tree-sitter Support: Beautiful, polyglot syntax highlighting for Neovim, Emacs, and VS Code.

T is still in Beta, testers more than welcome!

https://tstats-project.org/

#TLang #DataScience #Reproducibility #Nix #RStats #Python #OpenSource

📊 First-Class Visual Metadata:

• Deep Inspection: T now automatically "looks inside" your R (ggplot2) and Python (matplotlib, plotly, altair, seaborn, plotnine) plots.
• Transparent Results: read_node() on a plot node now returns a structured dictionary of titles, labels, and layers instead of an opaque binary artifact.
• show_plot(): Render and view plots from any language natively from the T REPL, handled safely and headlessly within the Nix sandbox.

🚀 T 0.51.4 is officially out! 🚀

The polyglot orchestration engine for reproducible data science just took a massive leap forward in both visualization and AI collaboration.

https://tstats-project.org/

What's New in 0.51.4?

🤖 LLM Onboarding: T is now the first language to treat LLMs as first-class collaborators. `t init` now interactively scaffolds tiered technical references (`small` to `huge`) and project-specific `AGENTS.md` guides.

https://tstats-project.org/index.html

https://tstats-project.org/