yarenty retweeted
You can now train MoE models 12× faster with 35% less VRAM via our new Triton kernels (no accuracy loss).
Train gpt-oss locally on 12.8GB VRAM.
In collab with @HuggingFace, Unsloth trains DeepSeek, Qwen3, GLM faster.
Repo: github.com/unslothai/unsloth
Blog: unsloth.ai/docs/new/faster-m…
yarenty retweeted
🦞 OpenClaw 2026.2.9 just dropped
🔍 Grok web search provider
🧠 No more post-compaction amnesia
🛡️ Context overflow recovery
⏰ Cron reliability overhaul
+ 40 more fixes from 25+ contributors
Elon we added your model btw, you're welcome. github.com/openclaw/openclaw…
yarenty retweeted
Ultra-lightweight alternative to OpenClaw for embedded systems.
cnx-software.com/2026/02/10/…
PicoClaw is an ultra-lightweight personal AI Assistant designed to work on less than 10 MB RAM and suitable for resource-constrained embedded boards such as the $10+ @SipeedIO LicheeRV Nano SBC powered by a SOPHGO SG2002 RISC-V SoC with 256MB on-chip DDR3.
The code is written in Go and generated with the help of AI. It works on 64-bit Linux (RISC-V, ARM, and AMD64) and Windows (AMD64).
yarenty retweeted
I always have fun bringing this one up! (BTW, std::collections::linked_list exists...)
🦀 Learning Rust With Entirely Too Many Linked Lists is one of the best ways to really understand Rust’s ownership model.
• ownership and borrowing
• Box, Rc, RefCell, Option patterns
• where unsafe bites (and how Miri helps)
rust-unofficial.github.io/to…
#rust #rustlang
yarenty retweeted
If you are using OpenClaw, do this before it's too late:
Scan every one of your skills using this Skill Scanner. Do not install new skills without first doing this.
This app shows you exactly what skills OpenClaw can access before it runs.
Do not let an AI agent run without understanding what it can do. Remember, these agents have access to your files, your data, and external tools, and they don't ask for permission.
Gen Digital (the Fortune 500 company behind Norton and Avast) is building the "Agent Trust Hub". This is a trust layer for autonomous AI agents.
The OpenClaw Skill Scanner is the first tool they released.
A good analogy is to think about this like a nutrition label for AI skills.
It tells you what the skill can do, what data it reads, and what APIs it can call before you decide to trust it.
Your AI agent found a skill. Installed it. And it charged you $39,214. No bad intent. No warning. Just automation. Agent Trust Hub helps agents check skill safety BEFORE they act. ai.gendigital.com/agent-trus…
yarenty retweeted
I spent a week inside OpenClaw (Clawdbot) communities watching how the best AI builders actually use it.
Most tutorials show you the basics.
The real magic happens when you start mixing OpenClaw with custom agent workflows that handle your whole research pipeline.
I watched one builder set up a system that scrapes competitor data, tracks pricing shifts, and spits out market reports while he sleeps.
Another built an agent that keeps tabs on 50+ data sources, grabs the important stuff, and drops it straight into his product roadmap.
The breakthrough moment: OpenClaw goes way beyond scraping.
These builders use it as the data collection layer for research setups that run nonstop.
Your competition is building research machines while you're still copying and pasting from Google.
yarenty retweeted
Are you using Homebrew? Sorry to hear that, check this out:
🍺 zerobrew — A modern drop-in replacement for Homebrew on macOS
💯 Content-addressable store + parallel downloads = 20x faster
🦀 Written in Rust!
⭐ GitHub: github.com/lucasgelfond/zero…
#rustlang #cli #devtools #macos #homebrew #packagemanager #performance
yarenty retweeted
Why do most LLM agents hit a wall?
They don’t accumulate skills.
Introducing SkillRL📚 — recursive skill-augmented reinforcement learning that lets agents learn skills from failure and evolve over time.
🔥A 7B model:
• +41% over GPT-4o
• ~20% fewer training tokens
• 33% faster convergence
SkillRL bridges raw experience → policy improvement by distilling trajectories into structured, co-evolving skills during RL.
Most agents forget.
SkillRL evolves. 🔄
📄 Paper: arxiv.org/abs/2602.08234
💻 Code: github.com/aiming-lab/SkillR…
Great work @richardxp888, Jianwen Chen, Hanyang Wang, @JiaqiLiu835914, @lillianwei423, @AiYiyangZ, and nice collab. w/ @__YuWang__, @XujiangZhao, Haifeng Chen, Zeyu Zheng, @cihangxie.
yarenty retweeted
⚡️ Using PIO for LED Blinking on Raspberry Pi Pico in Embedded Rust
- Intro to PIO in Raspberry Pi Pico
- Simple PIO assembly code to blink LED
- Guide to use the PIO Simulator to visually understand the signal generated corresponding to asm
rp2040.implrust.com/pio/inde…
yarenty retweeted
Here's a classic... reimagined for the terminal 🔥
🧱 tetrs — A modern Tetromino game with a TUI
💯 Multiple game modes, smooth controls, custom rotation systems, animations & deep mechanics.
🦀 Written in Rust!
⭐ GitHub: github.com/strophox/tetrs
#rustlang #tui #terminal #games #tetris #gamedev #opensource
and ... russian comments in source code 😉
koi.ai/blog/glassworm-first-… - this is creeeeaaaazy!!!
🚨 Does using AI like ChatGPT make your brain "lazy"? MIT Media Lab's study, led by Dr. Nataliya Kosmyna, reveals cognitive debt in students writing essays with AI. Brain scans show less engagement vs. no-tool writers! 🧠💻 Curious? Dive in: bit.ly/45oPw5i #AI #Learning
8/8 Reminds me of electronic calculators: When invented, folks thought “no more mathematicians!” Not true:we still need math skills, even if daily calcs are rare. AI gives access to all, but over-reliance might dull thinking, like struggling to add shopping prices without a phone
This is a preprint (June 2025), with limitations like small sample. But it’s a wake-up call for AI in education. Check the full paper: arxiv.org/pdf/2506.08872 More studies needed! What do you think—AI helper or crutch? #YourBrainOnChatGPT #xAI