Sign in

Alberto Torres Barrán

@albertotb.com
61 followers 526 following 25 posts

Machine Learning at komorebi.ai

PostsRepliesMedia
Alberto Torres Barrán @albertotb.com · 27/09/2026
Wow, I always used match and I think I didn't know this!
010
Reposted by Alberto Torres Barrán
Python Hub @pythonhub.dev · 27/09/2026
Soft-deprecating re.match() The article explains when to use re.prefixmatch(), re.search(), and re.fullmatch(), what soft deprecation means for existing code, and how to flag re.match() usage with Ruff. hugovk.dev/blog/2026/soft-deprecati…
121
Alberto Torres Barrán @albertotb.com · 02/09/2026
This looks really cool, I need to find an excuse to try it
020
Reposted by Alberto Torres Barrán
Sung Kim @sungkim.bsky.social · 31/08/2026
Google's TimesFM-3, a state-of-the-art time series foundation model that enables accurate multivariate time series forecasting in a single forward pass, significantly outperforming other forecasting models across major benchmarks. Blog: goo.gle/4x5WGpD
0486
Alberto Torres Barrán @albertotb.com · 24/08/2026
100% this. I start adding a small thing and end up with 5 new issues and 5 new PRs
020
Reposted by Alberto Torres Barrán
GitHub @github.com · 11/08/2026
A dozen Dependabot pull requests on a Monday morning is how important updates get ignored. On Microsoft's GCToolkit, roughly one in six commits were single-dependency version bumps. Three small changes to dependabot.yml fixed it 👇
github.blog
Tame Dependabot: Group your updates, slow the cadence, keep security fast
Here's how grouping updates, slowing the cadence, and keeping security fixes fast cut the noise on a Microsoft open source project.
1142
Reposted by Alberto Torres Barrán
Brett Cannon @snarky.ca · 10/08/2026
I self-nominated myself for the Python packaging council today. I have posted my nomination statement as a blog post. snarky.ca/my-nominatio...
snarky.ca
My nomination statement for the 2026 Python packaging council
I have decided to run for the inaugural/2026 Python packaging council (PPC). I will say I have the support of my employer (Microsoft) to do this, but they didn't ask me to and my usual thing that I wo...
3196
Reposted by Alberto Torres Barrán
Maike Osborne @maosbot.bsky.social · 30/07/2026
When the AIs take over, I hope they remember all the times I said please and thank you to Claude
2203
Reposted by Alberto Torres Barrán
Julia Evans @b0rk.jvns.ca · 17/07/2026
learning a few things about running SQLite jvns.ca/blog/2026/07...
2766
Reposted by Alberto Torres Barrán
Sebastián Ramírez @tiangolo.com · 20/06/2026
library-skills now supports uv workspaces and npm/Bun workspaces. ✨ Auto-repair and sync of skills. Improved support for .claude (additional to .agents), it will ask you interactively what to use. Plus several other bug fixes and improvements. library-skills.io
093
Reposted by Alberto Torres Barrán
FastAPI @fastapi.tiangolo.com · 20/06/2026
FastAPI can now serve your frontend app ✨ With support for client-side routing 😎 Great for React with TanStack Router, Astro static builds, Vite-based apps, etc. 🎉 FastAPI version 0.138.0 🔖 fastapi.tiangolo.com/tutorial/fro...
2172
Reposted by Alberto Torres Barrán
Julia Evans @b0rk.jvns.ca · 16/06/2026
the next version of Python is going to have a built in sampling profiler!!!! this is so exciting to see docs.python.org/3.15/library...
28311
Reposted by Alberto Torres Barrán
Jakub Nowosad @jakubnowosad.com · 14/06/2026
🌍 New paper on key challenges in spatial machine learning! We discuss six major topics shaping the field: model validation, performance metrics, uncertainty quantification, algorithmic diversity, software implementation, and modeling protocols. www.erdkunde.uni-bonn.de/article/view... #SpatialML
0156
Reposted by Alberto Torres Barrán
Ted Underwood @tedunderwood.com · 12/06/2026
So, naive question about these campuses that pay OpenAI $13mil/yr. Why don’t they pour that money into a cluster running the strongest available open-weight model, with API and interactive options. And if the answer is “overhead,” why aren’t we collaborating?
5549872
Alberto Torres Barrán @albertotb.com · 01/06/2026
But they don't have the same number of keys no? I guess you can get used to it, really hard for me now though
100
Alberto Torres Barrán @albertotb.com · 01/06/2026
Was it English layout? There are some many options for US layout but almost none for other layouts. For Spanish specifically I always end up with keycron. Does low free offer ISO variants?
100
Reposted by Alberto Torres Barrán
Bite Code! @bitecode.dev · 29/05/2026
The conker bad fur day vibes with actual good graphics and amazing animation paired with earthworm jim wonkiness and cup head challenge + aesthetic is such a killer combo.
031
Reposted by Alberto Torres Barrán
Richard Iannone @richmeister.bsky.social · 21/05/2026
YAML deserves a defense! opensource.posit.co/blog/2026-05... The complaints are real but historical as they describe YAML 1.1, not the 1.2 spec that fixed them in 2009. The post traces the full history and introduces py-yaml12, a fast Rust-backed Python library for the modern spec. Hope you enjoy!
opensource.posit.co
In Defense of YAML
A common refrain among developers: YAML is bad and TOML is good. This post argues otherwise, tracing the history of configuration formats, examining what YAML 1.2 actually fixed, and introducing py-ya...
2235
Alberto Torres Barrán @albertotb.com · 15/05/2026
At least the resulting citations exist
0100
Reposted by Alberto Torres Barrán
⟨ Jargonaut NPC Alex ⟩ @lastnpcalex.agency · 15/05/2026
if you were to tell me better outcomes occurred with more compute time that's not particularly strange
261
Alberto Torres Barrán @albertotb.com · 15/05/2026
Ok, since when LLMs and Claude Code/Codex is the same thing? Scaling laws started with the size of the model but now sudenly is tokens spent. An LLM + harness is much more than the single model.
010
Reposted by Alberto Torres Barrán
Mapas Milhaud @mapasmilhaud.com · 10/05/2026
Segunda lengua extranjera más estudiada en la Unión Europea (2016): mapasmilhaud.com/mapas-de-datos/seg… #Mapas #Maps
0185-Lenguas-extranjeras-mas-estudiadas-en-la-Union-Europea-2016
082
Reposted by Alberto Torres Barrán
Hugo van Kemenade @hugovk.dev · 05/05/2026
prek now supports cooldown in either user or project config: github.com/j178/prek/re... pre-commit closed and locked the feature request as one of many many duplicates, but I can't find any: github.com/pre-commit/p... #PreCommit #prek #cooldown
github.com
Release 0.3.12 · j178/prek
Release Notes Released on 2026-05-05. Highlights auto_update.cooldown_days is now available in both the user-level global config (~/.config/prek/prek.toml on Linux and macOS, or $XDG_CONFIG_HOME/pr...
0114
Reposted by Alberto Torres Barrán
conda @conda.org · 21/04/2026
This is the year conda CLI gets a glow-up! 💅 Faster operations, clear progress, errors that actually help you fix things, better integration with PyPI, and new environment workflows. Conda maintainer Dan Yeaw laid it all out in our latest blog post 👉️
buff.ly
Building a Better conda CLI: A Vision | conda.org
A vision for the conda CLI in 2026: making conda fast, trusted, and delightful.
031
Reposted by Alberto Torres Barrán
bjkeefe @bjkeefe.bsky.social · 26/03/2026
A related blog post for those using pip, from @sethmlarson.dev: sethmlarson.dev/pip-relative...
sethmlarson.dev
Relative “Dependency Cooldowns” in pip v26.0 with crontab
WARNING: Most of this blog post is a hack, everyone should probably just wait for relative dependency cooldowns to come to a future version of pip. pip v26.0 added support for th...
033
Alberto Torres Barrán @albertotb.com · 26/02/2026
I'll try, thanks! I was hoping I could set movement to something move direct, ie., no prefix, like Ctrl + arrows or something. I got working Ctrl + S to split vertically and Alt + S to split horizontally.
000
Alberto Torres Barrán @albertotb.com · 26/02/2026
I can't even set custom bindings. They work to create new horizontal and vertical splits, but Im incapable of setting a keybinding to move between the splits
100
Reposted by Alberto Torres Barrán
Sung Kim @sungkim.bsky.social · 12/02/2026
It prevents wasted effort, keeps me in control of architecture decisions, and produces significantly better results with minimal token usage than jumping straight to code." boristane.com/blog/how-i-u...
boristane.com
Boris Tane
The research-plan-implement workflow I use to build software with Claude Code, and why I never let it write code until I've approved a written plan.
0113
Reposted by Alberto Torres Barrán
Emma @emmatyping.dev · 06/02/2026
maxhalford.github.io/blog/text-cl... This is a cool blog post about using compression.zstd in Python as a surprisingly effective (and fast!) simple text classifier.
maxhalford.github.io
Text classification with Python 3.14's zstd module
Python 3.14 introduced the compression.zstd module. It is a standard library implementation of Facebook’s Zstandard (Zstd) compression algorithm. It was developed a decade ago by Yann Collet, who hold...
153
Reposted by Alberto Torres Barrán
atticus goldfinch @atticusgf.bsky.social · 21/01/2026
the more uncomfortable corollary to this is that I think the days of humans _reading_ the code is also mostly gone. you're producing so much code now that a human reviewer is a massive bottleneck. IMO, you need to shift validation strategies massively towards tests and flagging high risk code.
169611
Reposted by Alberto Torres Barrán
Rodrigo Girão Serrão 🐍🚀 @mathspp.com · 17/01/2026
This uv cheatsheet lists 40+ of the most and useful uv commands. It groups the commands into 9 sections that correspond to major features/capabilities that uv has. The cheatsheet is free to download and comes in light mode, dark mode, and high-contrast mode. mathspp.gumroad.com/l/cheatsheet...
uv cheatsheet with 40+ commands and variations split into 9 categories.
043
Reposted by Alberto Torres Barrán
Electronic Frontier Foundation @eff.org · 11/12/2025
#E2EE must be explicitly protected, and client-side scanning must be excluded entirely. www.eff.org/deeplinks/2...
eff.org
After Years of Controversy, the EU’s Chat Control Nears Its Final
After a years-long battle, the European Commission’s “Chat Control” plan, which would mandate mass scanning and other encryption-breaking measures, at last codifies agreement on a position within the
011340
Reposted by Alberto Torres Barrán
rmoff 🏃‍♂️🫖🥓 @rmoff.net · 14/11/2025
A step-by-step guide to building a key-value database from scratch: www.nan.fyi/database - love the explainer/interactive animation in this
0105
Reposted by Alberto Torres Barrán
derek guy @dieworkwear.bsky.social · 14/11/2025
crazy IG em_clarkson
157110472499
Alberto Torres Barrán @albertotb.com · 29/10/2025
I'm actually not sure. I installed it in PowerShell and /doctor command showsthe info in the picture. But I don't know if underneath it uses git bash, because I've just realized that I happened to habe that installed as well.
100
Alberto Torres Barrán @albertotb.com · 28/10/2025
Agree, I hate this. Most of the time I cannot find the content I saw earlier because the feed changed completely.
030
Reposted by Alberto Torres Barrán
Peter W🤓 @wang.social · 28/10/2025
Why is it that all modern apps with “feeds” like social media apps or even reading apps like Medium, Substack, etc. exhibit the anti pattern of refreshing and losing your spot if you switch away for even a minute? It’s absolutely infuriating.
5704
Alberto Torres Barrán @albertotb.com · 27/10/2025
Actually, if you go to alternative installation methods, it can also be installed in PowerShell. At least that is how I did it
100
Reposted by Alberto Torres Barrán
Achim Zeileis @zeileis.org · 17/10/2025
💡 Came across this nice tool today: 🎨 qualpal for algorithmically choosing maximally distinct colors under certain restrictions #dataviz JOSS paper, online tool, R package #rstats joss.theoj.org/papers/10.21...
joss.theoj.org
Qualpal: Qualitative Color Palettes for Everyone
Larsson, J., (2025). Qualpal: Qualitative Color Palettes for Everyone. Journal of Open Source Software, 10(114), 8936, https://doi.org/10.21105/joss.08936
15213
Reposted by Alberto Torres Barrán
Anil Dash @anildash.com · 17/10/2025
"I have a decent fluency in LLMs, and they have utility, but the absurd degree of over-hype, the way they're being forced on everyone, and the insistence on ignoring the many valid critiques about them make it very difficult to focus on legitimate uses where they might add value."
291435218
Alberto Torres Barrán @albertotb.com · 16/10/2025
Interesting, since everyone is mentioning this as a big deal I though I did not understand it correctly (great post btw). Maybe I'm thinking too much about Claude Code and this is more useful in Claude web and the such...
100
Alberto Torres Barrán @albertotb.com · 16/10/2025
What am I missing, isn't this just basically the same as having those tools/scripts in a folder and list them in CLAUDE.md, describing what they can do, with maybe an extra markdown file with instructions? I don't really see the big gains here... Are we really reinventing folders, mds and scripts?
claude.md
Claude Code overview - Claude Docs
Learn about Claude Code, Anthropic's agentic coding tool that lives in your terminal and helps you turn ideas into code faster than ever before.
120
Reposted by Alberto Torres Barrán
Rami Krispin @ramikrispin.bsky.social · 16/10/2025
(1/3) Introduction to Multi-Stage Build 🐳👇🏼 The size of the image is a function of its dependencies and the efficiency of the image build. This tutorial focuses on resizing a #Python image using minimal images as our baseline and the multi-stage method. medium.com/data-science... #docker #mlops
medium.com
Introduction to Multi-Stage Image Build for Python
This post introduces the Multi-Stage build approach for setting up a lightweight dockerized Python development environment.
132
Reposted by Alberto Torres Barrán
mr. TIM @timkellogg.me · 13/10/2025
yes
A humorous comparison meme split into two halves.

Top half:
The Stack Overflow logo (stacked orange lines in a gray tray) appears on the left, with the text on the right:
“you’re absolutely wrong.”

Bottom half:
The ChatGPT logo (green interlocking hexagonal knot) is on the left, with the text on the right:
“you’re absolutely right.”

A thin black line divides the two sections. The image playfully contrasts Stack Overflow’s often critical tone with ChatGPT’s agreeable responses.
3616
Reposted by Alberto Torres Barrán
Rodrigo Girão Serrão 🐍🚀 @mathspp.com · 10/10/2025
I've turned this blog post cheatsheet into a downloadable cheatsheet. You can get the cheatsheet from here: mathspp.com/blog/uv-chea...
High-contrast uv cheatsheet with common and useful commands to create and manage projects and its dependencies and lifecycle, working with scripts, managing Python installations, working with and installing tools, uv's interface for pip and venv, meta commands, and miscellaneous commands. You can find an HTML version and the link to download the high resolution images at https://mathspp.com/blog/uv-cheatsheet.Dark mode uv cheatsheet with common and useful commands to create and manage projects and its dependencies and lifecycle, working with scripts, managing Python installations, working with and installing tools, uv's interface for pip and venv, meta commands, and miscellaneous commands. You can find an HTML version and the link to download the high resolution images at https://mathspp.com/blog/uv-cheatsheet.Light mode uv cheatsheet with common and useful commands to create and manage projects and its dependencies and lifecycle, working with scripts, managing Python installations, working with and installing tools, uv's interface for pip and venv, meta commands, and miscellaneous commands. You can find an HTML version and the link to download the high resolution images at https://mathspp.com/blog/uv-cheatsheet.
072
Reposted by Alberto Torres Barrán
sustrato @sustrato.bsky.social · 02/10/2025
Regresa @picanumeros.bsky.social para poner los superíndices sobre los sumatorios: contra la brocha gorda estadística y la varianza de barra de bar: «No se están “cocinando” los datos de forma maquiavélica para mostrar una realidad que no es» www.sustrato.io/textos/homer...
sustrato.io
Homer y la estadística pública | Ramón Ferri (Picanúmeros)
Hay que decirlo más: la estadística pública está en el punto de mira y nos afecta más de lo que pensamos.
046
Reposted by Alberto Torres Barrán
Willem Hilhorst @whilhorst.bsky.social · 19/09/2025
For those on LinkedIn who aren't aware. By default all profiles have now 'Data for Generative AI Improvement' turned on. You can turn it off via: Settings > Data Privacy > Data for Generative AI Improvement > Select the OFF option. Do not feed the planet-destroying machine!✊
45029
Alberto Torres Barrán @albertotb.com · 14/09/2025
By the way, what’s a AA? open.substack.com/pub/hushcras...
open.substack.com
By the way, what’s a AA?
Instead of fuzzy labels like 'indie' and 'AA', we propose a new data-driven system to classify video games based on their actual production scope. This new framework reveals the real economic and crea...
000
Reposted by Alberto Torres Barrán
Arvy @arvy.bsky.social · 14/09/2025
I love this analytical take on video game categorization from Antoine Mayerowitz and Julie Belzanne: hushcrasher.substack.com/p/taxonomy-o... Instead of trying to vibe-intuit the definition of an "indie" game, the authors analyzed the data from the perspective of game size and credits length
A graph titled Scope of video games according to Hushcrasher Classification System. The Y axis shows Game Size (from 10 MB to 100GB), the X axis shows Credits count (sourced from Mobygames).

The clustered dots on the graph represent 4 distinct scale categories of games - Kei, Midi, AA, AAA. The larger correlation between the size and the credits means larger category of a game.
311464550