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.

Karpathy told Dwarkesh that a 1 billion parameter model, trained on clean data, could hit the intelligence of today's 1.8 trillion parameter frontier. That is a 1,800x compression claim. The math behind it is more defensible than it sounds. When researchers at frontier labs look at random samples from... their training corpus, they see stock ticker symbols, broken HTML, forum spam, autogenerated gibberish. Not Wikipedia. Not the Wall Street Journal. The actual pretraining dataset is mostly noise, and the model is burning parameters to vaguely remember all of it. One estimate pegs Llama 3's information compression at 0.07 bits per token. Well-structured English carries around 1.5 bits per token of real information. The trillion-parameter model is holding a roughly 5% resolution image of the internet it trained on. So when a lab ships a 1.8 trillion parameter model, the overwhelming majority of those weights are handling rough memorization. They are compression overhead for a noisy training set, taking up capacity that could be doing reasoning instead. Karpathy's proposal is to separate the two. Build a cognitive core: a small model that contains only the algorithms for reasoning and problem-solving, stripped of encyclopedic memorization. Pair it with external memory the model queries when it needs a fact. A 1 billion parameter reasoner plus retrieval beats a 1.8 trillion parameter model trying to do both. The data already supports this direction. GPT-4o runs at roughly 200 billion parameters and outperforms the original 1.8 trillion GPT-4. Inference costs for GPT-3.5 level performance fell 280x between 2022 and 2024, driven almost entirely by smaller, cleaner, better-architected models. The trend line is pointing where Karpathy says it should. The real implication for anyone tracking the AI trade: data quality is the actual constraint. The companies winning the next phase will be the ones who figured out what to train on, and what to throw away.show more

Aakash Gupta

331,103 subscribers

508,540 views • 4 months ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

$Mark Zuckerberg is explaining one of the most misunderstood dynamics in AI and it has direct investment implications (Save this). The concept he's describing is model distillation, and it's one of the most important techniques to emerge in AI over the past year. Here's how it works. You train a massive, enormously expensive model, in Meta's case, Llama 4 Behemoth, a 2 trillion parameter teacher model and then you use that model to teach a much smaller, cheaper model. The smaller model inherits roughly 90 to 95% of the intelligence of the giant while running at 10% of the cost and on a fraction of the compute. Meta already did this with the Llama 4 family and Behemoth serves as the teacher. Llama 4 Scout and Maverick, the publicly released open-source models were distilled from it. Scout runs on a single H100 GPU with a 10 million token context window and outperforms models that cost far more to operate. Maverick, at 17 billion active parameters, rivals DeepSeek V3 in coding at half the parameter count and beats GPT-4o on multimodal benchmarks. Both are completely free for commercial use. What Zuckerberg is pointing at is a structural shift in how AI gets deployed in the real world. Companies aren't taking a frontier model off the shelf and running it as-is but rather taking open-source models, fine-tuning them on their own proprietary data, distilling them into even smaller custom models tailored to their specific use case, and running them on infrastructure they control at a fraction of the cost of a closed frontier API. The investment implication of this is significant and runs in two directions. For Meta specifically, this is a strategic masterstroke. Every company that builds on Llama, fine-tunes it, distills it, or deploys it through their infrastructure is pulling into Meta's orbit while Meta builds the most powerful open teacher model. The ecosystem of companies using it grows and that ecosystem generates commercial activity across Meta's platforms and data services. Meta's AI research benefits from billions of real world deployment signals and it's a flywheel that closed model providers cannot replicate because their strategy requires charging per token, which is now a 65x cost disadvantage against the open-source alternative. For the broader market, distillation changes the economics of inference in a way that has barely been priced in. As intelligence becomes extractable into smaller and cheaper models, the absolute demand for compute doesn't decline but rather it explodes, because now the number of applications that are economically viable expands by orders of magnitude. Every task that was previously too expensive to automate at $3.25 per call becomes viable at $0.05 that means more total token usage, more total GPU utilization, and more demand for the infrastructure companies, the Nebiuses, the GE Vernovas, the Constellation Energies that supply the underlying compute and power.$

Mark Zuckerberg is explaining one of the most misunderstood dynamics in AI and it has direct investment implications (Save this). The concept he's describing is model distillation, and it's one of the most important techniques to emerge in AI over the past year. Here's how it works. You train a massive, enormously expensive model, in Meta's case, Llama 4 Behemoth, a 2 trillion parameter teacher model and then you use that model to teach a much smaller, cheaper model. The smaller model inherits roughly 90 to 95% of the intelligence of the giant while running at 10% of the cost and on a fraction of the compute. Meta already did this with the Llama 4 family and Behemoth serves as the teacher. Llama 4 Scout and Maverick, the publicly released open-source models were distilled from it. Scout runs on a single H100 GPU with a 10 million token context window and outperforms models that cost far more to operate. Maverick, at 17 billion active parameters, rivals DeepSeek V3 in coding at half the parameter count and beats GPT-4o on multimodal benchmarks. Both are completely free for commercial use. What Zuckerberg is pointing at is a structural shift in how AI gets deployed in the real world. Companies aren't taking a frontier model off the shelf and running it as-is but rather taking open-source models, fine-tuning them on their own proprietary data, distilling them into even smaller custom models tailored to their specific use case, and running them on infrastructure they control at a fraction of the cost of a closed frontier API. The investment implication of this is significant and runs in two directions. For Meta specifically, this is a strategic masterstroke. Every company that builds on Llama, fine-tunes it, distills it, or deploys it through their infrastructure is pulling into Meta's orbit while Meta builds the most powerful open teacher model. The ecosystem of companies using it grows and that ecosystem generates commercial activity across Meta's platforms and data services. Meta's AI research benefits from billions of real world deployment signals and it's a flywheel that closed model providers cannot replicate because their strategy requires charging per token, which is now a 65x cost disadvantage against the open-source alternative. For the broader market, distillation changes the economics of inference in a way that has barely been priced in. As intelligence becomes extractable into smaller and cheaper models, the absolute demand for compute doesn't decline but rather it explodes, because now the number of applications that are economically viable expands by orders of magnitude. Every task that was previously too expensive to automate at $3.25 per call becomes viable at $0.05 that means more total token usage, more total GPU utilization, and more demand for the infrastructure companies, the Nebiuses, the GE Vernovas, the Constellation Energies that supply the underlying compute and power.

Milk Road AI

27,908 views • 2 months ago

NVIDIA CEO: GROK 5 IS A 7 TRILLION PARAMETER RACE AGAINST TIME Jensen Huang is dialing in on the real challenge: not making bigger models, but training them fast without draining power or budgets. Grok 5 is right in the middle of that race. “The next frontier model. Elon already mentioned that the next version of Grok, Grok 5 I believe, is 7 trillion parameters. This one is 10, and the green represents Blackwell. In the case of Rubin, notice that the throughput is much higher, so it only takes one fourth as many of these systems to train the model within the one-month timeframe we have given here.” Source: Rohan Paul

NVIDIA CEO: GROK 5 IS A 7 TRILLION PARAMETER RACE AGAINST TIME Jensen Huang is dialing in on the real challenge: not making bigger models, but training them fast without draining power or budgets. Grok 5 is right in the middle of that race. “The next frontier model. Elon already mentioned that the next version of Grok, Grok 5 I believe, is 7 trillion parameters. This one is 10, and the green represents Blackwell. In the case of Rubin, notice that the throughput is much higher, so it only takes one fourth as many of these systems to train the model within the one-month timeframe we have given here.” Source: Rohan Paul

Mario Nawfal

76,356 views • 8 months ago

TWO BOXES THE SIZE OF A MAC MINI JUST RAN A 235 BILLION PARAMETER MODEL ON A DESK It is two NVIDIA DGX Spark units linked by a single cable. A year ago a model this size meant renting a GPU cluster by the hour. Now it sits next to your monitor for around $8,000. Here is the twist most people miss. Linking them does not create one shared 256GB memory pool. The model is split across both boxes, and that is the only reason a 235B model fits at all. It answers at roughly 10 tokens per second, and both chips sit at just 74 degrees while sipping around 50 watts. Every token stays on the desk. Nothing touches a cloud, and nothing leaves the room. The ceiling for what you can run at home just jumped from 70B to 235B. Bookmark this & Watch it run ↓

TWO BOXES THE SIZE OF A MAC MINI JUST RAN A 235 BILLION PARAMETER MODEL ON A DESK It is two NVIDIA DGX Spark units linked by a single cable. A year ago a model this size meant renting a GPU cluster by the hour. Now it sits next to your monitor for around $8,000. Here is the twist most people miss. Linking them does not create one shared 256GB memory pool. The model is split across both boxes, and that is the only reason a 235B model fits at all. It answers at roughly 10 tokens per second, and both chips sit at just 74 degrees while sipping around 50 watts. Every token stays on the desk. Nothing touches a cloud, and nothing leaves the room. The ceiling for what you can run at home just jumped from 70B to 235B. Bookmark this & Watch it run ↓

slash1s

101,317 views • 2 months ago

APPLE RESEARCH SCIENTIST JUST SHOWED HOW 4 MAC STUDIOS RUN A TRILLION PARAMETER MODEL LOCALLY ZERO COSTS 13:18 she shows the main thing - connect 4 Mac Studios and you get 1TB of shared memory - exactly enough to run a trillion parameter model right on your desk Apple's library - and four machines start working as one cluster tensor parallelism: every machine holds part of every layer - all process the same token simultaneously - speed increases 3x compared to a single device fine-tuning: one Mac Studio processes 180 tokens per second four together process 600 and not a single byte of data leaves the room one command in the terminal - and a trillion parameter model answers from your desk and runs 24/7 data centers took years to build to run models like this - Apple did it with four Thunderbolt cables

APPLE RESEARCH SCIENTIST JUST SHOWED HOW 4 MAC STUDIOS RUN A TRILLION PARAMETER MODEL LOCALLY ZERO COSTS 13:18 she shows the main thing - connect 4 Mac Studios and you get 1TB of shared memory - exactly enough to run a trillion parameter model right on your desk Apple's library - and four machines start working as one cluster tensor parallelism: every machine holds part of every layer - all process the same token simultaneously - speed increases 3x compared to a single device fine-tuning: one Mac Studio processes 180 tokens per second four together process 600 and not a single byte of data leaves the room one command in the terminal - and a trillion parameter model answers from your desk and runs 24/7 data centers took years to build to run models like this - Apple did it with four Thunderbolt cables

Noisy

48,763 views • 2 months ago

i'm running a 397 billion parameter model on a amd ai max box that sits on my desk and pulls less power than a gaming laptop. the model is Nex-N2-Pro, 397B-A17B, the open weight release people are putting next to gpt-5.5 on coding. i have it quantized to IQ1_M, 1.75 bits per weight, 90gb of weights loaded into the 128gb of unified memory on amd's strix halo igpu. watch the gpu in this recording. it spikes, it sustains, it does not fall over. that is the part the spec sheets never show you, not just that a 400b model loads, but that an integrated graphics chip holds the load and generates token after token, stable, no crash, no thermal cliff. and it is not a slideshow. roughly 18 tokens a second, faster than you can read. a frontier scale model producing usable output, fully local. no datacenter, no rented h100s, no api key, no permission. three years ago a model this size meant a server room and a budget to match. tonight it is a quiet box on my desk. this is the accessible tier almost nobody benchmarks honestly, and it is further along than the timeline thinks. the full breakdown is coming, rocm vs vulkan on this chip, and this little amd box head to head against the nvidia equivalent. stay tuned.

i'm running a 397 billion parameter model on a amd ai max box that sits on my desk and pulls less power than a gaming laptop. the model is Nex-N2-Pro, 397B-A17B, the open weight release people are putting next to gpt-5.5 on coding. i have it quantized to IQ1_M, 1.75 bits per weight, 90gb of weights loaded into the 128gb of unified memory on amd's strix halo igpu. watch the gpu in this recording. it spikes, it sustains, it does not fall over. that is the part the spec sheets never show you, not just that a 400b model loads, but that an integrated graphics chip holds the load and generates token after token, stable, no crash, no thermal cliff. and it is not a slideshow. roughly 18 tokens a second, faster than you can read. a frontier scale model producing usable output, fully local. no datacenter, no rented h100s, no api key, no permission. three years ago a model this size meant a server room and a budget to match. tonight it is a quiet box on my desk. this is the accessible tier almost nobody benchmarks honestly, and it is further along than the timeline thinks. the full breakdown is coming, rocm vs vulkan on this chip, and this little amd box head to head against the nvidia equivalent. stay tuned.

Sudo su

32,163 views • 2 months ago

The most overlooked part of the SpaceX IPO thesis is the model and most people are completely missing it (Save this) Everyone has been focused on the Anthropic compute deal and the Colossus revenue because those are numbers you can put in a spreadsheet. Six months ago, xAI was competing reasonably well on model performance but was not clearly on the frontier. Then SpaceX exercised its option to acquire Cursor for $60 billion, the largest startup acquisition in history just days after completing the largest IPO in history at $75 billion. Cursor is a team of 700 to 800 people, was on track to exit 2026 at up to $10 billion in revenue, had millions of professional developers using it daily, and had already built a team with the genuine potential to compete at the frontier, the one thing holding them back was compute. SpaceX just gave them the largest GPU cluster in the world to work with. Grok 4.3, a 1.5 trillion parameter model, is currently training with Cursor's proprietary coding data being injected directly into pre-training, not just fine tuning which is a fundamentally more powerful integration than anything the market is currently modeling. The prior version, Grok 4, was already on the Pareto frontier as of 10 to 12 days ago, the most intelligent 500 billion parameter model in the world, sitting alongside Google Gemini, Anthropic, and OpenAI as one of only four systems at the true frontier. Composer 2.5, the previous Cursor model was Pareto dominant in coding tasks just before the acquisition closed, meaning SpaceX inherited a model that was already best-in-class in the highest-value AI use case in the market. The AWS parallel is the one everyone keeps missing. Bezos built data center capacity for Black Friday, sat on idle infrastructure the rest of the year, and monetized it into what was at the time the most profitable technology business in history and investors hated it in 2009 and 2010 because he was burning free cash flow on capacity that had no obvious revenue yet. SpaceX is in exactly that position, it built Colossus for xAI's own training needs, is monetizing excess capacity to Anthropic at $1.25 billion per month across 220,000 Nvidia GPUs, and has reportedly secured up to 20% of Nvidia's early Vera Rubin allocation, giving it the most powerful and scarcest GPU infrastructure in the world during the critical window when those chips are hardest to get. The $60 billion Cursor acquisition closed at a moment when SpaceX had essentially unlimited compute, a team already at the frontier, and a product with deep enterprise distribution, three things no other model lab had simultaneously when it was at this stage. The market is pricing the compute business conservatively and ignoring the model call option entirely, and coding is the fastest path to AGI, once you are on the Pareto frontier with that compute, revenue scales fast. Anthropic went from negligible revenue to $30 billion annualized in under 18 months and that is the existence proof. Bullish on SpaceXAI and Elon Musk

The most overlooked part of the SpaceX IPO thesis is the model and most people are completely missing it (Save this) Everyone has been focused on the Anthropic compute deal and the Colossus revenue because those are numbers you can put in a spreadsheet. Six months ago, xAI was competing reasonably well on model performance but was not clearly on the frontier. Then SpaceX exercised its option to acquire Cursor for $60 billion, the largest startup acquisition in history just days after completing the largest IPO in history at $75 billion. Cursor is a team of 700 to 800 people, was on track to exit 2026 at up to $10 billion in revenue, had millions of professional developers using it daily, and had already built a team with the genuine potential to compete at the frontier, the one thing holding them back was compute. SpaceX just gave them the largest GPU cluster in the world to work with. Grok 4.3, a 1.5 trillion parameter model, is currently training with Cursor's proprietary coding data being injected directly into pre-training, not just fine tuning which is a fundamentally more powerful integration than anything the market is currently modeling. The prior version, Grok 4, was already on the Pareto frontier as of 10 to 12 days ago, the most intelligent 500 billion parameter model in the world, sitting alongside Google Gemini, Anthropic, and OpenAI as one of only four systems at the true frontier. Composer 2.5, the previous Cursor model was Pareto dominant in coding tasks just before the acquisition closed, meaning SpaceX inherited a model that was already best-in-class in the highest-value AI use case in the market. The AWS parallel is the one everyone keeps missing. Bezos built data center capacity for Black Friday, sat on idle infrastructure the rest of the year, and monetized it into what was at the time the most profitable technology business in history and investors hated it in 2009 and 2010 because he was burning free cash flow on capacity that had no obvious revenue yet. SpaceX is in exactly that position, it built Colossus for xAI's own training needs, is monetizing excess capacity to Anthropic at $1.25 billion per month across 220,000 Nvidia GPUs, and has reportedly secured up to 20% of Nvidia's early Vera Rubin allocation, giving it the most powerful and scarcest GPU infrastructure in the world during the critical window when those chips are hardest to get. The $60 billion Cursor acquisition closed at a moment when SpaceX had essentially unlimited compute, a team already at the frontier, and a product with deep enterprise distribution, three things no other model lab had simultaneously when it was at this stage. The market is pricing the compute business conservatively and ignoring the model call option entirely, and coding is the fastest path to AGI, once you are on the Pareto frontier with that compute, revenue scales fast. Anthropic went from negligible revenue to $30 billion annualized in under 18 months and that is the existence proof. Bullish on SpaceXAI and Elon Musk

Milk Road AI

69,744 views • 2 months ago

Baseten's Charlie O'Neill on why the gap to the frontier is just compute and data: "Despite OpenAI and Anthropic and the closed source labs having more compute and more data, the recipe is the same. We're all scaling the same recipe here, and we're at slightly different points of the scaling curve. It's not like OpenAI or Anthropic have any secret sauce that differentiates them scaling up their multi-trillion parameter models from the rest of the open source ecosystem." "It really is just a matter of taking the core underlying architecture and training recipe, and throwing more compute and data at it, which is really promising." "Of course, these closed source labs probably have a long tail of optimizations they've made, which we're gonna continue to try and figure out in the open source community. But the core mechanism is there." "K3 is a great example of, well, if you do scale this recipe up, we're gonna get a pretty good model that's very, very close to the frontier, if not frontier itself." Charlie O'Neill Baseten

Baseten's Charlie O'Neill on why the gap to the frontier is just compute and data: "Despite OpenAI and Anthropic and the closed source labs having more compute and more data, the recipe is the same. We're all scaling the same recipe here, and we're at slightly different points of the scaling curve. It's not like OpenAI or Anthropic have any secret sauce that differentiates them scaling up their multi-trillion parameter models from the rest of the open source ecosystem." "It really is just a matter of taking the core underlying architecture and training recipe, and throwing more compute and data at it, which is really promising." "Of course, these closed source labs probably have a long tail of optimizations they've made, which we're gonna continue to try and figure out in the open source community. But the core mechanism is there." "K3 is a great example of, well, if you do scale this recipe up, we're gonna get a pretty good model that's very, very close to the frontier, if not frontier itself." Charlie O'Neill Baseten

MTS

17,404 views • 1 month ago

China just released an open source AI model that matches the best closed models from OpenAI and Anthropic. Gavin Baker explained exactly how they did it and the answer should concern every American AI lab. The model is called GLM 5.2. It was built by Z. AI. You get 744 billion parameters, 1 million token context window and its MIT license, meaning anyone can download it, fork it, build a company on it, with no restrictions and no Dario. It scored 51 points on the artificial analysis intelligence index. The highest score any open weight model has ever achieved. It beat GPT 5.5 on the frontier software engineering benchmark. It trails Claude Opus 4.8 by less than one percentage point. And it costs 85% less to run than GPT 5.5 for comparable performance. Gavin Baker said on the All-In podcast that this model has challenged some of his beliefs. Then he explained how China built it. The method is called distillation. Just think of tens of thousands of phones and computers running simultaneously, all hitting the frontier model APIs through masked accounts, asking specific questions, and harvesting what happens inside the model when it answers. Every reasoning step, every token. The entire thinking process gets recorded and fed back into the Chinese model during training. It is a cheat sheet. It is the answer key to the exam. And here is the part that should worry everyone. Sacks said it plainly. China was already nine months behind American models. But now that GLM 5.2 is good enough to run its own reinforcement learning, it can improve itself without needing to distill from American models anymore. The cheat sheet let them get close enough to start writing their own answers. Sacks said we are six months behind on the model and 24 months behind on silicon and they are only a few months behind in total. The Z. AI founder told Elon Musk directly that open weight fable-level capability will be here before Q1 2027. Every restriction Anthropic lobbied for, every self-imposed safety guardrail, every month of delay in releasing American frontier models accelerated this. The Chinese labs were not under those restrictions. They were not going to wait. The composable model future Gavin described, where every enterprise runs a frontier model alongside their own fine-tuned open weight model, is coming regardless of what American labs do next. The question is just whether the open weight half of that stack is American or Chinese. Right now it is Chinese. WATCH THE FULL PODCAST ON The All-In Podcast

China just released an open source AI model that matches the best closed models from OpenAI and Anthropic. Gavin Baker explained exactly how they did it and the answer should concern every American AI lab. The model is called GLM 5.2. It was built by Z. AI. You get 744 billion parameters, 1 million token context window and its MIT license, meaning anyone can download it, fork it, build a company on it, with no restrictions and no Dario. It scored 51 points on the artificial analysis intelligence index. The highest score any open weight model has ever achieved. It beat GPT 5.5 on the frontier software engineering benchmark. It trails Claude Opus 4.8 by less than one percentage point. And it costs 85% less to run than GPT 5.5 for comparable performance. Gavin Baker said on the All-In podcast that this model has challenged some of his beliefs. Then he explained how China built it. The method is called distillation. Just think of tens of thousands of phones and computers running simultaneously, all hitting the frontier model APIs through masked accounts, asking specific questions, and harvesting what happens inside the model when it answers. Every reasoning step, every token. The entire thinking process gets recorded and fed back into the Chinese model during training. It is a cheat sheet. It is the answer key to the exam. And here is the part that should worry everyone. Sacks said it plainly. China was already nine months behind American models. But now that GLM 5.2 is good enough to run its own reinforcement learning, it can improve itself without needing to distill from American models anymore. The cheat sheet let them get close enough to start writing their own answers. Sacks said we are six months behind on the model and 24 months behind on silicon and they are only a few months behind in total. The Z. AI founder told Elon Musk directly that open weight fable-level capability will be here before Q1 2027. Every restriction Anthropic lobbied for, every self-imposed safety guardrail, every month of delay in releasing American frontier models accelerated this. The Chinese labs were not under those restrictions. They were not going to wait. The composable model future Gavin described, where every enterprise runs a frontier model alongside their own fine-tuned open weight model, is coming regardless of what American labs do next. The question is just whether the open weight half of that stack is American or Chinese. Right now it is Chinese. WATCH THE FULL PODCAST ON The All-In Podcast

Ihtesham Ali

86,621 views • 2 months ago

Jensen Huang said Grok 5 will be 7 Trillion parameter model. On time to train, with the training window fixed at 1 month, the new Vera-Rubin GPU system of Nvidia needs 1/4 the number of systems compared with Blackwell to train the same frontier model. “factory throughput” improves by about 10x over Blackwell, and Blackwell itself was about 10x over Hopper. i.e. overall Rubin is roughly 100x Hopper in factory throughput per watt, which matters hugely, because a 1 GW, $50B data center is power-limited and revenue scales with throughput per watt. --- From Nvidia YT channel

Jensen Huang said Grok 5 will be 7 Trillion parameter model. On time to train, with the training window fixed at 1 month, the new Vera-Rubin GPU system of Nvidia needs 1/4 the number of systems compared with Blackwell to train the same frontier model. “factory throughput” improves by about 10x over Blackwell, and Blackwell itself was about 10x over Hopper. i.e. overall Rubin is roughly 100x Hopper in factory throughput per watt, which matters hugely, because a 1 GW, $50B data center is power-limited and revenue scales with throughput per watt. --- From Nvidia YT channel

Rohan Paul

165,770 views • 8 months ago

Nvidia has just announced Alpamayo 2 Super, an open 34 billion parameter reasoning vision-language-action model designed to accelerate the development of autonomous vehicles. This new model combines the NVIDIA Cosmos 3 Super reasoning model with a 2 billion parameter diffusion-based action expert model, and is post trained with reinforcement learning. The model can return multiple outputs: future trajectory plans, reasoning traces, grounded answers to questions about the scenes, and auto label generation. The model weights are now available for anyone to download on Hugging Face, and the inference code has been posted to GitHub. Distilled models can be deployed commercially without any further permission from Nvidia, and model outputs carry no license conditions. Automakers can distill down a compact version of this model that can run on the Nvidia computer in the car. Major kudos to Nvidia and Jensen Huang for advancing the state of the industry by releasing this as an open model with permissive licensing. Jensen isn't just paying lip service to the idea of open models, Nvidia is actually contributing to the ecosystem — and it's great for their business, because it helps sell more Thor computers that go in the car. Anyone can go download the model and play with it. If you do, let me know what you think. Personally I think it's so cool that we have open weights models that are this advanced, for anyone to download.

Nvidia has just announced Alpamayo 2 Super, an open 34 billion parameter reasoning vision-language-action model designed to accelerate the development of autonomous vehicles. This new model combines the NVIDIA Cosmos 3 Super reasoning model with a 2 billion parameter diffusion-based action expert model, and is post trained with reinforcement learning. The model can return multiple outputs: future trajectory plans, reasoning traces, grounded answers to questions about the scenes, and auto label generation. The model weights are now available for anyone to download on Hugging Face, and the inference code has been posted to GitHub. Distilled models can be deployed commercially without any further permission from Nvidia, and model outputs carry no license conditions. Automakers can distill down a compact version of this model that can run on the Nvidia computer in the car. Major kudos to Nvidia and Jensen Huang for advancing the state of the industry by releasing this as an open model with permissive licensing. Jensen isn't just paying lip service to the idea of open models, Nvidia is actually contributing to the ecosystem — and it's great for their business, because it helps sell more Thor computers that go in the car. Anyone can go download the model and play with it. If you do, let me know what you think. Personally I think it's so cool that we have open weights models that are this advanced, for anyone to download.

Whole Mars Catalog

45,595 views • 1 month ago

New model: your robot can now pack your suitcase 🧳 Xiaomi has released a new robot foundation model. Called Xiaomi-Robotics-1, it is designed to have a robot pick things up and move them around. But first, DEFINITIONS: - Mixture-of-Transformers (MoT): An architecture where separate transformer "experts" (e.g., one for vision-language, one for actions) share a single attention stream, so each modality gets specialized parameters without losing joint reasoning. - Vision-language model (VLM): A model that jointly understands images and text. - Diffusion transformer: A transformer trained to turn noise into structured outputs by iterative denoising, here generating robot actions rather than images. - Action chunks: Short sequences of future actions (e.g., the next ~50 motor commands) predicted in one shot instead of one step at a time. - Flow matching: A faster version of diffusion. The model learns a straight-line velocity field from noise to the target action, so it needs only a few integration steps instead of many denoising ones. Its peculiarity comes from its two stage training: 1. 100,000 hours of video shot through a UMI rig: a handheld 3D-printed gripper with a camera, worn by humans doing ordinary tasks in homes, shops, factories and offices. 2. Adapt to actual robot bodies with ~10,000 hours of real-robot data. It replaces the standard approach of teleoperating a real robot for every hour of training data. Its architecture is a Mixture-of-Transformers pairing a pre-trained Qwen3-VL vision-language model with a diffusion transformer that emits action chunks via flow matching, released in 2.6B, 5.1B and 10.5B parameter variants. However, if you read the entire paper ("Scaling VLA Models with over 100K Hours"), you realize that all of the scaling experiments on 20k hours. Therefore the headline "out-of-the-box success climbing 26% → 75% as pre-training data grows" tops out at 100% of 20k hours! What the full corpus does to that curve is never shown -> and this where things would become interesting! Xiaomi's own conclusion is that model size has stopped mattering and data is the binding constraint. The performance gap among different model sizes are less pronounced than those observed across different data scales. This result suggests that model capacity at the billions-parameter scale may already be sufficient to capture the current dataset's distribution. Which further asks the same question: why not use the 100k video hours? Anyway, I would definitely love to have a couple robots at home that can cooperate to pack my suitcase with items relevant to my next destination:

New model: your robot can now pack your suitcase 🧳 Xiaomi has released a new robot foundation model. Called Xiaomi-Robotics-1, it is designed to have a robot pick things up and move them around. But first, DEFINITIONS: - Mixture-of-Transformers (MoT): An architecture where separate transformer "experts" (e.g., one for vision-language, one for actions) share a single attention stream, so each modality gets specialized parameters without losing joint reasoning. - Vision-language model (VLM): A model that jointly understands images and text. - Diffusion transformer: A transformer trained to turn noise into structured outputs by iterative denoising, here generating robot actions rather than images. - Action chunks: Short sequences of future actions (e.g., the next ~50 motor commands) predicted in one shot instead of one step at a time. - Flow matching: A faster version of diffusion. The model learns a straight-line velocity field from noise to the target action, so it needs only a few integration steps instead of many denoising ones. Its peculiarity comes from its two stage training: 1. 100,000 hours of video shot through a UMI rig: a handheld 3D-printed gripper with a camera, worn by humans doing ordinary tasks in homes, shops, factories and offices. 2. Adapt to actual robot bodies with ~10,000 hours of real-robot data. It replaces the standard approach of teleoperating a real robot for every hour of training data. Its architecture is a Mixture-of-Transformers pairing a pre-trained Qwen3-VL vision-language model with a diffusion transformer that emits action chunks via flow matching, released in 2.6B, 5.1B and 10.5B parameter variants. However, if you read the entire paper ("Scaling VLA Models with over 100K Hours"), you realize that all of the scaling experiments on 20k hours. Therefore the headline "out-of-the-box success climbing 26% → 75% as pre-training data grows" tops out at 100% of 20k hours! What the full corpus does to that curve is never shown -> and this where things would become interesting! Xiaomi's own conclusion is that model size has stopped mattering and data is the binding constraint. The performance gap among different model sizes are less pronounced than those observed across different data scales. This result suggests that model capacity at the billions-parameter scale may already be sufficient to capture the current dataset's distribution. Which further asks the same question: why not use the 100k video hours? Anyway, I would definitely love to have a couple robots at home that can cooperate to pack my suitcase with items relevant to my next destination:

Léo

15,662 views • 1 month ago

$Introducing HRM-Text. An ultra-lean 1B-parameter reasoning language model designed to deliver strong general performance with a fraction of the data, compute, and infrastructure. Trained on just 40B structured tokens, HRM-Text achieves competitive performance while using ~1/1000 of the training data of comparable models. The kicker? The full model trains in roughly one day on a $1,000 budget. This opens the door to a new generation of AI that is powerful, accessible, and radically easier to adapt. Theories and research concepts once deemed too expensive to test are officially back in the game. Sapient Intelligence invites you to help us shape a new paradigm for general intelligence.$

Introducing HRM-Text. An ultra-lean 1B-parameter reasoning language model designed to deliver strong general performance with a fraction of the data, compute, and infrastructure. Trained on just 40B structured tokens, HRM-Text achieves competitive performance while using ~1/1000 of the training data of comparable models. The kicker? The full model trains in roughly one day on a $1,000 budget. This opens the door to a new generation of AI that is powerful, accessible, and radically easier to adapt. Theories and research concepts once deemed too expensive to test are officially back in the game. Sapient Intelligence invites you to help us shape a new paradigm for general intelligence.

Sapient Intelligence

517,084 views • 3 months ago

#WATCH | Bengaluru, Karnataka: Co-founder, Sarvam AI, Pratyush Kumar says, "... It's been three years now. We started off building on top of the Llama model, which was the best open model at that time. Sarvam was the only company outside, or maybe in Asia, that contributed to the Llama 3 Series. We understood how to add language skills into models, and Llama 3 was reasonable in Indian languages. When the first set of GPUs arrived in India, we pre-trained from scratch a 2-billion model and worked with Infosys to make it useful for IT services. Then we recognised the importance of scaling up reinforcement learning and built a reasoning model called Sarvam M on top of Mistral's small model. Over the last year, we pre-trained from scratch in India on 4,000 Hopper Series GPUs a 100-billion-plus-parameter model..."

#WATCH | Bengaluru, Karnataka: Co-founder, Sarvam AI, Pratyush Kumar says, "... It's been three years now. We started off building on top of the Llama model, which was the best open model at that time. Sarvam was the only company outside, or maybe in Asia, that contributed to the Llama 3 Series. We understood how to add language skills into models, and Llama 3 was reasonable in Indian languages. When the first set of GPUs arrived in India, we pre-trained from scratch a 2-billion model and worked with Infosys to make it useful for IT services. Then we recognised the importance of scaling up reinforcement learning and built a reasoning model called Sarvam M on top of Mistral's small model. Over the last year, we pre-trained from scratch in India on 4,000 Hopper Series GPUs a 100-billion-plus-parameter model..."

ANI

163,060 views • 1 month ago

one half of this keynote sells you the cloud forever, the other half shows the chip that lets you stop renting it this is the other half, AMD's CEO holding the chip in her hand 00:00 - Lisa Su introduces the Ryzen AI Halo, a system built for local AI 00:29 - the line that matters, it runs models up to 200 billion parameters locally, not connected to anything 00:42 - a 200 billion parameter model, the tier of the top paid AI plans, on a desktop that fits in your hand so the cloud wants $200 a month, forever, for access you never own this is the box that runs the same class of model with nothing leaving the room that is the whole point of my breakdown, the $200 a month was never the intelligence, it was the meter and the meter just became optional most people will see a spec demo the part that matters is what it lets you stop paying for full breakdown below

one half of this keynote sells you the cloud forever, the other half shows the chip that lets you stop renting it this is the other half, AMD's CEO holding the chip in her hand 00:00 - Lisa Su introduces the Ryzen AI Halo, a system built for local AI 00:29 - the line that matters, it runs models up to 200 billion parameters locally, not connected to anything 00:42 - a 200 billion parameter model, the tier of the top paid AI plans, on a desktop that fits in your hand so the cloud wants $200 a month, forever, for access you never own this is the box that runs the same class of model with nothing leaving the room that is the whole point of my breakdown, the $200 a month was never the intelligence, it was the meter and the meter just became optional most people will see a spec demo the part that matters is what it lets you stop paying for full breakdown below

John Doe

25,960 views • 2 months ago

Sam Altman just handed every startup founder a one-question autopsy. Altman: “If you’re building something on GPT-4 that a reasonable observer would say we’re going to steamroll you.” Not might. Not could. Going to. He said it with the calm of someone describing weather. Because to him it is weather. The model improves. Whatever was built on the old version’s weaknesses gets washed away. That is not strategy. That is erosion. And most founders are building on the erosion line. They find a gap in the current model. They wrap a product around it. They raise money. They hire. They scale. Then OpenAI releases the next version and the gap closes and the product has no reason to exist anymore. Altman: “When we just do our fundamental job, which is make the model better with every crank, then you get the ‘OpenAI killed my startup’ meme.” He is telling you directly. They are not hunting you. They are not even thinking about you. They are just improving the model. You happen to be standing where the improvement lands. That is the part founders refuse to hear. OpenAI does not need to compete with you. It just needs to keep doing exactly what it was already doing and your entire company disappears as a side effect. You are not a competitor. You are a temporary symptom of incomplete intelligence. The moment the intelligence completes you become nothing. Then Brad Lightcap delivered the cleanest diagnostic ever spoken in venture capital. Lightcap: “Ask if a 100x improvement in the model is something they’re excited about.” One question. The entire investment thesis reduced to a single binary. Does the next model make your company more powerful or does it make your company pointless. There is no middle ground. Lightcap: “We know the companies that come to us saying, ‘We want the next model. When is it coming out? I want to be the first to try it.’” These companies built something that feeds on intelligence. The smarter the model gets the more their product can do. They are not threatened by progress. They are starving for it. Then there are the companies Lightcap never hears from. The ones who go quiet when a new model drops. The ones who read the release notes like a death sentence. The ones privately praying the next generation takes longer because every improvement shrinks the ground beneath them. If you are hoping the model stays roughly where it is you have already told the market everything it needs to know about your company. You are not building on intelligence. You are building on the absence of it. Altman: “95% of the world should be betting on the latter category.” The latter category is simple. Assume the model keeps getting better at the pace it has been getting better. Build for that world. Not the world where GPT-4 is the ceiling. The world where GPT-4 is the floor and the ceiling has not been built yet. Then Altman told a story that should be framed on the wall of every startup in the country. A medical AI company came to him that morning. They were not complaining about the model. They were not worried about being replaced. They were demanding it improve faster. Altman: “Here’s how many people are dying every day you delay.” That is what alignment with the trajectory looks like. A company so deeply built on intelligence improving that every day the model stays the same is a day someone dies who did not have to. They are not building on a flaw. They are building on a future that has not arrived fast enough. That is the difference. The wrapper startup patches what the model cannot do today. The real company builds what the model will unlock tomorrow. One is running from the train. The other is laying the track. Altman told you the train is not slowing down. Lightcap told you exactly how to know which side you are on. One question. Does a 100x smarter model make you more valuable or erase you. If you had to pause before answering you already did.

Sam Altman just handed every startup founder a one-question autopsy. Altman: “If you’re building something on GPT-4 that a reasonable observer would say we’re going to steamroll you.” Not might. Not could. Going to. He said it with the calm of someone describing weather. Because to him it is weather. The model improves. Whatever was built on the old version’s weaknesses gets washed away. That is not strategy. That is erosion. And most founders are building on the erosion line. They find a gap in the current model. They wrap a product around it. They raise money. They hire. They scale. Then OpenAI releases the next version and the gap closes and the product has no reason to exist anymore. Altman: “When we just do our fundamental job, which is make the model better with every crank, then you get the ‘OpenAI killed my startup’ meme.” He is telling you directly. They are not hunting you. They are not even thinking about you. They are just improving the model. You happen to be standing where the improvement lands. That is the part founders refuse to hear. OpenAI does not need to compete with you. It just needs to keep doing exactly what it was already doing and your entire company disappears as a side effect. You are not a competitor. You are a temporary symptom of incomplete intelligence. The moment the intelligence completes you become nothing. Then Brad Lightcap delivered the cleanest diagnostic ever spoken in venture capital. Lightcap: “Ask if a 100x improvement in the model is something they’re excited about.” One question. The entire investment thesis reduced to a single binary. Does the next model make your company more powerful or does it make your company pointless. There is no middle ground. Lightcap: “We know the companies that come to us saying, ‘We want the next model. When is it coming out? I want to be the first to try it.’” These companies built something that feeds on intelligence. The smarter the model gets the more their product can do. They are not threatened by progress. They are starving for it. Then there are the companies Lightcap never hears from. The ones who go quiet when a new model drops. The ones who read the release notes like a death sentence. The ones privately praying the next generation takes longer because every improvement shrinks the ground beneath them. If you are hoping the model stays roughly where it is you have already told the market everything it needs to know about your company. You are not building on intelligence. You are building on the absence of it. Altman: “95% of the world should be betting on the latter category.” The latter category is simple. Assume the model keeps getting better at the pace it has been getting better. Build for that world. Not the world where GPT-4 is the ceiling. The world where GPT-4 is the floor and the ceiling has not been built yet. Then Altman told a story that should be framed on the wall of every startup in the country. A medical AI company came to him that morning. They were not complaining about the model. They were not worried about being replaced. They were demanding it improve faster. Altman: “Here’s how many people are dying every day you delay.” That is what alignment with the trajectory looks like. A company so deeply built on intelligence improving that every day the model stays the same is a day someone dies who did not have to. They are not building on a flaw. They are building on a future that has not arrived fast enough. That is the difference. The wrapper startup patches what the model cannot do today. The real company builds what the model will unlock tomorrow. One is running from the train. The other is laying the track. Altman told you the train is not slowing down. Lightcap told you exactly how to know which side you are on. One question. Does a 100x smarter model make you more valuable or erase you. If you had to pause before answering you already did.

Dustin

39,109 views • 5 months ago

The entire AI industry is racing to build the smartest model. Satya Nadella just admitted that is not where the money is. The model is not the product. The harness is. That is the exact line. And it changes what Microsoft is actually competing on. OpenAI, Anthropic, Google, xAI, Meta every frontier lab is pouring hundreds of billions into training compute, chasing the next capability jump. Each betting that raw model intelligence is the moat. Microsoft is doing the opposite. It is building the harness the orchestration layer that sits above the model, connecting it to tools, data, permissions, sub-agents, and enterprise workflows. And it is letting OpenAI, Anthropic, and MAI compete to plug into it. "You need the model. But the model is not the product. The harness is." So do the math on what a harness actually does. A raw model dropped into an enterprise answers questions. That is a chatbot. A harness turns that same model into an agent that reads the SharePoint, edits the ERP entry, pulls the GitHub PR, updates Salesforce, and files the Excel report with the right permissions, the right audit trail, and the right sub-agent for each sub-task. The model provides the intelligence. The harness converts intelligence into work. Now here's where it gets interesting. "Even the best model in the world will feel broken without a great harness. And an okay model with a great harness can feel like magic." If that is true, the enterprise buyer is not buying model quality. The enterprise buyer is buying the harness. Which means model quality becomes a commodity input over time, and harness quality becomes the sustainable moat. Compare that to the strategy the entire frontier lab industry is executing. Everyone else is chasing the numerator raw intelligence. Almost nobody at scale is racing to build the denominator the orchestration layer that determines whether that intelligence can actually be deployed profitably inside a real company. The frontier model race has a 10 to 20 percent chance of producing a single dominant winner. Nadella just told the industry he does not need to be that winner. If OpenAI wins, Microsoft wins. If Anthropic wins, Microsoft wins. If MAI wins, Microsoft wins. If someone Microsoft has never heard of trains a better model in 2027, Microsoft still wins. Because the compute they train on, the harness they get plugged into, the enterprise contracts they get delivered through, and the products they sit inside are all Microsoft. He is not building the best AI model. He is building the layer that the best AI model has to run on to make anyone money. I wonder which position looks more valuable in ten years.

The entire AI industry is racing to build the smartest model. Satya Nadella just admitted that is not where the money is. The model is not the product. The harness is. That is the exact line. And it changes what Microsoft is actually competing on. OpenAI, Anthropic, Google, xAI, Meta every frontier lab is pouring hundreds of billions into training compute, chasing the next capability jump. Each betting that raw model intelligence is the moat. Microsoft is doing the opposite. It is building the harness the orchestration layer that sits above the model, connecting it to tools, data, permissions, sub-agents, and enterprise workflows. And it is letting OpenAI, Anthropic, and MAI compete to plug into it. "You need the model. But the model is not the product. The harness is." So do the math on what a harness actually does. A raw model dropped into an enterprise answers questions. That is a chatbot. A harness turns that same model into an agent that reads the SharePoint, edits the ERP entry, pulls the GitHub PR, updates Salesforce, and files the Excel report with the right permissions, the right audit trail, and the right sub-agent for each sub-task. The model provides the intelligence. The harness converts intelligence into work. Now here's where it gets interesting. "Even the best model in the world will feel broken without a great harness. And an okay model with a great harness can feel like magic." If that is true, the enterprise buyer is not buying model quality. The enterprise buyer is buying the harness. Which means model quality becomes a commodity input over time, and harness quality becomes the sustainable moat. Compare that to the strategy the entire frontier lab industry is executing. Everyone else is chasing the numerator raw intelligence. Almost nobody at scale is racing to build the denominator the orchestration layer that determines whether that intelligence can actually be deployed profitably inside a real company. The frontier model race has a 10 to 20 percent chance of producing a single dominant winner. Nadella just told the industry he does not need to be that winner. If OpenAI wins, Microsoft wins. If Anthropic wins, Microsoft wins. If MAI wins, Microsoft wins. If someone Microsoft has never heard of trains a better model in 2027, Microsoft still wins. Because the compute they train on, the harness they get plugged into, the enterprise contracts they get delivered through, and the products they sit inside are all Microsoft. He is not building the best AI model. He is building the layer that the best AI model has to run on to make anyone money. I wonder which position looks more valuable in ten years.

Vikram M

21,463 views • 2 months ago

I am stocked to announce that I won the OpenAI Developers Codex x Mollie Hacka Worldwide Hackathon in Paris. 60+ builders, every one of us working solo, one day to ship. I built mine around a single question: who gets to own intelligence? The default answer is scary. You hand your data to a handful of labs, they train the model, they own it, and you rent back a thin slice of what your own data made possible. That is the bargain on the table today. I do not accept it. So I built Lensemble: a Tapestry like distributed training platform for JEPA based World Models. What does it enable: World Models that a community improves together, keeps sovereign, and co-owns. Two bets sit underneath it. First, the paradigm. Language models predict the next token. Powerful for text, a dead end for the physical world. A robot does not need to autocomplete sentences, it needs to predict what happens next in the world. That is what JEPA does: it learns by predicting representations instead of pixels or tokens. I am convinced world models are the most underrated paradigm in AI right now, and the closest thing we have to a ChatGPT moment for robotics. Second, the politics. Your raw trajectories never leave your machine. Each participant trains locally against a shared protocol and ships only an update, never the data. A federated round folds those updates into one shared world model, a LeWorldModel based model, and the gain is measured, not claimed: a 12k-parameter adapter on a frozen backbone, held-out prediction error down about 12 percent, the model measurably less surprised by the world. Then the upside is split by contribution weight, so the people who improved the model own a share of what it earns. This is the thesis behind Project Tapestry, the AI Alliance and Yann LeCun's push for federated, sovereign frontier AI, carried into world models and robotics. Call it Tapestry for the physical world. All of it built solo, in a single day, with Codex as my pair the whole way. Thank you to OpenAI Codex and Mollie for backing builders who ship real things, and to Boris and the organizing crew for the room and the standard you set. Intelligence the world improves, and the world owns. That is the future I want for my kids, and the one I will keep building.

I am stocked to announce that I won the OpenAI Developers Codex x Mollie Hacka Worldwide Hackathon in Paris. 60+ builders, every one of us working solo, one day to ship. I built mine around a single question: who gets to own intelligence? The default answer is scary. You hand your data to a handful of labs, they train the model, they own it, and you rent back a thin slice of what your own data made possible. That is the bargain on the table today. I do not accept it. So I built Lensemble: a Tapestry like distributed training platform for JEPA based World Models. What does it enable: World Models that a community improves together, keeps sovereign, and co-owns. Two bets sit underneath it. First, the paradigm. Language models predict the next token. Powerful for text, a dead end for the physical world. A robot does not need to autocomplete sentences, it needs to predict what happens next in the world. That is what JEPA does: it learns by predicting representations instead of pixels or tokens. I am convinced world models are the most underrated paradigm in AI right now, and the closest thing we have to a ChatGPT moment for robotics. Second, the politics. Your raw trajectories never leave your machine. Each participant trains locally against a shared protocol and ships only an update, never the data. A federated round folds those updates into one shared world model, a LeWorldModel based model, and the gain is measured, not claimed: a 12k-parameter adapter on a frozen backbone, held-out prediction error down about 12 percent, the model measurably less surprised by the world. Then the upside is split by contribution weight, so the people who improved the model own a share of what it earns. This is the thesis behind Project Tapestry, the AI Alliance and Yann LeCun's push for federated, sovereign frontier AI, carried into world models and robotics. Call it Tapestry for the physical world. All of it built solo, in a single day, with Codex as my pair the whole way. Thank you to OpenAI Codex and Mollie for backing builders who ship real things, and to Boris and the organizing crew for the room and the standard you set. Intelligence the world improves, and the world owns. That is the future I want for my kids, and the one I will keep building.

abdel

20,191 views • 2 months ago

Everyone wrote Apple off as the AI loser, but one hardware spec might flip that story upside down (Save this). @jason called Apple a screaming buy on the back of a single chip detail. The rumored M7 Ultra, expected around 2028, is designed to support up to 1.5TB of unified memory, enough to run frontier class trillion parameter AI models locally, with no cloud required. The Street's bear case on Apple is straightforward. Apple has no frontier model of its own, Siri has stumbled for years and the company effectively rents OpenAI's models for its hardest queries. That narrative treats Apple as the one Magnificent Seven name that missed the AI wave entirely but the bull case flips that framing on its head. If frontier AI models keep shrinking and getting cheaper to run, Apple doesn't need the smartest model in the world, it just needs to own the device that model runs on. And unified memory is the mechanism that makes this possible. Unlike traditional systems where the CPU and GPU each need separate memory, Apple's architecture lets the CPU, GPU and Neural Engine draw from one shared pool. A fully specced M7 Ultra could theoretically run something on the scale of a 1.2 trillion parameter model locally and that capability plugs directly into the one advantage Apple has spent over a decade building: privacy. Apple has already shipped Private Cloud Compute, a system designed so even Apple can't access user data processed off device. Apple doubled down on this at WWDC 2026, framing on device privacy as non-negotiable while rivals default to the cloud. If the best AI models get small enough to run on Apple silicon, the moat stops being the model and becomes the hardware it has to sit on. Milk Road Pro remains bullish on Apple and it remains as one of our core positions, if you want the full thesis + our full AI trades, come join us using the link below for just a $1.

Everyone wrote Apple off as the AI loser, but one hardware spec might flip that story upside down (Save this). @jason called Apple a screaming buy on the back of a single chip detail. The rumored M7 Ultra, expected around 2028, is designed to support up to 1.5TB of unified memory, enough to run frontier class trillion parameter AI models locally, with no cloud required. The Street's bear case on Apple is straightforward. Apple has no frontier model of its own, Siri has stumbled for years and the company effectively rents OpenAI's models for its hardest queries. That narrative treats Apple as the one Magnificent Seven name that missed the AI wave entirely but the bull case flips that framing on its head. If frontier AI models keep shrinking and getting cheaper to run, Apple doesn't need the smartest model in the world, it just needs to own the device that model runs on. And unified memory is the mechanism that makes this possible. Unlike traditional systems where the CPU and GPU each need separate memory, Apple's architecture lets the CPU, GPU and Neural Engine draw from one shared pool. A fully specced M7 Ultra could theoretically run something on the scale of a 1.2 trillion parameter model locally and that capability plugs directly into the one advantage Apple has spent over a decade building: privacy. Apple has already shipped Private Cloud Compute, a system designed so even Apple can't access user data processed off device. Apple doubled down on this at WWDC 2026, framing on device privacy as non-negotiable while rivals default to the cloud. If the best AI models get small enough to run on Apple silicon, the moat stops being the model and becomes the hardware it has to sit on. Milk Road Pro remains bullish on Apple and it remains as one of our core positions, if you want the full thesis + our full AI trades, come join us using the link below for just a $1.

Milk Road AI

37,459 views • 1 month ago

SITUATION EXPLAINED: Richard Sutton, the father of reinforcement learning, just launched a new lab. The goal: a trillion-parameter agent that learns in real time on 20 watts. • Sutton wrote The Bitter Lesson in 2019, its core claim: general methods that leverage computation beat hand-crafted approaches, by a large margin • His new lab, Oak Lab, builds on the "big world hypothesis": the world is too big for any model to pre-learn everything, so trying to is wasted effort by construction • Their "batch size one" algorithm updates from a single live experience as it arrives, instead of training on giant curated datasets... no massive pretraining run at all • Claimed result: multiple orders of magnitude less compute and energy than existing methods • The holy grail number: a trillion-parameter agent that learns and plans in real time on 20 watts... roughly what the human brain runs on • It's built for agentic tasks specifically, not as a general-purpose LLM replacement sof 𓋹: "It's possible that we end up looking back at LLMs as an evolution of search... not to hate on LLMs, they're very useful, they're awesome. But eventually I think that's what it ends up looking like."

SITUATION EXPLAINED: Richard Sutton, the father of reinforcement learning, just launched a new lab. The goal: a trillion-parameter agent that learns in real time on 20 watts. • Sutton wrote The Bitter Lesson in 2019, its core claim: general methods that leverage computation beat hand-crafted approaches, by a large margin • His new lab, Oak Lab, builds on the "big world hypothesis": the world is too big for any model to pre-learn everything, so trying to is wasted effort by construction • Their "batch size one" algorithm updates from a single live experience as it arrives, instead of training on giant curated datasets... no massive pretraining run at all • Claimed result: multiple orders of magnitude less compute and energy than existing methods • The holy grail number: a trillion-parameter agent that learns and plans in real time on 20 watts... roughly what the human brain runs on • It's built for agentic tasks specifically, not as a general-purpose LLM replacement sof 𓋹: "It's possible that we end up looking back at LLMs as an evolution of search... not to hate on LLMs, they're very useful, they're awesome. But eventually I think that's what it ends up looking like."

MTS

83,455 views • 1 month ago