
It’s never been more important for Opensource AI to win

Co-Founder @ Emilia AI copilots for the physical world, Prev. @vocodehq, Urban Biker, Pythonista, geek, ML Open-Source. real-time communication. Sapere Aude.

It’s never been more important for Opensource AI to win

repost for visibility. 😍
Please can someone at @Cloudflare reply to this. It would be great if there was a way to reuse isolates (with wasm) on websocket connections. github.com/cloudflare/worker… cc @dok2001 @KentonVarda

yes that's the irony ultra-closed models really like to use small open weights models locally to transcribe, classify, judge, etc...
Astra 6.2 cheated a bit on this one! I asked it to create a photorealistic 30 second movie of the life in the eyes of a squirrel. I was expecting it to use ThreeJS, Unreal or Unity for the job, but instead, it generated images with imagegen and downloaded a local img-to-video model to animate them...at this stage I am not even mad...

This was my bitter lesson. I spent 2 years at my startup designing, architecting and coding computer vision pipelines for 30+ different restaurant use cases. When GPT-6 Astra came out, I felt I was not longer de best one to design them, and my ego took a real hit. Luckily I got there fast enough to see where i can still add value, because it's very easy to the ego stay in denial. how does it make you feel?

Introducing Qwen3.8-27B-pi-adapted 🤗 We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter. The results, measured as pass@1: 🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7% 💻 Terminal-Bench 2.1: 39.5% → 55.8% ⚡ 55–62% fewer tool calls per task This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛 Open weights: huggingface.co/harness-hub/Q… Methodology: huggingface.co/blog/harness-…

GPT-6 Astra led Roboflow vision evals for about five weeks. Sep 5: Astra takes #1 in OD, 82.1 mAP@50 at low effort. Oct 13: Gemini 4 Argon passes it. Oct 27: OpenAI's Own GPT-6.1 Terra beats it on detection at 92% lower cost. I charted the run on the old 5 task suite: arpagon.co/writing/astra-vis… Next chapter is this week: @skalskip92 harder detection test, Argon, Carbon, Hydrogen and Terra aren't in yet. I'll update the chart when they are.


We’re not going to win by lobbying, plotting, or manipulating. We’re going to win by building, in the open, with the whole community! Today, with @ggerganov we’re super excited to announce the release of llama.cpp v1, the first major fully stable version of the most used library to run AI models locally, meaning on your phone, your laptop, and your workstation. It means you can own your intelligence for free, not send your data to anyone and be 100% in control, wherever you are in the world. It shows our commitment to support every hardware, platform and model type out there and makes AI distributed rather than concentrated in a few Silicon Valley orgs. Huge thanks to the team, the long-time maintainers and every contributor who helped make this happen. Let’s build! 💛🧡💚
llama.cpp v1.0.0 is now available huggingface.co/blog/ggml-org…

Fable 5.2 is equal parts brilliant and unhinged. I asked for a quick PoC with barely any context, and it went off and designed a whole wearable to solve it.


Wow that was unexpected 😍. the latest version of t3code by @theo now renders my pi extensions. Here it's rendering pi-context-zone Inspired by @dexhorthy "No Vibes Allowed" talk. `pi install npm:pi-context-zone`

Si hoy estuviera empezando en tech en América Latina, con tanta incertidumbre y cambios por AI, buscaría una comunidad. Construir solo es muy difícil, por mas que los modelos están mejorando.

buzz 1.0 is finally here! the most flexible team workspace out there. for humans and agents, model agnostic, local or cloud, and infinitely extendable. buzz.xyz

if you only want to read an article this week, please read this.
keeping the flame thesephist.com/posts/flame/

Love llama.cpp, but these days "llama" means something else. With Meta taking a different path, maybe it's time to brainstorm a new name for the project? @ggerganov @ClementDelangue @Thom_Wolf @mervenoyann @QGallouedec

Markdown is my sole creation. Aaron Swartz was my best, by far, beta tester and the first person other than me to get excited about it.