Make sure you own your AI. AI in the cloud is not aligned with you; it's aligned with the company that owns it.
Sofia
Joined August 2013
- Tweets6.7K
- Following1.5K
- Followers5.1K
- Likes12.9K
Mitko Vasilev retweeted
Selling my two NVIDIA RTX PRO 6000 Blackwell 96GB:
m.olx.bg/d/ad/nvidia-rtx-pro…
And
m.olx.bg/149715038/?bs=olx_p…
Mitko Vasilev retweeted
This is pretty ridiculous. Their code was licensed permissively, and other projects (notably @OmarchyLinux) used their code. As how Open Source works.
Now they close shop when Omarchy is making their vision a reality. Completely unserious.
Asahi Linux is moving everything from the MIT licence to GPLv2-or-later.
Asahi is the volunteer project that reverse-engineers Apple Silicon Macs so they can run Linux. In today's Linux 7.3 progress report, developer James Calligeros explains why: people have been using the permissive licence to take the team's work. "In some extreme cases, in-development feature code was being lifted straight out of our personal forks and published without attribution or lazily rewritten with LLM assistance to obfuscate its origin."
The report points to WINE, which switched from MIT to LGPL in 2002 for the same reason: "Those who do not learn from history are doomed to repeat it."
The same report brings real wins too. M3 Macs now install without the EXPERT flag, Thunderbolt support (USB3 and networking for now) is shipping to users, and the team got GTA IV and Fallout: New Vegas running outside the muvm VM using Google's per-process page size patches.
Mitko Vasilev retweeted
From GW-scale DCs to single-container POD solutions, Supermicro’s DCBBS helps build AI Factories faster, better, greener, and at a lower cost! Customer-proven DCBBS integrates optimized hardware, management software, and services to minimize TTD and TTO while maximizing compute uptime, directly saving time and increasing ROI - A true game changer!
Mitko Vasilev retweeted
who's building Tinker for weight-space programming
the infra needs for WSP may allow for great economics
Mitko Vasilev retweeted
Today I'm releasing my LLM fine-tuning recipe, so anyone can build their own models.
I've created over a hundred models over the past few years (Tess and Synthia), and it has never been a better time than today to fine-tune your own models.
Here's a skills repo, to import to your coding agent. It has all the instructions you need, with Python scripts. Use your own Claude Code sessions, and Tinker for fine-tuning.
Repo: github.com/drost-ai/tinker-a…
Go build some models! 🚀🎉
Mitko Vasilev retweeted
Bun to RCE!!
@jarredsumner @bunjavascript
Fuck you and your bun project.
You marked my report INVALID. Told me bin linking "has nothing to do with trustedDep" and waved it off as "not a bun issue."
1.4.4 patched it but I Bypassed anyway in empty folder.
github.com/X3r0Day/BunnyHija…
Mitko Vasilev retweeted
Qwen3.8-Flash-Next GGUFs now run faster with less thinking and less token use.
Unsloth Dynamic V3.1 GGUFs bring thinking & token use to match 95% of full precision Qwen3.8-Flash.
Results: 50% fewer tokens + 15% higher accuracy vs same size quants.
GGUF: huggingface.co/unsloth/Qwen3…
We’re releasing new Qwen3.8-Flash-Next GGUFs with 15% higher accuracy.
Unsloth Dynamic V3.1 outperforms others by >15% on Terminal Bench, DeepSWE & more benchmarks (256K context).
Run on 32GB+ RAM setups.
GGUF: huggingface.co/unsloth/Qwen3…
Blog: unsloth.ai/docs/models/qwen3…
Mitko Vasilev retweeted
🚀 Zvec-grep 0.3.0 is here — rebuilt in Rust.
Local search for developers and AI agents, now with less waiting and a lighter footprint.
⚡ Less time indexing
Across four BEIR datasets, total index build time dropped from 219 s to 78 s—about 65% less time than the Node.js version.
🪶 Lower memory use
Average peak indexing memory fell by 53%, leaving more room for your editor, agents, and other tools.
🔎 Smoother everyday use
An rg-inspired CLI, clearer model selection and indexing progress, plus fixes for file tracking, result locations, and index management.
🤖 More agent integrations
Now supports GitHub Copilot, Grok Build, VS Code, and Pi, alongside existing integrations.
Try it in your workspace and tell us how it goes.
zvec.org/en/blog/2026-11-03-…
Mitko Vasilev retweeted
18 of the 20 best open models today are Mixture-of-Experts. Dense is basically over.
And yet the know-how to train a good MoE still lives mostly in the heads of a handful of teams.
So we spent 800k GPU hours running the experiments ourselves and wrote everything down. Super happy to finally release "The Ultimate MoE Training Handbook", free and open, organized around the questions training teams hit:
- how many experts, how big, shared or not?
- why is half my step spent in all-to-all, and how do I hide it?
- why is my run 25x slower with no error in the logs?
- softmax, sigmoid or sqrt(softplus) router, and why keep it in fp32?
- aux loss or DeepSeek's loss-free bias?
- upcycle a dense checkpoint or train from scratch?
- can I serve with fewer experts per token than I trained with?
⚠️ Warning: it's long (~70k words, budget a few days, sorry) but with 55 interactive figures, code you can run and numbers from our own runs.
Read it here: huggingface.co/spaces/huggin…
Mitko Vasilev retweeted
Changing one letter bypassed an entire admin API in @huggingface chat.
One request could trigger bulk internal exports of every user’s private chats, completely unauthenticated.
Here’s how. 🧵
There is no perfume quite like “New GPU Smell"
It hits different at 10 PM when you realize you will to be up at 7 AM.
But the 8xB300 GPU cluster is green, the latency is low and the Kimi-K3.1-Pro token buffet is officially open.
Mitko Vasilev retweeted
llama.cpp is featured in the Github Octoverse report
The project remains consistently one of the top open-source projects on Github over the last few years. An attestation to it's great quality and awesome community!
Mitko Vasilev retweeted
Official now! I'll be in SF 11/29-12/5! I hope to be able to meet amazing people while there and see some cool AI projects and ideas! 🚀
I'll probably come again in future!
DM me to grab a coffee and talk about Local AI (cloud is ok too)!
Mitko Vasilev retweeted
Kimi K3.1 Pro is now open-weight!
Releasing the model weights of Kimi K3.1 Pro: huggingface.co/moonshotai/Ki…
We'll keep advancing open intelligence.
Mitko Vasilev retweeted
R9700 AI PRO is now $3,000.
Gorgon Halo 495 192GB is now $7,000.
Lucebox kept the same pricing and is still $8,100 for both cards combined: shop.lucebox.com
We got very lucky and locked supply back in July, so we’re still selling at July costs. We don’t have many units left, and once they’re gone, our price goes up.
Could we make way more by raising the price today? Yes. But our goal has always been to provide community the best price/performance local setup.
Mitko Vasilev retweeted
(wild) predictions after today's MIT Weight-Space Programming work:
1. Model weights become a programming language that AI can understand, but humans can't.
2. Training and inference will merge. Like human brain, AI could continuously rewrite its own weights while thinking.
3. A new path to RSI: AIs will directly edit each other's weights, evolving together without backprop.
