Hector Him
63 posts
MiMo-V2-Pro & Omni & TTS is out. Our first full-stack model family built truly for the Agent era.
I call this a quiet ambush — not because we planned it, but because the shift from Chat to Agent paradigm happened so fast, even we barely believed it. Somewhere in between was a process that was thrilling, painful, and fascinating all at once.
The 1T base model started training months ago. The original goal was long-context reasoning efficiency. Hybrid Attention carries real innovation, without overreaching — and it turns out to be exactly the right foundation for the Agent era. 1M context window. MTP inference for ultra-low latency and cost. These architectural decisions weren't trendy. They were a structural advantage we built before we needed it.
What changed everything was experiencing a complex agentic scaffold — what I'd call orchestrated Context — for the first time. I was shocked on day one. I tried to convince the team to use it. That didn't work. So I gave a hard mandate: anyone on MiMo Team with fewer than 100 conversations tomorrow can quit. It worked. Once the team's imagination was ignited by what agentic systems could do, that imagination converted directly into research velocity.
People ask why we move so fast. I saw it firsthand building DeepSeek R1. My honest summary:
— Backbone and Infra research has long cycles. You need strategic conviction a year before it pays off.
— Posttrain agility is a different muscle: product intuition driving evaluation, iteration cycles compressed, paradigm shifts caught early.
— And the constant: curiosity, sharp technical instinct, decisive execution, full commitment — and something that's easy to underestimate: a genuine love for the world you're building for.
We will open-source — when the models are stable enough to deserve it.
From Beijing, very late, not quite awake.
English
The next chapter of America’s journey to explore the solar system begins TONIGHT.
Artemis II and the SLS rocket roll out of the Vehicle Assembly Building to Launch Complex 39B as we target a launch attempt as early as April 1. This mission will potentially send astronauts farther into space than any human has traveled before - around the Moon and safely back home.
And, under @POTUS’ National Space Policy Directive, we’re just getting started.
English
Very proud of our new documentary on rare disease — the families and the path toward a cure. Join us at 7 pm ET tonight on @CNBC.
English
You know a technology is good when you don't even realize you've used it.
For this scene,
- I created the character as an image with Grok Imagine, then the landscape with Grok, and combined everything using Grok Imagine's image references.
- I then used video extensions to avoid cuts and style changes.
The character is perfectly integrated into the scene, proportions respected and style preserved.
English
@ylecun @ronbrachman Looking forward to a podcast or an article about how you see world models shaping the current AI landscape. Good luck!
English
The basic idea of world models is very old.
Optimal control folks were using model-based planning in the 1960s (using the "adjoint state" methods, which deep learning people would now call "backprop through time").
But the real question is what you do with this idea and how you reduce it to practice.
English
Congrats, @ylecun!! Well done.
Of course many of us AI researchers have been working on world models since the 1970's, so let's make sure all of that great historical work doesn't get forgotten or reinvented...
Rohan Paul@rohanpaul_ai
📢 BREAKING: FT reports that Yann LeCun’s startups AMI Labs raises $1.03 bn to build world models, at a pre-money valuation of $3.5bn. Congratulations @ylecun 🚀 The financing positions the company as a test of LeCun's belief that today's large language models fall short of human-level reasoning and autonomy. LeCun earlier said AMI aims to build systems capable of reasoning and planning in complex real-world settings. AMI Labs (Advanced Machine Intelligence Labs) aims to solve the limitations of standard language models by building world models using the Joint Embedding Predictive Architecture to observe spatial data. This visual framework helps the AI internalize how objects behave so it can safely plan complex actions. Relying exclusively on text limits AI to human linguistic output while ignoring the massive bandwidth of unspoken physical laws. Building predictive spatial architectures is the mandatory leap required to achieve reliable autonomous agents. Building predictive spatial architectures is the mandatory leap required to achieve reliable autonomous agents. This fundraising included backing from a global group of investors, including France’s Cathay Innovation, Amazon founder Jeff Bezos’s Bezos Expeditions, Singapore’s Temasek, Seoul-based SBVA and US chip giant Nvidia. The company's near-term target customers are organizations operating complex systems, including manufacturers, automakers, aerospace companies, biomedical firms and pharmaceutical groups. Over time, he added, the technology could also support consumer applications. "What consumers could be interacting with is a domestic robot. You need a domestic robot to have some level of common sense to really understand the physical world." LeCun said he was also talking with Meta about potentially deploying the technology in its Ray-Ban Meta smart glasses. "That's probably one of the shorter term potential applications," he said. --- ft .com/content/e5245ec3-1a58-4eff-ab58-480b6259aaf1
English
Hector Him đã retweet
The video at simbrain.net shows some of the main new features. The attached pictures show: a chaotic attractor, a simple retina reacting to a flower input, and part of a transformer model.
A full discussion of what’s new in Simbrain 4 is at docs.simbrain.net/docs/whatsnew
English
I’m thrilled to announce the release of #Simbrain 4, which I have been working on for over 10 years! For the last five of those years I have met almost every weeknight with Yulin Li and we have pair-programmed our way through the entire app, adding hundreds of new features. 1/
English
@mkristensen Vim like navigation, better terminal experience and a stronger AI and agents support
Slow sometimes, but I am just starting to use the newer version and it appears a lot faster
English
What features or extensions make you jump from Visual Studio to other IDEs and editors to perform certain tasks?
English
I sat down with @mignano in my hometown to share stories about my childhood and what life is like when you’re customer support for 500 million people.
English
@Lister38 @MilHistNow The F-117A was tested in Panama for the first time during operation just cause. There were some issues because some targets were missed
English
@MilHistNow I think this was also the first use of the F-117 stealth fighter as well.
English
On this day in 1989, 27,000 American troops invade Panama to topple the dictator and former U.S. ally Manuel Noriega. The global community condemns the operation as a violation of international law.
English
All of my Deep RL course lecture videos from Spring 2025 are now online! 🥳
Youtube playlist: youtube.com/watch?v=EvHRQh…
YouTube
English
A little video of my PDP-11/83 doin' stuff. Compiling code, being cool, blinking its lights.
English
Courant is now a full-fledged school within NYU: the Courant Institute School of Mathematics, Computing, and Data Science.
NYU Courant@NYU_Courant
NYU (@nyuniversity) announced the creation of the Courant Institute School of Mathematics, Computing, and Data Science today, signaling the university’s enthusiastic commitment to mathematics, computing, and data science over the coming decades: nyu.edu/about/news-pub…
English
Hello everyone,
I’m currently deciding on the topic for my next Substack post and would love to get your input.
Here are the three ideas I’m considering:
1. How did Samsung rise from a latecomer to defeat Japan’s once-dominant DRAM giants?
2. How did Korea’s OSAT (semiconductor packaging) industry collapse? Korea’s semiconductor packaging sector was sliced up and sold like a cake to China and the U.S.
3. Should we really see India as the alternative to China? Could India become the next China—turning into our competitor or even challenging U.S. dominance?
If you have other topics you’d like to read about, feel free to suggest them below.
Thank you!
English
Strongly disagree with the original post, and agree with that Berkeley, Stanford, and UCSD actually do offer many good courses that are cutting edge and timely.
For example, this Winter I offered this machine learning systems course hao-ai-lab.github.io/cse234-w25/ at UCSD (all materials are public available btw) which attracted 220+ students across UCSD CSE/HDSI/ECE
I covered how DeepSeek V3 was made literally 2 weeks after its release, and one of the programming assignment my TA team designed was to implement an all2all primitive, which precisely was one of the core innovation made by DSK-v3.
Also got very encouraging feedback from students, too, to let you appreciate the happiness of being a teacher 😀
Probably will make a Youtube course with more latest content in both English and Chinese next quarter when I have more time to make it more accessible!
Jelani Nelson@minilek
At @Berkeley_EECS we always work to keep our curriculum fresh. Our intro ML course CS 189 just got a drastic makeover this semester (thanks @profjoeyg @NargesNorouzi!) and now includes ~12 lectures on e.g. Adam, PyTorch, various NN architectures, LLMs, and more (see eecs189.org/fa25/). We also this semester launched a modern, brand new mezzanine NLP course by @alsuhr (EECS 183/283A, cal-nlp-class.github.io/fa25/), with a made over Advanced NLP course launching in Sp26 by @sewon__min.
English
Headed for my 2nd emergency surgery in less than a year this evening, send some good vibes this way if you can folks
English
Despite handling unprecedented levels of fame, the grind never stops.
I just published 2+ hours of hardcore Functional Programming in Modern C++.
Enjoy.
Dmitrii Kovanikov@ChShersh
Day in the Life of a Senior SWE at Bloomberg
English
@ChShersh I don’t think it will affect current H-1B visa holders.
It is my understanding that F-1 (students in the USA) visas to H-1B may not have to pay this
English
I'm sending hugs to all my American friends on H-1B visas.
This is going to make a huge impact on the tech sector.
Watcher.Guru@WatcherGuru
JUST IN: 🇺🇸 President Trump officially signs executive order to impose $100,000 annual fee on H-1B visas.
English