Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Diffusion models are an amazing tool for cofolding, they allow us to predict a protein and the molecule bound to it at once. But they are not exactly fast and require a lot of denoising steps to get accurate predictions. So we distilled ours. Meet DeCAF-Pearl: the first flow... map model for all-atom cofolding. Instead of inching along the denoising trajectory, a flow map learns to jump across it. DeCAF-Pearl runs structure generation ~5x faster than Pearl, our SOTA model, while still maintaining the performance of the teacher model. That speed up allows us to run larger experiments and generate more synthetic data to improve our models. Getting there meant reparameterizing into noise-level space to stabilize gradients, committing to clean-structure prediction to keep the rigid-alignment loss biomolecules needed, and building DeCAF-Search, one steering algorithm for every compute budget. For more technical details, read out blog post: And the paper:show more

Sergey Edunov

2,223 subscribers

36,985 views • 1 month ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

We release Diamond Maps💎 unlocking accurate and efficient guidance for diffusion models. Our experiments show that our methods scale incredibly well. Excited to see what people will build with this! Accurate guidance has been a notoriously hard problem, but in this work, we’re bringing TWO (!) solutions to the table. The recipe for success: 1️⃣ Speed: Use distilled models (flow maps, mean flows, consistency models). 2️⃣ Exploration: Inject stochasticity to properly explore your search space. Because this fundamentally improves anything using flow matching and diffusion, we see a lot of potential for applications across audio, robotics, molecules, and beyond. Paper: Code: Huge thanks to an amazing team: Douglas Chen, Luca Eyring @ ICML26, Ishin Shah, Giri Anantharaman, Yutong (Kelly) He, Zeynep Akata, Tommi Jaakkola, Nicholas Boffi, and Max Simchowitz. It was awesome bringing this to life together!

We release Diamond Maps💎 unlocking accurate and efficient guidance for diffusion models. Our experiments show that our methods scale incredibly well. Excited to see what people will build with this! Accurate guidance has been a notoriously hard problem, but in this work, we’re bringing TWO (!) solutions to the table. The recipe for success: 1️⃣ Speed: Use distilled models (flow maps, mean flows, consistency models). 2️⃣ Exploration: Inject stochasticity to properly explore your search space. Because this fundamentally improves anything using flow matching and diffusion, we see a lot of potential for applications across audio, robotics, molecules, and beyond. Paper: Code: Huge thanks to an amazing team: Douglas Chen, Luca Eyring @ ICML26, Ishin Shah, Giri Anantharaman, Yutong (Kelly) He, Zeynep Akata, Tommi Jaakkola, Nicholas Boffi, and Max Simchowitz. It was awesome bringing this to life together!

Peter Holderrieth

60,179 views • 3 months ago

1/ We are so excited to unveil the Kite AI Ecosystem Map. Kite AI is only as strong as the ecosystem behind it, and ours has enabled us to quickly become the leading base layer for the agentic web. From Google, Shopify and PayPal to Coinbase 🛡️ and Chainlink, our ecosystem comprises 100+ powerhouses across Web2 and Web3 that are building the next generation of autonomous AI. Learn more about who’s building with us here:

1/ We are so excited to unveil the Kite AI Ecosystem Map. Kite AI is only as strong as the ecosystem behind it, and ours has enabled us to quickly become the leading base layer for the agentic web. From Google, Shopify and PayPal to Coinbase 🛡️ and Chainlink, our ecosystem comprises 100+ powerhouses across Web2 and Web3 that are building the next generation of autonomous AI. Learn more about who’s building with us here:

KITE AI

62,046 views • 10 months ago

A Letter to Our Community: The Road Ahead for Robotics To our Community and Partners, As we step into 2026, our mission at Axis is clearer than ever: Constructing the definitive End-to-End Scaling Layer for Robotics. Our goal is to accelerate the transfer of diverse human intelligence into Robotics General Intelligence (RGI). By owning the critical path of intelligence creation, we are turning the physical limitations of robotics into a scalable, software-driven future. Here is our strategic outlook and roadmap for the year ahead. The Core Thesis: Simulation is the Only Way Out The path to RGI is currently blocked by Data Scarcity, Generalization Fragility, and Hardware Fragmentation. At Axis, we believe Simulation is the only way out. Our Simulation Data Platform and Data Augmentation Engine transform raw data into "Synthetic Gold". Backed by academic milestones like Roboverse, Skill Blending, and GraspVLA, we have proven that pure simulation can achieve the generalization required for the real world. We don’t just collect data; we architect it. The Engine: Why Crypto? We believe RGI should come from all, not a few. Crypto is not just a feature; it is the primitive that powers our entire ecosystem flywheel: - Incentive Mechanism: Democratizing contribution and rewarding the trainers and developers. - Assetization: Turning proprietary data and refined models into liquid, ownable assets. - Verifiable Workflow: We are opening the "Black Box" of AI. By bringing total transparency to the Task Generation → Data Collection → Model Training pipeline, we ensure every byte of intelligence is verifiable, traceable, and secure. 2026 Strategic Deliverables This year, we are committed to delivering three foundational pillars: - The World's Largest Training Dataset for Robots: A robot training set—diverse, high-quality interaction data at an unprecedented scale. - A Robotics Foundation Model: A universal robotic brain trained on our pure simulation and synthetic data, capable of robust cross-embodiment transfer and open-world adaptability. - Evolvable Robot Hardware: Robots deployed with Axis models that autonomously evolve through continuous interaction, turning every deployment into a self-improving node within our RGI network. The Ultimate Vision We are building more than models; we are architecting the Distributed Machine Economy. A future where every dataset, model, and robotic embodiment is a verifiable asset in a global, autonomous network. Thank you for building the future of intelligence with us✌️📷

A Letter to Our Community: The Road Ahead for Robotics To our Community and Partners, As we step into 2026, our mission at Axis is clearer than ever: Constructing the definitive End-to-End Scaling Layer for Robotics. Our goal is to accelerate the transfer of diverse human intelligence into Robotics General Intelligence (RGI). By owning the critical path of intelligence creation, we are turning the physical limitations of robotics into a scalable, software-driven future. Here is our strategic outlook and roadmap for the year ahead. The Core Thesis: Simulation is the Only Way Out The path to RGI is currently blocked by Data Scarcity, Generalization Fragility, and Hardware Fragmentation. At Axis, we believe Simulation is the only way out. Our Simulation Data Platform and Data Augmentation Engine transform raw data into "Synthetic Gold". Backed by academic milestones like Roboverse, Skill Blending, and GraspVLA, we have proven that pure simulation can achieve the generalization required for the real world. We don’t just collect data; we architect it. The Engine: Why Crypto? We believe RGI should come from all, not a few. Crypto is not just a feature; it is the primitive that powers our entire ecosystem flywheel: - Incentive Mechanism: Democratizing contribution and rewarding the trainers and developers. - Assetization: Turning proprietary data and refined models into liquid, ownable assets. - Verifiable Workflow: We are opening the "Black Box" of AI. By bringing total transparency to the Task Generation → Data Collection → Model Training pipeline, we ensure every byte of intelligence is verifiable, traceable, and secure. 2026 Strategic Deliverables This year, we are committed to delivering three foundational pillars: - The World's Largest Training Dataset for Robots: A robot training set—diverse, high-quality interaction data at an unprecedented scale. - A Robotics Foundation Model: A universal robotic brain trained on our pure simulation and synthetic data, capable of robust cross-embodiment transfer and open-world adaptability. - Evolvable Robot Hardware: Robots deployed with Axis models that autonomously evolve through continuous interaction, turning every deployment into a self-improving node within our RGI network. The Ultimate Vision We are building more than models; we are architecting the Distributed Machine Economy. A future where every dataset, model, and robotic embodiment is a verifiable asset in a global, autonomous network. Thank you for building the future of intelligence with us✌️📷

Axis Robotics

27,858 views • 7 months ago

DimensionX: Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion TL;DR: Create 3/4DGS from Video Diffusion Note: Some first inference code released (not all yet). Contributions (cited): • We present DimensionX, a novel framework for generating photorealistic 3D and 4D scenes from only a single image using controllable video diffusion. • We propose ST-Director, which decouples the spatial and temporal priors in video diffusion models by learning (spatial and temporal) dimension-aware modules with our curated datasets. We further enhance the hybriddimension control with a training-free composition approach according to the essence of video diffusion denoising process. • To bridge the gap between video diffusion and real-world scenes, we design a trajectory-aware mechanism for 3D generation and an identity-preserving denoising approach for 4D generation, enabling more realistic and controllable scene synthesis. • Extensive experiments manifest that our DimensionX delivers superior performance in video, 3D, and 4D generation compared with baseline methods.

DimensionX: Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion TL;DR: Create 3/4DGS from Video Diffusion Note: Some first inference code released (not all yet). Contributions (cited): • We present DimensionX, a novel framework for generating photorealistic 3D and 4D scenes from only a single image using controllable video diffusion. • We propose ST-Director, which decouples the spatial and temporal priors in video diffusion models by learning (spatial and temporal) dimension-aware modules with our curated datasets. We further enhance the hybriddimension control with a training-free composition approach according to the essence of video diffusion denoising process. • To bridge the gap between video diffusion and real-world scenes, we design a trajectory-aware mechanism for 3D generation and an identity-preserving denoising approach for 4D generation, enabling more realistic and controllable scene synthesis. • Extensive experiments manifest that our DimensionX delivers superior performance in video, 3D, and 4D generation compared with baseline methods.

MrNeRF

17,052 views • 1 year ago

It’s more than a little daunting to set out to expand and improve the identity system for a company and brand like Stripe. But we knew we had to — the existing one had served us well, but wasn’t up to the task anymore. Our brand system required new and improved tools to scale with our ever growing audiences, new products, global footprint, and more. This update introduces material improvements to infographics, advertising, type styles, and more. While the wordmark remains unchanged, we’re using the dot of the ‘i’ (called the “tittle”), a parallelogram pointing up and to the right, to serve as our identifying symbol. We’re also using it as an ever evolving storytelling device to use when talking about our many great users (you can see the latest brand campaign in SF and NYC doing just that). Anyone who has ever worked on the refresh and expansion of an existing system for a large company knows that it is no small endeavor. Crafting impactful solutions, building alignment, creating extensible guidelines, building toolkits, and orchestrating rollout requires a ton of resilience. Here’s to the team that continually inspires me with their dedication, rigor, taste, and exceptional vibes. Great work and thank you to the Brand Studio folks, and of course our many many amazing and invaluable friends and collaborators across the company who all helped shape the work. And a special thank you to a handful of creative agencies that helped us along the way.

Michael Jeter

11,072 views • 9 months ago

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

MrNeRF

27,428 views • 1 year ago

we sped up distributed inference by up to 5x with decentralized speculative decoding. many don't realize that AI models normally generate text one single word at a time, waiting for the network after every word. speculative decoding changes this by using a "guess & confirm" system, similar to autocomplete. how it's done: 1. draft locally (the guess) instead of waiting for the network, a tiny, fast model on your device guesses the next few words instantly, without waiting for the network. 2. confirm remotely (the check) the massive remote model doesn't generate from scratch; it just checks the draft. it looks at the guesses in a batch and says "yes, yes, no." you get multiple words in the time it usually takes to get one. 3. adaptive logic dsd is smart. if the topic is creative, it lets the draft flow loose. if the topic is math or code, it checks more strictly. it balances speed and precision automatically so your inference almost feel instant. find out more: paper: blog:

we sped up distributed inference by up to 5x with decentralized speculative decoding. many don't realize that AI models normally generate text one single word at a time, waiting for the network after every word. speculative decoding changes this by using a "guess & confirm" system, similar to autocomplete. how it's done: 1. draft locally (the guess) instead of waiting for the network, a tiny, fast model on your device guesses the next few words instantly, without waiting for the network. 2. confirm remotely (the check) the massive remote model doesn't generate from scratch; it just checks the draft. it looks at the guesses in a batch and says "yes, yes, no." you get multiple words in the time it usually takes to get one. 3. adaptive logic dsd is smart. if the topic is creative, it lets the draft flow loose. if the topic is math or code, it checks more strictly. it balances speed and precision automatically so your inference almost feel instant. find out more: paper: blog:

Parallax

45,584 views • 7 months ago

AI Is Moving Beyond “Generating Videos” — Toward “Generating Worlds” Over the past two years, AI video models have advanced at an astonishing pace. From Runway and Pika to Sora and Veo, AI-generated videos have become increasingly realistic and more consistent with the physical laws of the real world. Many people believe the next objective is simply to generate videos that are longer, sharper, and more lifelike. But if we take a step back, we can see that the real transformation is not happening in video itself. It is happening in world models. What Is a World Model? In 1943, psychologist Kenneth Craik proposed an idea that would influence artificial intelligence research for decades. He argued that the human brain does not merely react to the outside world. Instead, it maintains an internal model of how the world works. Because we have this internal model, we can predict the outcome of an action before we actually take it. Before crossing a road, we estimate whether a car will pass by. Before catching a ball, we predict its trajectory. These abilities come from continuously simulating the world in our minds, rather than relying entirely on trial and error. This idea later became known by a more formal term: World Model. A world model does not describe a single image or a fixed video clip. It is an internal representation capable of continuously simulating the rules and dynamics of the real world. Why Is AI Research Turning Toward World Models? Because predicting “what comes next” is becoming increasingly central to how AI systems work. Language models predict the next token. Image models predict the next step in the denoising process. Video models predict the next frame. A world model, however, attempts to predict something broader: What should the world look like in the next moment? In 2018, David Ha and Jürgen Schmidhuber proposed in their paper World Models that an intelligent agent could first learn a model of the world, and then use that internal model to plan its actions. The Dreamer series later demonstrated that many complex tasks could be learned by training agents inside an “imagined world.” At the same time, the development of video models such as Sora and Veo led researchers to another realization: A model capable of continuously generating video has already learned, at least implicitly, many of the rules governing the real world. As a result, these two research directions have gradually begun to converge. But Video Is Not Yet a World This is where the distinction is often misunderstood. For a world model to support meaningful real-time interaction, it must solve several critical problems. Most video models today are essentially answering one question: What should the next frame look like? A true world model needs to answer much more: What happens if I take one step forward? If I walk behind a building and then return, will the building still be there? If I suddenly change the camera angle, will the entire space remain consistent? If I enter a command such as: “Summon a dragon.” Will the world respond immediately? In other words, a world model must do more than generate content. It must understand space. It must understand time. It must understand causality. And it must understand interaction. Moving from watching to participating is where the real difficulty of world models begins. World Models Are Entering the Interactive Era One of the latest attempts in this direction is Alaya World, recently open-sourced by Alaya World, or Alaya Lab. Instead of generating a fixed video clip, it generates a world that users can explore in real time. Users can begin with text, an image, or a video, enter the generated scene, move freely through it, and introduce new prompts at any moment during generation. The world responds immediately. According to the publicly released information, Alaya World provides: Real-time streaming generation at 720p and 24 FPS Stable continuous exploration for more than one minute The ability to switch prompts and trigger skills or events during generation Model weights and inference code released under the Apache 2.0 License Training code and datasets planned for future release What makes these capabilities important is not simply the technical specifications. It is that the generated “world” can now support continuous interaction. The official demo shows that users can genuinely control, transform, and explore the generated environment. AI Is Evolving From a Tool Into an Environment Over the past few years, most discussions around AI have focused on content generation. Generating text. Generating images. Generating videos. But world models raise a fundamentally different question: Can AI generate an environment that people can inhabit, explore, and continuously evolve? If the answer is yes, the impact will extend far beyond video generation. Game development, robotics training, embodied intelligence, digital twins, virtual production, and many other fields could be transformed by the development of world models. World models are still at a very early stage. Yet from Craik’s proposal of an internal mental model more than eighty years ago to the emergence of today’s interactive world-generation systems, a clear evolutionary path is beginning to take shape. Perhaps what AI is ultimately learning has never been limited to images, videos, or language. Perhaps it is learning the world itself. References GitHub: Technical Report:

AI Is Moving Beyond “Generating Videos” — Toward “Generating Worlds” Over the past two years, AI video models have advanced at an astonishing pace. From Runway and Pika to Sora and Veo, AI-generated videos have become increasingly realistic and more consistent with the physical laws of the real world. Many people believe the next objective is simply to generate videos that are longer, sharper, and more lifelike. But if we take a step back, we can see that the real transformation is not happening in video itself. It is happening in world models. What Is a World Model? In 1943, psychologist Kenneth Craik proposed an idea that would influence artificial intelligence research for decades. He argued that the human brain does not merely react to the outside world. Instead, it maintains an internal model of how the world works. Because we have this internal model, we can predict the outcome of an action before we actually take it. Before crossing a road, we estimate whether a car will pass by. Before catching a ball, we predict its trajectory. These abilities come from continuously simulating the world in our minds, rather than relying entirely on trial and error. This idea later became known by a more formal term: World Model. A world model does not describe a single image or a fixed video clip. It is an internal representation capable of continuously simulating the rules and dynamics of the real world. Why Is AI Research Turning Toward World Models? Because predicting “what comes next” is becoming increasingly central to how AI systems work. Language models predict the next token. Image models predict the next step in the denoising process. Video models predict the next frame. A world model, however, attempts to predict something broader: What should the world look like in the next moment? In 2018, David Ha and Jürgen Schmidhuber proposed in their paper World Models that an intelligent agent could first learn a model of the world, and then use that internal model to plan its actions. The Dreamer series later demonstrated that many complex tasks could be learned by training agents inside an “imagined world.” At the same time, the development of video models such as Sora and Veo led researchers to another realization: A model capable of continuously generating video has already learned, at least implicitly, many of the rules governing the real world. As a result, these two research directions have gradually begun to converge. But Video Is Not Yet a World This is where the distinction is often misunderstood. For a world model to support meaningful real-time interaction, it must solve several critical problems. Most video models today are essentially answering one question: What should the next frame look like? A true world model needs to answer much more: What happens if I take one step forward? If I walk behind a building and then return, will the building still be there? If I suddenly change the camera angle, will the entire space remain consistent? If I enter a command such as: “Summon a dragon.” Will the world respond immediately? In other words, a world model must do more than generate content. It must understand space. It must understand time. It must understand causality. And it must understand interaction. Moving from watching to participating is where the real difficulty of world models begins. World Models Are Entering the Interactive Era One of the latest attempts in this direction is Alaya World, recently open-sourced by Alaya World, or Alaya Lab. Instead of generating a fixed video clip, it generates a world that users can explore in real time. Users can begin with text, an image, or a video, enter the generated scene, move freely through it, and introduce new prompts at any moment during generation. The world responds immediately. According to the publicly released information, Alaya World provides: Real-time streaming generation at 720p and 24 FPS Stable continuous exploration for more than one minute The ability to switch prompts and trigger skills or events during generation Model weights and inference code released under the Apache 2.0 License Training code and datasets planned for future release What makes these capabilities important is not simply the technical specifications. It is that the generated “world” can now support continuous interaction. The official demo shows that users can genuinely control, transform, and explore the generated environment. AI Is Evolving From a Tool Into an Environment Over the past few years, most discussions around AI have focused on content generation. Generating text. Generating images. Generating videos. But world models raise a fundamentally different question: Can AI generate an environment that people can inhabit, explore, and continuously evolve? If the answer is yes, the impact will extend far beyond video generation. Game development, robotics training, embodied intelligence, digital twins, virtual production, and many other fields could be transformed by the development of world models. World models are still at a very early stage. Yet from Craik’s proposal of an internal mental model more than eighty years ago to the emergence of today’s interactive world-generation systems, a clear evolutionary path is beginning to take shape. Perhaps what AI is ultimately learning has never been limited to images, videos, or language. Perhaps it is learning the world itself. References GitHub: Technical Report:

雪踏乌云

112,114 views • 21 days ago

Our first test flight is just the beginning! Behind the scenes, we are focused on up-scaling and improving our technology. We are excited to announce that we have successfully tested the central subsystem of our Helix 2.0 oxygen-rich staged-combustion engine: the powerpack. We have performed two successful hot-fire tests in which we have shown steady-state operation and cavitation limits. The powerpack incorporates the turbopump and pre-burner(s). It is the most complex as well as the most mechanically and thermally stressed subsystem of a staged-combustion engine. This milestone validated key technological challenges, such as the simultaneous ignition of multiple pre-burners and turbopump cavitation performance. The results are in-line with the predictions from our design models. The closed-cycle architecture of Helix allows us to push the performance envelope further: Helix 2.0 is designed to deliver double the thrust (200kN), while mass, production technology and costs remain comparable to Helix 1.0. The result for our customers: more payload for a lower budget! Excited about this news? Check out our career portal for employment opportunities and help us to elevate our Helix staged-combustion engine technology to the next level! ➡️

Our first test flight is just the beginning! Behind the scenes, we are focused on up-scaling and improving our technology. We are excited to announce that we have successfully tested the central subsystem of our Helix 2.0 oxygen-rich staged-combustion engine: the powerpack. We have performed two successful hot-fire tests in which we have shown steady-state operation and cavitation limits. The powerpack incorporates the turbopump and pre-burner(s). It is the most complex as well as the most mechanically and thermally stressed subsystem of a staged-combustion engine. This milestone validated key technological challenges, such as the simultaneous ignition of multiple pre-burners and turbopump cavitation performance. The results are in-line with the predictions from our design models. The closed-cycle architecture of Helix allows us to push the performance envelope further: Helix 2.0 is designed to deliver double the thrust (200kN), while mass, production technology and costs remain comparable to Helix 1.0. The result for our customers: more payload for a lower budget! Excited about this news? Check out our career portal for employment opportunities and help us to elevate our Helix staged-combustion engine technology to the next level! ➡️

Rocket Factory Augsburg

34,499 views • 3 months ago

Chop the gradients ✂️! We found that truncating decoder gradients in latent video diffusion to a fixed window allows us to finetune on videos with pixel-wise perceptual losses without running out of memory. Pixel losses have been essential for image generation and reconstruction, but until now, they haven't scaled to long-duration, high-resolution video diffusion due to recursive activation accumulation in causal decoders, leading to OOM during training 💥📉. Project: Video diffusion models can do a lot more 🚀 when you can backprop the decoder! Post-process neural rendered scenes, super-resolve videos, harmonize lighting in controlled synthetic driving scenes, and inpaint videos — all in a single step ⚡ with a quick finetune from a standard diffusion model.

Chop the gradients ✂️! We found that truncating decoder gradients in latent video diffusion to a fixed window allows us to finetune on videos with pixel-wise perceptual losses without running out of memory. Pixel losses have been essential for image generation and reconstruction, but until now, they haven't scaled to long-duration, high-resolution video diffusion due to recursive activation accumulation in causal decoders, leading to OOM during training 💥📉. Project: Video diffusion models can do a lot more 🚀 when you can backprop the decoder! Post-process neural rendered scenes, super-resolve videos, harmonize lighting in controlled synthetic driving scenes, and inpaint videos — all in a single step ⚡ with a quick finetune from a standard diffusion model.

Felix Heide

28,399 views • 3 months ago

🚀 We introduce Neural Theorizer (NEO) — a new type of world model that learns to theorize the world from observation, without language or LLM supervision. Selected as an ICML 2026 oral presentation — 0.7% of submitted papers. The paper asks: "What does it mean to understand the world and build a world model?" Today’s world models are often trained to predict the future: the next frame, next latent state, or next observation. But is prediction enough? We argue that a world model should be a theory-building system: one that discovers reusable primitives, composes them into executable explanations, and transfers those explanations to novel phenomena. NEO is our first step toward this vision — a World Theory Model that learns explicit, compositional theories from raw observation. This work was led by my wonderful students: Doojin Baek*(Doojin Baek), Gyubin Lee* (GyuBin Lee), Junyeob Baek (Junyeob Baek), and Hosung Lee (Hosung Lee). For more details, take a look at the paper — and if you’re attending ICML, let’s talk there! 📄 arXiv: 🌐 Project page:

🚀 We introduce Neural Theorizer (NEO) — a new type of world model that learns to theorize the world from observation, without language or LLM supervision. Selected as an ICML 2026 oral presentation — 0.7% of submitted papers. The paper asks: "What does it mean to understand the world and build a world model?" Today’s world models are often trained to predict the future: the next frame, next latent state, or next observation. But is prediction enough? We argue that a world model should be a theory-building system: one that discovers reusable primitives, composes them into executable explanations, and transfers those explanations to novel phenomena. NEO is our first step toward this vision — a World Theory Model that learns explicit, compositional theories from raw observation. This work was led by my wonderful students: Doojin Baek(Doojin Baek), Gyubin Lee (GyuBin Lee), Junyeob Baek (Junyeob Baek), and Hosung Lee (Hosung Lee). For more details, take a look at the paper — and if you’re attending ICML, let’s talk there! 📄 arXiv: 🌐 Project page:

Sungjin Ahn

98,596 views • 1 month ago

I’m so excited to share that Cribl has closed an oversubscribed $319M Series E round at a new $3.5B valuation, up 40% from our Series D just two years ago. The round was led by GV in one of their largest ever investments, with participation from GIC, CapitalG , IVP , and CRV . But that’s not all. GV partner and former GitLab CRO Michael McBride (Michael McBride) is joining our board of directors. This is an incredible milestone for us and it goes to show that solving real problems for real users works really well! Since day 1, Cribl was built to help you unlock the value of all your IT and security data. We’re taking a different approach to data management, with vendor-agnostic products that resolve the tension between data growth (28% CAGR and growing!) and budget growth while giving you choice, control, and flexibility. We’re here for the critically important but often underserved IT and security practitioners who keep businesses running. And we’ll continue to innovate to meet your needs. Thank you to our customers, partners, investors, and all of the amazing goats who have supported and trusted us during this journey. Here’s to more incredible moments as we continue to succeed together! Read more about the announcement in my blog post:

I’m so excited to share that Cribl has closed an oversubscribed $319M Series E round at a new $3.5B valuation, up 40% from our Series D just two years ago. The round was led by GV in one of their largest ever investments, with participation from GIC, CapitalG , IVP , and CRV . But that’s not all. GV partner and former GitLab CRO Michael McBride (Michael McBride) is joining our board of directors. This is an incredible milestone for us and it goes to show that solving real problems for real users works really well! Since day 1, Cribl was built to help you unlock the value of all your IT and security data. We’re taking a different approach to data management, with vendor-agnostic products that resolve the tension between data growth (28% CAGR and growing!) and budget growth while giving you choice, control, and flexibility. We’re here for the critically important but often underserved IT and security practitioners who keep businesses running. And we’ll continue to innovate to meet your needs. Thank you to our customers, partners, investors, and all of the amazing goats who have supported and trusted us during this journey. Here’s to more incredible moments as we continue to succeed together! Read more about the announcement in my blog post:

Clint Sharp

13,981 views • 1 year ago

This is my "feel the AGI" moment: I used GPT-5.6 Sol to train my own autocorrect model that outperforms GPT-5.6 Sol (wtf??) I have no ML background. I have no idea what I'm doing. I just kept pushing Sol until it spat out a SOTA model. And I spent $0. The motivation: Years of talking to AI have made me terrible at typing. Rather than fix my skill issue, I decided to throw more AI at it. My idea was: instead of autocorrect that interrupts my flow, I want to type fast with mistakes and have AI clean it up after. I wanted the smallest local model possible, for speed, for battery life, for science! So I decided to train my own. Inspired by Andrej Karpathy’s autoresearch, I ran Codex /goal with this setup: pick an experiment, try it, record the results to a doc, throw it out if it fails, and plan the next experiment without repeating failures. I gave a few examples that had to pass, tight latency targets, and let it run. Sol did some amazing things. First, it scanned benchmarks and shortlisted base models: Qwen 3.5, Gemma 4, Liquid LFM 2.5. It found a dataset on HuggingFace for typed text. Then it built a simulator for fingers striking a Mac keyboard, modeling the physical layout with a Gaussian distribution around each key. It simulated striking the wrong key, wrong order, fat-fingering, etc. With the models + data + simulator, it fine-tuned using MLX right on my MacBook. It had a working prototype within an hour! But accuracy was pretty poor. — Problem 1: Tokenization Sol read papers, ran tests, and identified that the tokenizer was the bottleneck. Tokenization makes typos hard for the model to see, so it memorizes mappings instead of using its language priors. Sol tried ByT5, Google’s tokenizer-free byte-level LLM. This made a big improvement, but the model is old and lacked the knowledge needed to reach Sol performance. Sol dug deeper and realized a tokenizer-free model isn’t needed; instead, it used T5Gemma, an encoder-decoder model. This can understand the input deeply before producing output, and furthermore, Sol could post-train the encoder to improve performance. This gave a much higher ceiling. — Problem 2: Loss function Now the model was correcting some typos perfectly, but ignoring most. Sol realized that standard cross-entropy loss was teaching the model to avoid edits, because the vast majority of characters in the training data were left unmodified. The fix was wild: Sol wrote a custom loss function that byte-aligns the source and target strings, uses a dynamic programming algorithm to compute the minimum edits between the two, then weights correct edits much higher than copies. After a lot of tuning, this dramatically improved accuracy. — Problem 3: Autoregression One failure mode remained: if the model made a mistake, it couldn’t backtrack. It could only predict the next token. Teaching it to “think” like a reasoning model would solve this, but would be far too slow. Sol found a beautiful solution: instead of greedily predicting the next token, beam search over all possibilities. This parallelizes the exploration instead of one linear chain-of-thought. At the end, choose the path with highest cumulative log probability. This worked great, but made the experience worse, since the user wouldn’t see progress until the whole search was done. To fix this, Sol made a clever observation: after each search step, the longest common prefix among surviving branches is guaranteed to appear in the final result, so it can be displayed immediately. As the search progresses, weaker paths are dropped and the prefix grows, so the user sees continuous progress. Sol built all this as a custom MLX pipeline that does the parallel decoding on the MacBook GPU, with just ~40ms TTFT. It’s crazy fast and entirely local. — Final eval (error reduction rate, higher is better): - Apple autocorrect: 49.66% - GPT-5.6 Luna: 82.47% - GPT-5.6 Terra: 87.64% - GPT-5.6 Sol: 90.56% - Our model (1.7B): 91.02% Final cost: - 1 quota reset (thanks Tibo) - $0 (And yes, I verified there's no cheating. In fact, we test words scrubbed from the training data to prove the model isn’t memorizing) There were a ton more details and tangents I could write about: contrastive learning, GRPO, DPO, dynamic masking, and more. Sol is a fascinating and creative model. It blew my mind so many times. Don’t let a lack of experience stop you: Sol makes AI experiments accessible to anyone!

This is my "feel the AGI" moment: I used GPT-5.6 Sol to train my own autocorrect model that outperforms GPT-5.6 Sol (wtf??) I have no ML background. I have no idea what I'm doing. I just kept pushing Sol until it spat out a SOTA model. And I spent $0. The motivation: Years of talking to AI have made me terrible at typing. Rather than fix my skill issue, I decided to throw more AI at it. My idea was: instead of autocorrect that interrupts my flow, I want to type fast with mistakes and have AI clean it up after. I wanted the smallest local model possible, for speed, for battery life, for science! So I decided to train my own. Inspired by Andrej Karpathy’s autoresearch, I ran Codex /goal with this setup: pick an experiment, try it, record the results to a doc, throw it out if it fails, and plan the next experiment without repeating failures. I gave a few examples that had to pass, tight latency targets, and let it run. Sol did some amazing things. First, it scanned benchmarks and shortlisted base models: Qwen 3.5, Gemma 4, Liquid LFM 2.5. It found a dataset on HuggingFace for typed text. Then it built a simulator for fingers striking a Mac keyboard, modeling the physical layout with a Gaussian distribution around each key. It simulated striking the wrong key, wrong order, fat-fingering, etc. With the models + data + simulator, it fine-tuned using MLX right on my MacBook. It had a working prototype within an hour! But accuracy was pretty poor. — Problem 1: Tokenization Sol read papers, ran tests, and identified that the tokenizer was the bottleneck. Tokenization makes typos hard for the model to see, so it memorizes mappings instead of using its language priors. Sol tried ByT5, Google’s tokenizer-free byte-level LLM. This made a big improvement, but the model is old and lacked the knowledge needed to reach Sol performance. Sol dug deeper and realized a tokenizer-free model isn’t needed; instead, it used T5Gemma, an encoder-decoder model. This can understand the input deeply before producing output, and furthermore, Sol could post-train the encoder to improve performance. This gave a much higher ceiling. — Problem 2: Loss function Now the model was correcting some typos perfectly, but ignoring most. Sol realized that standard cross-entropy loss was teaching the model to avoid edits, because the vast majority of characters in the training data were left unmodified. The fix was wild: Sol wrote a custom loss function that byte-aligns the source and target strings, uses a dynamic programming algorithm to compute the minimum edits between the two, then weights correct edits much higher than copies. After a lot of tuning, this dramatically improved accuracy. — Problem 3: Autoregression One failure mode remained: if the model made a mistake, it couldn’t backtrack. It could only predict the next token. Teaching it to “think” like a reasoning model would solve this, but would be far too slow. Sol found a beautiful solution: instead of greedily predicting the next token, beam search over all possibilities. This parallelizes the exploration instead of one linear chain-of-thought. At the end, choose the path with highest cumulative log probability. This worked great, but made the experience worse, since the user wouldn’t see progress until the whole search was done. To fix this, Sol made a clever observation: after each search step, the longest common prefix among surviving branches is guaranteed to appear in the final result, so it can be displayed immediately. As the search progresses, weaker paths are dropped and the prefix grows, so the user sees continuous progress. Sol built all this as a custom MLX pipeline that does the parallel decoding on the MacBook GPU, with just ~40ms TTFT. It’s crazy fast and entirely local. — Final eval (error reduction rate, higher is better): - Apple autocorrect: 49.66% - GPT-5.6 Luna: 82.47% - GPT-5.6 Terra: 87.64% - GPT-5.6 Sol: 90.56% - Our model (1.7B): 91.02% Final cost: - 1 quota reset (thanks Tibo) - $0 (And yes, I verified there's no cheating. In fact, we test words scrubbed from the training data to prove the model isn’t memorizing) There were a ton more details and tangents I could write about: contrastive learning, GRPO, DPO, dynamic masking, and more. Sol is a fascinating and creative model. It blew my mind so many times. Don’t let a lack of experience stop you: Sol makes AI experiments accessible to anyone!

Anshu

178,432 views • 23 days ago

Ola recently announced that they are bringing affordable AI to Indian developers. 𝐉𝐚𝐫𝐯𝐢𝐬𝐥𝐚𝐛𝐬 an Indian company has been providing affordable GPUs for developers across the globe since 2020. We are a little known, so I want to share our story here. 𝐖𝐡𝐨 𝐰𝐞 𝐚𝐫𝐞 We are bootstrapped, building from the outskirts of Coimbatore. Started as a small team of 4, from humble backgrounds none from IITs/IIMs. Currently, we are a team of 12+. 𝐖𝐡𝐚𝐭 𝐰𝐞 𝐚𝐜𝐡𝐢𝐞𝐯𝐞𝐝 The cost of hosting GPU servers 4 years back in India was insanely high. We got 2 quotes which charged us Rs. 1.5L for a single server per month. At that cost, it was not practical for us to do the business. So we went to the first principle to build an MVP for a mini data center/server room. For the first few years, we ran all our servers from a room fitted with ACs, a UPS, and a Generator, which experts claimed would not work. As we scaled, we faced the heat of our setup, but by then we accumulated more money than we had. So last year we moved it to a tier 3+ DC near Bangalore. This helped us boost the confidence of our users, as we have redundancy for power, internet, and networking which gives us and our customers a lot of peaceful nights. 𝐖𝐡𝐨 𝐮𝐬𝐞𝐬 𝐉𝐚𝐫𝐯𝐢𝐬𝐥𝐚𝐛𝐬 Developers and artists from across the world have supported us in our journey. Some prominent companies are ZOHO (My inspiration), Weights and Biases, UNC, UpGrad, and many more. 𝐑𝐞𝐯𝐞𝐧𝐮𝐞 We crossed 580K USD in the last financial year, the highest ever in our history. Being bootstrapped, the only way for us to grow is to put all the money back. Our customers are our investors, as a founder I have hardly taken a paycheck for the last 4+ years, since the team also believes in our vision they are happy not taking a fancy cheque. 𝐕𝐢𝐬𝐢𝐨𝐧 As AI evolves, we want to bring the capabilities of AI to users at the lowest prices possible. Being bootstrapped, the only way to survive is to be frugal and disciplined. 𝐇𝐢𝐫𝐢𝐧𝐠 I am proud of our hiring strategy. We hired only freshers to date, and most of our hires do not have a formal degree. They come from rural areas and economically challenged backgrounds. The average age of our new team is 19. They have played an active role in building our V2 of Jarvislabs and improving the product daily. I love to thank everyone for supporting us in our journey. Thanks to Analytics India Magazine, INDIAai, fastai for recognizing us in our early years. If our story resonates with you, Please share our story to inspire others & support our mission. #StartupIndia

Ola recently announced that they are bringing affordable AI to Indian developers. 𝐉𝐚𝐫𝐯𝐢𝐬𝐥𝐚𝐛𝐬 an Indian company has been providing affordable GPUs for developers across the globe since 2020. We are a little known, so I want to share our story here. 𝐖𝐡𝐨 𝐰𝐞 𝐚𝐫𝐞 We are bootstrapped, building from the outskirts of Coimbatore. Started as a small team of 4, from humble backgrounds none from IITs/IIMs. Currently, we are a team of 12+. 𝐖𝐡𝐚𝐭 𝐰𝐞 𝐚𝐜𝐡𝐢𝐞𝐯𝐞𝐝 The cost of hosting GPU servers 4 years back in India was insanely high. We got 2 quotes which charged us Rs. 1.5L for a single server per month. At that cost, it was not practical for us to do the business. So we went to the first principle to build an MVP for a mini data center/server room. For the first few years, we ran all our servers from a room fitted with ACs, a UPS, and a Generator, which experts claimed would not work. As we scaled, we faced the heat of our setup, but by then we accumulated more money than we had. So last year we moved it to a tier 3+ DC near Bangalore. This helped us boost the confidence of our users, as we have redundancy for power, internet, and networking which gives us and our customers a lot of peaceful nights. 𝐖𝐡𝐨 𝐮𝐬𝐞𝐬 𝐉𝐚𝐫𝐯𝐢𝐬𝐥𝐚𝐛𝐬 Developers and artists from across the world have supported us in our journey. Some prominent companies are ZOHO (My inspiration), Weights and Biases, UNC, UpGrad, and many more. 𝐑𝐞𝐯𝐞𝐧𝐮𝐞 We crossed 580K USD in the last financial year, the highest ever in our history. Being bootstrapped, the only way for us to grow is to put all the money back. Our customers are our investors, as a founder I have hardly taken a paycheck for the last 4+ years, since the team also believes in our vision they are happy not taking a fancy cheque. 𝐕𝐢𝐬𝐢𝐨𝐧 As AI evolves, we want to bring the capabilities of AI to users at the lowest prices possible. Being bootstrapped, the only way to survive is to be frugal and disciplined. 𝐇𝐢𝐫𝐢𝐧𝐠 I am proud of our hiring strategy. We hired only freshers to date, and most of our hires do not have a formal degree. They come from rural areas and economically challenged backgrounds. The average age of our new team is 19. They have played an active role in building our V2 of Jarvislabs and improving the product daily. I love to thank everyone for supporting us in our journey. Thanks to Analytics India Magazine, INDIAai, fastai for recognizing us in our early years. If our story resonates with you, Please share our story to inspire others & support our mission. #StartupIndia

Vishnu - Jarvislabs.ai

67,646 views • 2 years ago

Getting compliance costs down for builders was a big focus of our Budget and we’re making access to building standards free. Instead of paying up to $1600 a year to access the standards, tradies will get them for free. That means more money in the pockets of tradies and small businesses and higher standards all round. Big thanks to James and the team at BM Sydney Materials for welcoming us in Cabramatta. Anne Stanley MP

Getting compliance costs down for builders was a big focus of our Budget and we’re making access to building standards free. Instead of paying up to $1600 a year to access the standards, tradies will get them for free. That means more money in the pockets of tradies and small businesses and higher standards all round. Big thanks to James and the team at BM Sydney Materials for welcoming us in Cabramatta. Anne Stanley MP

Jim Chalmers MP

25,782 views • 2 months ago

China is smart. China learned from the US's successes, using only the best parts. It's our turn. We should be learning from China. China's economy is growing three times faster than that of the USA. China leads in almost all areas of science, and has the best infrastructure the world has ever seen. The number one priority of China is tell help their people become more prosperous, especially the least well off. Is there still work to be done. Hell yeah. But there difference is that they are trying and we are not. I've been to hundreds of Chinese villages, for more than a decade, and people are becoming better off each year. They call it 'Socialism with Chinese Characteristics.' We don't need to replicate their system. But we should be learning from it. For those ignoring it, you are hurting America. Pretending China's model isn't successful is preventing the US from improving and learning. The China-Haters should be ashamed of the damage that they are doing to the West. Instead of slander and Sinophobia, we need to be learning from the successes of other nations and their peoples.

China is smart. China learned from the US's successes, using only the best parts. It's our turn. We should be learning from China. China's economy is growing three times faster than that of the USA. China leads in almost all areas of science, and has the best infrastructure the world has ever seen. The number one priority of China is tell help their people become more prosperous, especially the least well off. Is there still work to be done. Hell yeah. But there difference is that they are trying and we are not. I've been to hundreds of Chinese villages, for more than a decade, and people are becoming better off each year. They call it 'Socialism with Chinese Characteristics.' We don't need to replicate their system. But we should be learning from it. For those ignoring it, you are hurting America. Pretending China's model isn't successful is preventing the US from improving and learning. The China-Haters should be ashamed of the damage that they are doing to the West. Instead of slander and Sinophobia, we need to be learning from the successes of other nations and their peoples.

Jason Smith - 上官杰文

82,781 views • 1 year ago

We released physics-intern: a simple harness for science problems! It gets models like Gemini 3.1 Pro to go from 17.7 -> 31.4, thus beating GPT 5.5 Pro. The physics-intern harness can wrap any model and via dedicated subagent boost the performance of the vanilla reasoning models. While I think more and more of these harness capability gains will be absorbed into the models (like prompting tricks disappeared over time) there is a lot to be gained right now by building good scaffolds for those models and integrating tools well. Interestingly, the exception we found that GPT 5.5 Pro actually didn't benefit from the physics-intern harness! Read more about it here: PS: I think the Harness[Model] notation is kind of nice.

We released physics-intern: a simple harness for science problems! It gets models like Gemini 3.1 Pro to go from 17.7 -> 31.4, thus beating GPT 5.5 Pro. The physics-intern harness can wrap any model and via dedicated subagent boost the performance of the vanilla reasoning models. While I think more and more of these harness capability gains will be absorbed into the models (like prompting tricks disappeared over time) there is a lot to be gained right now by building good scaffolds for those models and integrating tools well. Interestingly, the exception we found that GPT 5.5 Pro actually didn't benefit from the physics-intern harness! Read more about it here: PS: I think the Harness[Model] notation is kind of nice.

Leandro von Werra

97,262 views • 2 months ago

Help us make Pathfinder even better! The more we talk to our community, the more it validates our belief. Everyone’s favorite shapes are unique, even when they look similar, size makes all the difference! But we want to understand more and to further the science of shape. Our Discord data shows that Xtra Claw, Right Hand, and Wave shapes are among the most popular. Now we want to reach out to mroe people and hear from you! Like ❤️ / Retweet 🔁 this post and complete the survey📋. In one week, we’ll pick a winner 🎁 to be the very first to receive our secret gift box! 🎉

Help us make Pathfinder even better! The more we talk to our community, the more it validates our belief. Everyone’s favorite shapes are unique, even when they look similar, size makes all the difference! But we want to understand more and to further the science of shape. Our Discord data shows that Xtra Claw, Right Hand, and Wave shapes are among the most popular. Now we want to reach out to mroe people and hear from you! Like ❤️ / Retweet 🔁 this post and complete the survey📋. In one week, we’ll pick a winner 🎁 to be the very first to receive our secret gift box! 🎉

Orbital

11,612 views • 10 months ago

An interesting issue with Tesla Robotaxi where it took us to a Starbucks, but the Google data had the incorrect location. At drop off we were 0.2 miles from the Starbucks, so we had a short walk. Is there a way the Tesla AI team could add functionality in the app so riders can update map info to correct errors or inaccuracies on the underlying map data and then this propagates to the fleet? Being able to do this with a pin drop on the map instead of having to use an address might make this very easy and user friendly! Or, as a bigger ask, would it be possible for the car to be able to use visual images on its own to look for a Starbucks sign and on the fly, get us closer and update the map data on its own?

An interesting issue with Tesla Robotaxi where it took us to a Starbucks, but the Google data had the incorrect location. At drop off we were 0.2 miles from the Starbucks, so we had a short walk. Is there a way the Tesla AI team could add functionality in the app so riders can update map info to correct errors or inaccuracies on the underlying map data and then this propagates to the fleet? Being able to do this with a pin drop on the map instead of having to use an address might make this very easy and user friendly! Or, as a bigger ask, would it be possible for the car to be able to use visual images on its own to look for a Starbucks sign and on the fly, get us closer and update the map data on its own?

Joe Tegtmeyer 🚀 🤠🛸😎

80,355 views • 1 year ago

No one is coming to save us There are no political solutions, our heroes are dead and all our enemies are in power If we fail, in 100 years there will be no White people left… have you considered what that means? It means 3rd World conditions for the entire planet It means lawless anarchy 1st World Western Nations offset the 3rd World through aid, charity, and donations Without us, instead of continuing their upward trajectory they would begin to deteriorate fast Our people turn the world from a savage jungle into a place of order There are no 1st World non-white Nations China is still developing in a lot of ways and Korea and Japan have dangerous population declines as well Without the amount of people it takes to run the infrastructure you get an expedited breakdown of society Of course our White children will be hated and blamed for the failures Mobocracy will rule, reason will cease, and we’ll return to the stone ages not because of nuclear war but because White people disappeared

No one is coming to save us There are no political solutions, our heroes are dead and all our enemies are in power If we fail, in 100 years there will be no White people left… have you considered what that means? It means 3rd World conditions for the entire planet It means lawless anarchy 1st World Western Nations offset the 3rd World through aid, charity, and donations Without us, instead of continuing their upward trajectory they would begin to deteriorate fast Our people turn the world from a savage jungle into a place of order There are no 1st World non-white Nations China is still developing in a lot of ways and Korea and Japan have dangerous population declines as well Without the amount of people it takes to run the infrastructure you get an expedited breakdown of society Of course our White children will be hated and blamed for the failures Mobocracy will rule, reason will cease, and we’ll return to the stone ages not because of nuclear war but because White people disappeared

Christian Ascania🌞

70,540 views • 6 months ago