Chazyl Arfi

CTO & Technical Architect
Informatique, Technologies et Citoyenneté.

[Chazyl Arfi @Chazyl]

Hâte de couvrir le @hellfestopenair avec le nouveau Galaxy S26 Ultra ! 🤘 J'espère juste que la logistique @SamsungFR pourra me livrer ma commande avant mon départ le 18 juin !

[Chazyl Arfi @Chazyl]

In February, I introduced "Steered-BMAD": using Sparse Autoencoders (SAEs) to steer local agents (SLMs) as a lightweight alternative to VRAM-heavy Fine-Tuning. Fast forward to late May 2026, and the research this exact hypothesis! https://t.co/JPlEzlRW0t
#SparseAutoencoders

[Tom Dörr @tom_doerr]

Fits 10 million documents into 4GB of RAM

https://t.co/ylQzuS7SuN

[Jake Fitzgerald @earthtojake]

new release for text-to-cad, an open source CAD harness and skills for codex / claude:

- mechanism validation (go from text prompt to functional mechanical design)
- parameters + animations for step files
- extended sdf, srdf, urdf support

3k stars, 10k downloads, we cooking

[Chazyl Arfi @Chazyl]

To optimize Ai Agents token efficiency the key factor is:
Fix Distance
Why language choice beats raw model size in many cases and how we can build better compilers for them:
https://t.co/usxM0HDiV6
#Rust #AgenticAI #AIEngineering #LLM

[Science girl @sciencegirl]

Kame is a compact, open-source, 3D-printable quadruped robot that mimics animal-like movement.

[Haocheng Xi @HaochengXiUCB]

𝗞-𝗺𝗲𝗮𝗻𝘀 𝗶𝘀 𝘀𝗶𝗺𝗽𝗹𝗲. 𝗠𝗮𝗸𝗶𝗻𝗴 𝗶𝘁 𝗳𝗮𝘀𝘁 𝗼𝗻 𝗚𝗣𝗨𝘀 𝗶𝘀𝗻’𝘁.

That’s why we built Flash-KMeans — an IO-aware implementation of exact k-means that rethinks the algorithm around modern GPU bottlenecks.

By attacking the memory bottlenecks directly, Flash-KMeans achieves 30x speedup over cuML and 200x speedup over FAISS — with the same exact algorithm, just engineered for today’s hardware. At the million-scale, Flash-KMeans can complete a k-means iteration in milliseconds.

A classic algorithm — redesigned for modern GPUs.

Paper: https://t.co/z0z0d3vrlp
Code: https://t.co/BqRfVKGH0K

[Rohan Paul @rohanpaul_ai]

Alibaba just open-sourced OpenSandbox ( a general-purpose execution environment ) to give AI agents an isolated environment to run code safely.

8k+ Github stars ⭐️

This stops your AI Agent based applications from accessing your actual host infrastructure.

By removing the hardest security roadblock, this release will massively accelerate how fast developers can build autonomous Agent based tools.

OpenSandbox puts the agent inside isolated runtimes like gVisor or Firecracker.

You can run it locally using Docker or scale it up using Kubernetes.

The system includes a code interpreter and a file system that the agent uses to complete tasks.

It also manages network traffic so you control exactly what the agent accesses online.

I think this will become the standard infrastructure for autonomous systems because building custom sandboxes is too dangerous for most teams.

[0xMarioNawfal @RoundtableSpace]

LuxTTS clones any voice from 3 seconds of audio on a 4GB GPU.

- 150x realtime speed
- 48khz output vs industry standard 24khz
- Fits in 1GB VRAM
- Works on CPU too

No ElevenLabs subscription. No cloud. Just open source.

The voice cloning barrier just hit zero.

link: https://t.co/7qg9hIdDEU

[Chazyl Arfi @Chazyl]

Maximising Local Inference Performance on Commodity RDNA2 Hardware https://t.co/4DmfQMEhYd via @LinkedIn

[Ryan Hart @thisdudelikesAI]

🚨BREAKING: Someone just open-sourced a headless browser that runs 11x faster than Chrome and uses 9x less memory.

It's called Lightpanda and it's built from scratch specifically for AI agents, scraping, and automation.

Not a Chromium fork. Not a hack. A completely new browser written in Zig.

Here's why this changes everything for AI builders: ↓

[Mistral AI for Developers @MistralDevs]

🎙️Run Voxtral Realtime locally with ExecuTorch!

💻Thanks to the ExecuTorch team, you can now deploy and run Voxtral Realtime efficiently and fully offline on your MacBook.

[Unsloth AI @UnslothAI]

You can now fine-tune Qwen3.5 with our free notebook! 🔥

You just need 5GB VRAM to train Qwen3.5-2B LoRA locally!

Unsloth trains Qwen3.5 1.5x faster with 50% less VRAM.

GitHub: https://t.co/2kXqhhvLsb
Guide: https://t.co/JCPGIRo99s

Qwen3.5-4B Colab: https://t.co/2Aj1mZ3f5j

[Ihtesham Ali @ihtesham2005]

🚨 Someone built a full Perplexity clone that runs 100% locally for $0.

It's called Perplexica.

→ Searches the web in real-time
→ Cites every source it uses
→ Works with Ollama local models
→ Multiple search modes (general, academic,
YouTube, Reddit, writing)
→ Zero API costs. Zero data collection.

Perplexity charges $20/month for this.

This runs on your machine for free.

29K stars. MIT license.

(Link in the comments)

[Linus ✦ Ekenstam @LinusEkenstam]

We just turned WiFi signals into a radar that can see through walls and estimate exact poses of people.

Surveillance just got order of magnitude more easy todo. No need for cameras.

Git hub repo close to 12k ⭐️
https://t.co/WTLX54egRi

https://t.co/89GAFr7f4r

[forloop @forloopcodes]

I cant believe this guy just made a permanent solution to context bloat and open sourced it all!

when we tested this tool (Context+) for solving an issue on the OpenCode repository, the agent using this tool used ~6.5k fewer tokens, found the code and fixed it in half the time!

the results were surprising: 6 to 10k tokens saved per prompt, completed task in ~2 minutes while the agent running without the tool took ~4 mins for the same and got stuck in loops

bro built an entire beast by using all the modern tools that we could think of: undo trees, semantic search by meaning (by haskellforall), advanced refactoring, blast radius, advanced file context trees, restore points... i can keep going on

semantic code search and context trees are the future of agentic coding and this tool proves it

the feature i loved the most is semantic search and how it gets things done 2x faster with least possible tokens

it makes an agent that actually knows what it’s doing and not just guessing, it makes meaning from your code similar to RAG. if you aren't optimizing your context, you are just burning money

the developer says this tool is still under development, it can have unexpected behavior and the docs need updates but the video shows the reality of how fast it can be

github: https://t.co/M0nwGDubAT
get here: https://t.co/PIJrM0KYa4

[Muhammad Ayan @socialwithaayan]

🚨 BREAKING: Alibaba just handed the AI agent community a production-grade sandbox for free.

OpenSandbox is a full-stack platform for running untrusted agent code safely:

→ Unified APIs across multi-language SDKs
→ Docker and Kubernetes runtimes purpose-built for agents
→ Browser automation, VS Code desktop, and network isolation included
→ Designed for coding agents, GUI agents, evaluation, and beyond

Not a side project. Built by Alibaba. Open source.

1.5k stars (+1,100 this week). The secure agent infra you didn't have to build yourself.

[Elon Musk @elonmusk]

People giving OpenClaw root access to their entire life

[Kruptos @KuptoKosmos]

🛠️🌐 WORLD MONITOR

Un nouvel outil vraiment impressionnant pour du gratuit/open-source :

Une salle de crise géopolitique en temps réel, maintenant accessible à TOUT LE MONDE !!

Avant, ce niveau d’info stratégique (comme un Bloomberg Terminal à 25000$/an !) était réservé aux gouvernements, banques et ultra-riches.

Aujourd’hui World Monitor offre cela gratuitement, open-source et 100% légal ! (tout repose uniquement sur des données publiques ouvertes)

Sur une carte du monde en 3D ultra-fluide, tu vois en direct :

👁️ Zones de conflits actifs avec scoring d’escalade

👁️ +220 bases militaires

👁️ Avions militaires en vol

👁️ Navires de guerre, y compris les dark ships qui disparaissent des radars !

👁️ Installations nucléaires mondiales

👁️ Câbles internet sous-marins, pipelines pétrole, clusters de data centers IA

👁️ Manifestations, sanctions, coupures d’internet, incendies détectés par satellite

👁️ Marchés de prédiction (Polymarket) comme signaux d’alerte précoce

L’IA intégrée analyse +150 sources news en continu, donne un Indice d’Instabilité (0-100) par pays et t’alerte quand plusieurs signaux convergent !

Avantages : gratuit, légal, puissant, fonctionne entièrement en local avec Ollama (zéro donnée envoyée), version desktop native, alertes ultra-intelligentes et améliorations très régulières grâce à la communauté active.

Inconvénients : dépend de sources publiques (quelques retards ou lacunes possibles sur des événements très locaux)

Près de 10 000 étoiles sur GitHub en quelques jours seulement !

Le projet s’améliore chaque jour 💪

#OpenSource

[Wildminder @wildmindai]

llmfit. Useful tool that probes hardware and tells you exactly which LLMs will actually run.
- handles MoE expert offloading, picks the best quantization for your RAM, estimates tokens/sec before you even pull the weights.
Essential for local dev.
https://t.co/qn8PmqQNbV