←
Yuandong Tian1,189 posts
Yuandong Tian avatar

Yuandong Tian

@tydsh

Co-founder of @Recursive_SI. ex-Meta FAIR Director. ex-Google. Reasoning, Optimization and Understanding LLM. Novelist in spare time. PhD in @CMU_Robotics.

◷ Joined December 2009▧ 129 Photos and videos
Yuandong Tian avatar
Yuandong Tian@tydsh·Oct 21

Today, we release our first model, Flint-40B-A2B, a fast local coding model built with our automated research system. huggingface.co/RecursiveSI/f… 1. Strong performance. 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified. 2. It is trained from scratch (tokenizer->Pretrain->Midtrain->SFT->RL). 3. Fast in your local laptop. Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark. 4. We build the complete model training pipeline in 4 month!

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

2:41
Yuandong Tian avatar
Yuandong Tian@tydsh·15h

🫡

Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen

Yuandong Tian avatar
Yuandong Tian@tydsh·Nov 16

Thanks for the invitation! Looking forward to tomorrow's discussion!

The day is almost here! We're all geared up for the Backprop Summit. And we've got a few more incredible names joining the lineup! @sarahookr, Co-Founder @adaption_ai @adityagrover_, Co-founder & CTO @_inception_ai @AlexGDimakis, Co-Founder & Chief Scientist @bespokelabsai @xiao_ted , Founding Member of Technical Staff @PrometheusInc Gagan Bhat, Member of Technical Staff @AnthropicAI @tydsh, Co-founder @Recursive_SI @bekacru, Member of Technical Staff @vercel amongst @MichelTricot and @HamelHusain We're almost sold out. Last few spots left luma.com/backprop

Post attachment
Yuandong Tian avatar
Yuandong Tian@tydsh·Nov 6

Great to see @Recursive_SI ranked #7 for Top Startup that attracts top talents!

Top 30 Startups and BigTech companies based purely on ~50,000 engineer job changes in the last 3 years. If engineers leave a company to join another, it ranks higher especially if the company they come from ranks highly. Here are the final results:

Post attachment
Yuandong Tian avatar
Yuandong Tian@tydsh·Nov 3

Interesting work! I wonder whether this also works for other models than Qwen3.5-0.8B? It could be possible that the mech interp analysis of the Qwen model has been incorporated into the training set of Astra. E.g., can this approach make our recently released Flint 40B-A2B model better 😀 There could be also simple reward hacking like closing the thinking mode of the model to avoid hitting the thinking tokens limit..

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…

Post attachment
Cline avatar
⟳ Yuandong Tian retweeted
Cline@cline·Oct 21

We’re about to see a breakout in local model capabilities. Flint scores 54.9% on Terminal-Bench 2.1 with ~2B active parameters. That’s ahead of every other small MoE and two of the three dense ~30B models. It only needs about 24GB and runs at 125+ tok/s on a MacBook Pro.

Post attachment
Recursive avatar
⟳ Yuandong Tian retweeted
Recursive@Recursive_SI·Oct 21

Recursive Flint is here: our first model, built with the Eureka Machine. Fast, local coding. Open weights available on Hugging Face.

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

2:41
Yuandong Tian avatar
Yuandong Tian@tydsh·Oct 21

👉👉 AI Scientific Discovery Demos with our Recursive Flint-40B-A2B 🪨🪨!

1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵

Yuandong Tian avatar
Yuandong Tian@tydsh·Oct 17

A great read. The blog post captures years of progress in mechanistic interpretability and offers an optimistic vision: perhaps we really can open the black box and understand how neural networks work. I share that optimism, but with some reservations. I'm skeptical of explanations that stop at individual neurons, accompanied by compelling visualizations that seem to "explain everything." Humans are remarkably good at convincing themselves. It's easy to cherry-pick evidence that supports a hypothesis, especially when the visualizations look convincing. I'm also not convinced that "Recursive Generalization" is fundamentally different from hierarchical representation learning. As representations become more abstract, they discard details while preserving invariances. Isn't that essentially what hierarchical representations have always been about? IMHO, we need to move beyond a largely descriptive, behavioral approach toward a principled mathematical framework. And we have a natural starting point: gradient descent. Rather than just asking what a network has learned, we should understand why learning dynamics produce these structures in the first place. Of course, this is extraordinarily difficult. Look at OpenAI's 719 mathematical proofs. We see plausible proofs tackling problems at the Fields Medal level, yet no comparable breakthrough that cracks deep learning theory. That contrast is striking. Part of the difficulty is that there are countless plausible theories of how deep networks work. Proving even one can be mathematically formidable. And a proof alone isn't enough. Each theory needs empirical validation, often requiring substantial GPU compute. It's a daunting combination of mathematical and experimental challenges. Still, I believe this is the way forward. Interpretability gives us observations and hypotheses. But ultimately, we need a theory that explains why these phenomena emerge from the learning process itself. That's how we turn compelling stories about neural networks into actual understanding. And hopefully recursive self-improvement can help with that 😀

interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works already-clean.goodfire.com/

Yuandong Tian avatar
Yuandong Tian@tydsh·Oct 6

Thanks for the interview @ziv_ravid! It is a fun discussion 😀

New episode on The Information Bottleneck, with Yuandong Tian @tydsh (previously at Meta and now a co-founder of Recursive Superintelligence). He told us that after writing his last paper at Meta with GPT-5, he realized his job could be replaced in five years. "Instead of losing my job involuntarily, I think it's good to start a company." But he also says the models still can't come up with good ideas. You ask for one, and you get something mediocre that everyone already thought of. We also talked a lot about RL, Go, Continuous Latent Space CoT reasoning (Coconut), and the calls to restrict self-improving AI. I’m sure you will enjoy it. You can find the episode on our website, YouTube, and all the apps

Post attachment
Yuandong Tian avatar
Yuandong Tian@tydsh·Sep 2

We got exact the same comments when we published Coconut😀. However, even with explicit CoTs, the model can still learn to hide the true thinking process. The ultimate solution is always to understand how the model works.

new: OpenAI & others quietly using loop transformers that don't show their 'thinking' when scaled up a leap forward on performance, but sparking concerns inside & outside OpenAI re: security as this takes off

Post attachment
Yuandong Tian avatar
Yuandong Tian@tydsh·Sep 1

😅 so our StreamingLLM (the original attention sink paper) strikes back again? 😅 -- Caveat: it is in post-training setting.

Simple beats complicated: We show that switching to a sliding-window attention mask with attention sinks (at no cost) beats linear attention post-training. Huge thanks to my collaborators @RheaSukthanker, @CameronPashmina, and @Emy_Aze. Paper: arxiv.org/abs/2608.28444