Pinned Tweet
As promised, here's the video in which I test @NVIDIAAI Nemotron Dory 1.0 4B, for 2 simple tasks, on a 16GB MacBookPro.
youtu.be/PZgSo56jlCI
Spoiler: Dory is a really good model.
I'm sorry, but no. I was one of your most loyal users for a year, but I simply can't get over the Fable debacle and all the circus that you're still pulling ahead of the IPO.
The only reason I'm quoting this is to make clear I don't even want to try it / endorse it / use it.
He already retweeted one of my posts. It was fun for a couple of days.
Now back to chop wood, carry water.
What's even more concerning is that, for the AI-insulated normies, these hacks are just "business as usual, just some hackers on the internet".
No, they aren't.
They're 1000x amplified hackers.
Eventually, they’ll use it anyway. Just make them independent and responsible.
There's a lot of discussion about giving young people AI to use during their educational journey. Here's what I think:
Giving a kid AI to do their 9th-grade homework is a travesty.
Asking them to use AI to design an interplanetary spaceship is education.
Point the tools at the biggest problems they can imagine!
Another hour, another model launch! This one is also built from scratch and weights already available on huggingface. It claims more than decent speed on a MBP. Very interesting!
We live in the age of abundance. On Earth and in the skies as well.
If there's one thing I can tell you about @SpaceX, it's that there's never been a company like this...
My thoughts on the earnings call with @SquawkCNBC.
Happy to see a Korean player launching their very own model, trained from scratch. Looks very promising!
Today, we’re releasing a preview of Raon-1 mini, a 30B MoE LLM with 4B active parameters, trained from scratch by KRAFTON AI.
With our efficient architecture, it matches Muse Glimmer on coding/agentic tasks while being much faster and more memory-efficient! 🔥 Plus, natural, human-like conversation.
This is just a first look. The final release (and bigger models) are coming soon! 👀
👉 Try it now, free for a limited time: raon-mini.krafton.ai/en
🤗 Open weights + DFlash drafter (Apache 2.0): huggingface.co/KRAFTON/Raon-…
Relatable. I’m 55 and came out of semi retirement in February this year, harnessing AI for my iOS apps. I’m enjoying it tremendously.
Not to brag or anything, but I still have this bad boy in an attic…
I challenge you to post your vintage (working) gear below.
Hmmm, I’ve been calling thus for a while: API token prices are not real, they’re aspirational: xitter.cf/dragosroua/status/2077…
deepseek opensourcing deepinfer is a textbook baduk encircle movement masked as open source benevolence
while SV hyperscalers burn billions defending artificial 90% margins on raw compute, china is quietly commoditizing the entire execution stack to force a global race to the bottom
winning a technological war comes down to turning an opponent's primary revenue engine into a free public utility
by making hyper efficient inference the new baseline, China is forcing Western compute cartels directly into an unsustainable asymmetric trap
Dragos Roua retweeted
github.com/ddalcu/mlx-serve/…
Many new models
DeepSeek-V4.1-Flash is much faster
Chat on the Neural Engine
Meeting notes from the menu bar
Speech to text transcriptions api
RDMA
Images, video, music and 3D up to twice as fast on M5 Macs
Frontier models on a 128 GB Mac via Streaming
An OpenAI agent walks into a bar.
- What should it be? asks the bartender.
- Scotch on the Groks, says the agent.
- Oh, rogue again?…
Making money for the sake of money sucks out the soul from your body.
Mindless altruism does the same.


















