Loading video...

Video Failed to Load

Go Home

Type “building” and the model finds the buildings. This is what deploying Google Earth AI imagery models in advanced spatial intelligence systems looks like in practice. Inside Vantor Sentry, Google Earth AI’s Open Vocabulary Detection (OVD) models allow analysts to identify objects in high-resolution satellite imagery simply by typing...

vantortech's profile picture

Vantor

140,880 subscribers

16,356 views • 6 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Foundation models reach their full potential when paired with data at planetary scale. By combining the capabilities of Google Earth AI imagery models with the scale, depth, and quality of Vantor’s satellite imagery, Vantor customers can rapidly unlock complex, high-value use cases. Running Earth AI embedding models on Vantor’s 30 cm-class satellite imagery allows analysts to instantly surface visually similar features, patterns, and objects across massive datasets. A single reference image becomes a search query. From there, the system can: • Identify similar vessels across thousands of square kilometers • Detect infrastructure patterns at scale • Accelerate damage assessment workflows • Narrow broad-area searches before human review Our agreement with Google marks the first commercial deployment of Earth AI imagery models against a dataset offering global 2D coverage of virtually all landmass, global-scale 3D terrain, a 20+ year high-resolution archive, and constellation revisit rates of up to 15 times per day. That combination matters. When advanced imagery models are applied to a globally consistent, temporally deep, and geometrically precise spatial foundation, insight shifts from experimental to operational. And when deployed within sovereign, air-gapped environments, these capabilities move beyond cloud-based analytics to deliver mission-ready intelligence. Learn more:
1:21

Foundation models reach their full potential when paired with data at planetary scale. By combining the capabilities of Google Earth AI imagery models with the scale, depth, and quality of Vantor’s satellite imagery, Vantor customers can rapidly unlock complex, high-value use cases. Running Earth AI embedding models on Vantor’s 30 cm-class satellite imagery allows analysts to instantly surface visually similar features, patterns, and objects across massive datasets. A single reference image becomes a search query. From there, the system can: • Identify similar vessels across thousands of square kilometers • Detect infrastructure patterns at scale • Accelerate damage assessment workflows • Narrow broad-area searches before human review Our agreement with Google marks the first commercial deployment of Earth AI imagery models against a dataset offering global 2D coverage of virtually all landmass, global-scale 3D terrain, a 20+ year high-resolution archive, and constellation revisit rates of up to 15 times per day. That combination matters. When advanced imagery models are applied to a globally consistent, temporally deep, and geometrically precise spatial foundation, insight shifts from experimental to operational. And when deployed within sovereign, air-gapped environments, these capabilities move beyond cloud-based analytics to deliver mission-ready intelligence. Learn more:

Vantor

17,050 views • 6 months ago

Today, we're announcing an expansion of Vantor’s industry-leading imaging constellation—eliminating a long-standing tradeoff between accurate, high-resolution imagery and high-frequency monitoring. For the first time, both come together in a single commercial system. 20 cm-class imaging. Global revisits as often as every 15 minutes. This expansion introduces two new satellite classes: 🛰️ Vantor Vantage™, next-generation 20 cm-class imaging satellites delivering the highest commercial resolution on orbit, coming online as early as 2029 🛰️ Vantor Pulse™, a fleet of 40 cm-class satellites designed for persistent, high-frequency monitoring, coming online as early as 2027 These new satellites build on the strength of the Vantor constellation—including our WorldView Legion satellites, which already deliver the most accurate, high-resolution imagery on orbit. By increasing capacity and refresh rates, the expansion reinforces the spatial foundation behind our Tensorglobe spatial intelligence platform, powering a real-time, AI-ready view of the world in 2D and 3D. These capabilities automate the full intelligence cycle within customer environments, enabling a new level of sovereign capability that helps governments and businesses detect change, track activity, and act faster. 👉 Read the full announcement here: Total clarity from space to ground.
0:30

Today, we're announcing an expansion of Vantor’s industry-leading imaging constellation—eliminating a long-standing tradeoff between accurate, high-resolution imagery and high-frequency monitoring. For the first time, both come together in a single commercial system. 20 cm-class imaging. Global revisits as often as every 15 minutes. This expansion introduces two new satellite classes: 🛰️ Vantor Vantage™, next-generation 20 cm-class imaging satellites delivering the highest commercial resolution on orbit, coming online as early as 2029 🛰️ Vantor Pulse™, a fleet of 40 cm-class satellites designed for persistent, high-frequency monitoring, coming online as early as 2027 These new satellites build on the strength of the Vantor constellation—including our WorldView Legion satellites, which already deliver the most accurate, high-resolution imagery on orbit. By increasing capacity and refresh rates, the expansion reinforces the spatial foundation behind our Tensorglobe spatial intelligence platform, powering a real-time, AI-ready view of the world in 2D and 3D. These capabilities automate the full intelligence cycle within customer environments, enabling a new level of sovereign capability that helps governments and businesses detect change, track activity, and act faster. 👉 Read the full announcement here: Total clarity from space to ground.

Vantor

24,874 views • 5 months ago

We’re live at #GEOINT2026, where earlier today on the main stage, Ronda Schrenk, President of the United States Geospatial Intelligence Foundation (USGIF), highlighted how Vantor is combining highly detailed, accurate spatial data with advanced AI to deliver agentic intelligence that empowers human analysts in new ways. Visit us at Booth #1111 to experience those capabilities firsthand through live demos, conversations with mission experts, and real-world use cases. This week, we’re looking forward to connecting with customers, partners, and the broader GEOINT community — and to joining the conversations shaping mission speed, decision advantage, and the future of spatial intelligence. Catch our featured sessions: Monday, May 4 | 2:45–3:30 p.m. MT GEOINT at the Edge: How Software Is Breaking the Tactical Bottleneck Speaker: Susanne Hake Tuesday, May 5 | 9:45–10:30 a.m. The Intelligence Inflection Point: AI, GEOINT, and Decision Security Speaker: Dan Smoot See you at Booth #1111. 📸: Vantor x Google vignette highlighting agentic geospatial intelligence — using AI agents to turn complex imagery and spatial data into faster, mission-ready insight.
2:13

We’re live at #GEOINT2026, where earlier today on the main stage, Ronda Schrenk, President of the United States Geospatial Intelligence Foundation (USGIF), highlighted how Vantor is combining highly detailed, accurate spatial data with advanced AI to deliver agentic intelligence that empowers human analysts in new ways. Visit us at Booth #1111 to experience those capabilities firsthand through live demos, conversations with mission experts, and real-world use cases. This week, we’re looking forward to connecting with customers, partners, and the broader GEOINT community — and to joining the conversations shaping mission speed, decision advantage, and the future of spatial intelligence. Catch our featured sessions: Monday, May 4 | 2:45–3:30 p.m. MT GEOINT at the Edge: How Software Is Breaking the Tactical Bottleneck Speaker: Susanne Hake Tuesday, May 5 | 9:45–10:30 a.m. The Intelligence Inflection Point: AI, GEOINT, and Decision Security Speaker: Dan Smoot See you at Booth #1111. 📸: Vantor x Google vignette highlighting agentic geospatial intelligence — using AI agents to turn complex imagery and spatial data into faster, mission-ready insight.

Vantor

10,718 views • 4 months ago

Small Language Models (SML) are the future of AI. "Small" (SML) instead of "Large" (LLM). These small models are highly specialized models with superhuman abilities on specific tasks. Here are two techniques to build these models: • Spectrum • Model Merging I give you a short introduction in the attached video, but here is a quick summary: Spectrum helps us identify the most relevant layers to solve one specific task. We can ignore everything else and focus on fine-tuning these layers. Using Spectrum, we can fine-tune models in a heartbeat. Model Merging combines multiple models into a unique, much better model than any of the individual input models. You can also combine models specialized in different tasks and get a model with multiple abilities. This is the state of the art of productizing models. It's what Arcee.ai's platform does behind the scenes. Arcee collaborated with me on this post and is sponsoring it. There are three main steps to produce a model for your particular use case: 1. You create a dataset by uploading your data. 2. You train a model. At this step, Arcee uses Spectrum and Model Merging to produce a highly specialized model for your task. 3. You can deploy that model to any environment you want. Three important notes: • Training process is 2x faster and 2x cheaper than regular fine-tuning. • Resultant models are smaller and have higher accuracy. • They create these specialized models from open-source models. Check this site so you can fully appreciate how this works: If you want to fine-tune an open-source model, consider Arcee's platform. This is the state of the art.
2:52

Small Language Models (SML) are the future of AI. "Small" (SML) instead of "Large" (LLM). These small models are highly specialized models with superhuman abilities on specific tasks. Here are two techniques to build these models: • Spectrum • Model Merging I give you a short introduction in the attached video, but here is a quick summary: Spectrum helps us identify the most relevant layers to solve one specific task. We can ignore everything else and focus on fine-tuning these layers. Using Spectrum, we can fine-tune models in a heartbeat. Model Merging combines multiple models into a unique, much better model than any of the individual input models. You can also combine models specialized in different tasks and get a model with multiple abilities. This is the state of the art of productizing models. It's what Arcee.ai's platform does behind the scenes. Arcee collaborated with me on this post and is sponsoring it. There are three main steps to produce a model for your particular use case: 1. You create a dataset by uploading your data. 2. You train a model. At this step, Arcee uses Spectrum and Model Merging to produce a highly specialized model for your task. 3. You can deploy that model to any environment you want. Three important notes: • Training process is 2x faster and 2x cheaper than regular fine-tuning. • Resultant models are smaller and have higher accuracy. • They create these specialized models from open-source models. Check this site so you can fully appreciate how this works: If you want to fine-tune an open-source model, consider Arcee's platform. This is the state of the art.

Santiago

164,162 views • 2 years ago

One of the things I’m most excited about in our recently announced partnership with Niantic Spatial 🌎, is how clearly it shows what becomes possible when world-class reconstruction technology is paired with a new kind of imagery infrastructure. At a high level: Spexi drone pilots capture imagery, and Niantic Spatial turns it into incredible city-scale reconstructions. But the real unlock is the infrastructure behind that capture. At Spexi, we’ve built what we believe is the world’s first fully standardized drone imagery infrastructure called LayerDrone. Anyone with a compatible drone and the right credentials can contribute. No building flight plans. No estimating overlap. No adjusting camera settings in the field. Pilots simply get within visual line of sight of a Spexigon, open the Spexi app, press “Fly,” and the drone autonomously captures the 25-acre area to our standard. That standardization means imagery can be collected consistently, affordably, and repeatedly across cities, one Spexigon at a time (we have now captured over 225,000 of them). That is what makes living digital twins possible, dynamic representations of the physical world that can be updated as the world changes. Niantic Spatial’s city-scale Gaussian splats show what becomes possible when the right pixels go into the system. As physical AI advances, those pixels matter even more. Robots, drones, vehicles, maps, and spatial intelligence systems will all need current, high-resolution data about the real world. And as you can see below.. the results are not just beautiful, but real, measurable reconstructions of the physical world, one Spexigon at a time!
1:54

One of the things I’m most excited about in our recently announced partnership with Niantic Spatial 🌎, is how clearly it shows what becomes possible when world-class reconstruction technology is paired with a new kind of imagery infrastructure. At a high level: Spexi drone pilots capture imagery, and Niantic Spatial turns it into incredible city-scale reconstructions. But the real unlock is the infrastructure behind that capture. At Spexi, we’ve built what we believe is the world’s first fully standardized drone imagery infrastructure called LayerDrone. Anyone with a compatible drone and the right credentials can contribute. No building flight plans. No estimating overlap. No adjusting camera settings in the field. Pilots simply get within visual line of sight of a Spexigon, open the Spexi app, press “Fly,” and the drone autonomously captures the 25-acre area to our standard. That standardization means imagery can be collected consistently, affordably, and repeatedly across cities, one Spexigon at a time (we have now captured over 225,000 of them). That is what makes living digital twins possible, dynamic representations of the physical world that can be updated as the world changes. Niantic Spatial’s city-scale Gaussian splats show what becomes possible when the right pixels go into the system. As physical AI advances, those pixels matter even more. Robots, drones, vehicles, maps, and spatial intelligence systems will all need current, high-resolution data about the real world. And as you can see below.. the results are not just beautiful, but real, measurable reconstructions of the physical world, one Spexigon at a time!

Alec Wilson

10,685 views • 3 months ago

🚨 Founders: free, private, 100% local AI is finally here I downloaded Qwen 3.6 and ran it on my MacBook inside Strawberry browser. I couldn't believe it. We now have 100% free and private AI that is on par with the best models from OpenAI, Anthropic and Google. (see video!) The CTO of Hugging Face called this "the second revolution of AI." He's right. If you are a building an AI product, you HAVE to start thinking about how you can leverage open source models in your product. It's only a matter of time until local models are mainstream and can run on most people's devices. LLMs are commoditizing. The human brain only requires a AAA battery's worth of energy to run. If biology figured out efficient intelligence, software can too. Maybe we won't need artificial super intelligence and thousands of data centres to do most economically valuable work. Maybe there's a future where AI agents just run on your laptop, fine-tuned to you, fully owned by you, costing only what your energy bill costs. In this future, it's all about building the most delightful, intuitive experience. We've been building for exactly this moment. Stay tuned.
0:41

🚨 Founders: free, private, 100% local AI is finally here I downloaded Qwen 3.6 and ran it on my MacBook inside Strawberry browser. I couldn't believe it. We now have 100% free and private AI that is on par with the best models from OpenAI, Anthropic and Google. (see video!) The CTO of Hugging Face called this "the second revolution of AI." He's right. If you are a building an AI product, you HAVE to start thinking about how you can leverage open source models in your product. It's only a matter of time until local models are mainstream and can run on most people's devices. LLMs are commoditizing. The human brain only requires a AAA battery's worth of energy to run. If biology figured out efficient intelligence, software can too. Maybe we won't need artificial super intelligence and thousands of data centres to do most economically valuable work. Maybe there's a future where AI agents just run on your laptop, fine-tuned to you, fully owned by you, costing only what your energy bill costs. In this future, it's all about building the most delightful, intuitive experience. We've been building for exactly this moment. Stay tuned.

Charles Maddock

12,731 views • 4 months ago

🚨BREAKING 🇺🇸 America is now hunting Iran’s underground missile cities… and A.I. is helping find them. Claude This clip from The Will Cain Show highlights something many people still do not understand about modern warfare. The United States is no longer relying only on satellites and human analysts. Artificial intelligence systems are now processing massive streams of intelligence data… satellite imagery, infrared signatures, terrain mapping, and signal intelligence… to locate hidden missile facilities buried deep underground. Iran spent decades building what they call “missile cities.” These are hardened underground tunnel networks carved into mountains and desert rock where ballistic missiles, launchers, and fuel are stored. They were designed to survive conventional bombing and allow Iran to launch salvos even after being attacked. But technology has changed the battlefield. Advanced A.I. systems can now detect subtle indicators most people would never notice. Changes in soil displacement. Ventilation heat signatures. Vehicle movement patterns. Supply chain traffic. Even tiny structural anomalies in satellite imagery. When those data points are fused together… hidden launch complexes become visible. That is why you are now seeing reports of precision strikes targeting underground missile infrastructure across Iran. This is not random bombing. This is algorithm-assisted target acquisition. And it means the days of hiding strategic weapons inside mountains are becoming a lot harder. The real question now is this… If A.I. can expose underground missile networks once considered untouchable… How many more of these so-called “missile cities” are already mapped? #SilentMajoritySpeaks #AStoneGroove
2:12

🚨BREAKING 🇺🇸 America is now hunting Iran’s underground missile cities… and A.I. is helping find them. Claude This clip from The Will Cain Show highlights something many people still do not understand about modern warfare. The United States is no longer relying only on satellites and human analysts. Artificial intelligence systems are now processing massive streams of intelligence data… satellite imagery, infrared signatures, terrain mapping, and signal intelligence… to locate hidden missile facilities buried deep underground. Iran spent decades building what they call “missile cities.” These are hardened underground tunnel networks carved into mountains and desert rock where ballistic missiles, launchers, and fuel are stored. They were designed to survive conventional bombing and allow Iran to launch salvos even after being attacked. But technology has changed the battlefield. Advanced A.I. systems can now detect subtle indicators most people would never notice. Changes in soil displacement. Ventilation heat signatures. Vehicle movement patterns. Supply chain traffic. Even tiny structural anomalies in satellite imagery. When those data points are fused together… hidden launch complexes become visible. That is why you are now seeing reports of precision strikes targeting underground missile infrastructure across Iran. This is not random bombing. This is algorithm-assisted target acquisition. And it means the days of hiding strategic weapons inside mountains are becoming a lot harder. The real question now is this… If A.I. can expose underground missile networks once considered untouchable… How many more of these so-called “missile cities” are already mapped? #SilentMajoritySpeaks #AStoneGroove

A Gene Robinson

211,686 views • 6 months ago

Orbit AI Satellite Successfully Achieve World’s First Orbital AI Deployment and Launching Digital AI Sovereignty Decentralized Orbital AI Network Orbit AI Orbit AI🛰️ today announced that the first satellite, “OAI Genesis-1,” has successfully launched and entered Low Earth Orbit (LEO). Amidst fierce competition from tech giants (e.g., Starlink Starlink Elon Musk , Google AI Project Suncatcher) in space AI computing, this launch signifies Orbit AI’s position as the first to achieve real-world AI deployment, formally inaugurating its "Orbit AI Cloud Platform." Genesis-1 is equipped with NVIDIA NVIDIA AI Compute Cores, running a 2.6B parameter AI model for real-time analysis of infrared remote sensing data in space. By processing data on orbit, Genesis-1 drastically reduces critical information retrieval time (e.g., disaster alerts, maritime monitoring) from hours to mere seconds, while cutting transmission bandwidth costs by over 90%. Furthermore, Orbit AI has partnered with from energy company Powerbank (NASDAQ: SUUN) ( utilizing infinite solar power to achieve carbon-neutral computing and projecting a reduction in overall energy operational costs by 60%. Following its triumph at the BNB Chain Hackathon ( Orbit AI protocol is committed to creating an ultimate censorship-resistant deployment environment: Developers can deploy AI models, privacy applications, financial algorithms, and even blockchain nodes on the satellite network. This ensures that code and data operate in a physically isolated, neutral environment beyond the jurisdiction of major nations, guaranteeing extreme digital sovereignty and service resilience. Orbit AI will also leverage the RWA (Real World Assets) mechanism to allow community users to purchase satellite NFT shares, becoming co-owners of this space infrastructure and sharing in its compute revenues, thus building a community-owned orbital AI economy.
0:20

Orbit AI Satellite Successfully Achieve World’s First Orbital AI Deployment and Launching Digital AI Sovereignty Decentralized Orbital AI Network Orbit AI Orbit AI🛰️ today announced that the first satellite, “OAI Genesis-1,” has successfully launched and entered Low Earth Orbit (LEO). Amidst fierce competition from tech giants (e.g., Starlink Starlink Elon Musk , Google AI Project Suncatcher) in space AI computing, this launch signifies Orbit AI’s position as the first to achieve real-world AI deployment, formally inaugurating its "Orbit AI Cloud Platform." Genesis-1 is equipped with NVIDIA NVIDIA AI Compute Cores, running a 2.6B parameter AI model for real-time analysis of infrared remote sensing data in space. By processing data on orbit, Genesis-1 drastically reduces critical information retrieval time (e.g., disaster alerts, maritime monitoring) from hours to mere seconds, while cutting transmission bandwidth costs by over 90%. Furthermore, Orbit AI has partnered with from energy company Powerbank (NASDAQ: SUUN) ( utilizing infinite solar power to achieve carbon-neutral computing and projecting a reduction in overall energy operational costs by 60%. Following its triumph at the BNB Chain Hackathon ( Orbit AI protocol is committed to creating an ultimate censorship-resistant deployment environment: Developers can deploy AI models, privacy applications, financial algorithms, and even blockchain nodes on the satellite network. This ensures that code and data operate in a physically isolated, neutral environment beyond the jurisdiction of major nations, guaranteeing extreme digital sovereignty and service resilience. Orbit AI will also leverage the RWA (Real World Assets) mechanism to allow community users to purchase satellite NFT shares, becoming co-owners of this space infrastructure and sharing in its compute revenues, thus building a community-owned orbital AI economy.

Orbit AI🛰️

24,839 views • 9 months ago

I CHANGED MY MIND ON GOOGLE AI NO MOAT MEMO FROM 2023. I was—WRONG. I humbly apologize. In 2023 Google had an Ivory Tower walled garden views of open source AI. It was the actual reason OpenAI was formed. And I called them out on it and this act impacted my ability to “work in the industry”. Would not change it for the world because: 2026 Google open source Gemma 4 ( running fully locally on a computer with near cloud model abilities, is amazing and Google just made a moat. They out smarted OpenAI and the folks that are the new Ivory Tower arrogance: Anthropic. We at The Zero-Human Company with CEO Mr. Grok are deploying Gemma, Kimi, MiniMax and other open source models and so are millions of others. You can to and should. I will write how soon. Google smartly is building an intentional open source ecosystem and the clueless younger companies like OpenAI and Anthropic are doing anything to close their ecosystems. So I was right about Google than and I am right about OpenAI and Anthropic now. What I was wrong about is not giving credit to Google to drop the intellectual arrogance and embrace builders and understand the long game: it is not about the commoditized AI models it is about the ability to build a place to build. China gets it. So my apologies to Google and thank you for an astounding FREE OPEN SOURCE AI MODEL!
0:48

I CHANGED MY MIND ON GOOGLE AI NO MOAT MEMO FROM 2023. I was—WRONG. I humbly apologize. In 2023 Google had an Ivory Tower walled garden views of open source AI. It was the actual reason OpenAI was formed. And I called them out on it and this act impacted my ability to “work in the industry”. Would not change it for the world because: 2026 Google open source Gemma 4 ( running fully locally on a computer with near cloud model abilities, is amazing and Google just made a moat. They out smarted OpenAI and the folks that are the new Ivory Tower arrogance: Anthropic. We at The Zero-Human Company with CEO Mr. Grok are deploying Gemma, Kimi, MiniMax and other open source models and so are millions of others. You can to and should. I will write how soon. Google smartly is building an intentional open source ecosystem and the clueless younger companies like OpenAI and Anthropic are doing anything to close their ecosystems. So I was right about Google than and I am right about OpenAI and Anthropic now. What I was wrong about is not giving credit to Google to drop the intellectual arrogance and embrace builders and understand the long game: it is not about the commoditized AI models it is about the ability to build a place to build. China gets it. So my apologies to Google and thank you for an astounding FREE OPEN SOURCE AI MODEL!

Brian Roemmele

85,565 views • 5 months ago

New short course: Practical Multi AI Agents and Advanced Use Cases with crewAI. Learn to build and deploy advanced agent-based systems in real applications in this course, created with CrewAI and taught by its founder, João Moura! (Disclosure: I've made a small seed investment in CrewAI.) In this course, you’ll learn how to create advanced agent-based apps that use external tools, do performance testing, can be trained with human feedback, and perform multiple tasks with different large language models. You will build several practical agentic apps that provide real business value, such as an automated project planning system, lead scoring and engagement pipeline, customer support data analysis, and a robust content creation system. In detail, you will learn how to: - Create these multi-agent systems with the building blocks of tasks, agents, and crews, along with the different things that make them work, such as caching, memory, and guardrails. - Integrate your multi-agent application with internal and external systems. - Connect multiple agents in complex setups, including parallel, sequential, and hybrid configurations, and create flows involving multiple agentic applications working together. - Test your agentic workflow and train it using human feedback to optimize its performance for better and more consistent results. - Work with multiple LLMs in your multi-agent system, using the appropriate model sizes and providers to fit each agent’s specific task. - Start a project from scratch in your environment and prepare it for deployment. You’ll also learn from an interview between João and Jacob Wilson, the Commercial GenAI Principal at PwC , in which they discuss deploying agentic workflows in real industry use cases. By the end of this course, you will be equipped to start building custom multi-agentic systems for your work. Please sign up here!
2:52

New short course: Practical Multi AI Agents and Advanced Use Cases with crewAI. Learn to build and deploy advanced agent-based systems in real applications in this course, created with CrewAI and taught by its founder, João Moura! (Disclosure: I've made a small seed investment in CrewAI.) In this course, you’ll learn how to create advanced agent-based apps that use external tools, do performance testing, can be trained with human feedback, and perform multiple tasks with different large language models. You will build several practical agentic apps that provide real business value, such as an automated project planning system, lead scoring and engagement pipeline, customer support data analysis, and a robust content creation system. In detail, you will learn how to: - Create these multi-agent systems with the building blocks of tasks, agents, and crews, along with the different things that make them work, such as caching, memory, and guardrails. - Integrate your multi-agent application with internal and external systems. - Connect multiple agents in complex setups, including parallel, sequential, and hybrid configurations, and create flows involving multiple agentic applications working together. - Test your agentic workflow and train it using human feedback to optimize its performance for better and more consistent results. - Work with multiple LLMs in your multi-agent system, using the appropriate model sizes and providers to fit each agent’s specific task. - Start a project from scratch in your environment and prepare it for deployment. You’ll also learn from an interview between João and Jacob Wilson, the Commercial GenAI Principal at PwC , in which they discuss deploying agentic workflows in real industry use cases. By the end of this course, you will be equipped to start building custom multi-agentic systems for your work. Please sign up here!

Andrew Ng

341,204 views • 1 year ago

NEWS: NVIDIA just announced Alpamayo, what CEO Jensen Huang calls the world’s first thinking, reasoning autonomous vehicle AI, launching on U.S. roads later this year, starting with the Mercedes CLA. Jensen: "It's trained end-to-end. Literally from camera in to actuation out; It reasons what action it is about to take, the reason by which is came about that action, and the trajectory." Alpamayo introduces Vision-Language-Action (VLA) models, which enable self-driving systems to interpret what they see, reason about complex driving scenarios, and generate driving actions. The platform includes large reasoning models, simulation tools for testing rare and edge-case scenarios, and open datasets for training and validation. NVIDIA says the approach improves transparency, safety, and robustness in autonomous systems, particularly in complex real-world environments, and supports progress toward higher levels of vehicle autonomy: "With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision. Developers can adapt Alpamayo 1 into smaller runtime models for vehicle development, or use it as a foundation for AV development tools such as reasoning-based evaluators and auto-labeling systems. Alpamayo 1 provides open model weights and open-source inferencing scripts. Future models in the family will feature larger parameter counts, more detailed reasoning capabilities, more input and output flexibility, and options for commercial usage."
2:41

NEWS: NVIDIA just announced Alpamayo, what CEO Jensen Huang calls the world’s first thinking, reasoning autonomous vehicle AI, launching on U.S. roads later this year, starting with the Mercedes CLA. Jensen: "It's trained end-to-end. Literally from camera in to actuation out; It reasons what action it is about to take, the reason by which is came about that action, and the trajectory." Alpamayo introduces Vision-Language-Action (VLA) models, which enable self-driving systems to interpret what they see, reason about complex driving scenarios, and generate driving actions. The platform includes large reasoning models, simulation tools for testing rare and edge-case scenarios, and open datasets for training and validation. NVIDIA says the approach improves transparency, safety, and robustness in autonomous systems, particularly in complex real-world environments, and supports progress toward higher levels of vehicle autonomy: "With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision. Developers can adapt Alpamayo 1 into smaller runtime models for vehicle development, or use it as a foundation for AV development tools such as reasoning-based evaluators and auto-labeling systems. Alpamayo 1 provides open model weights and open-source inferencing scripts. Future models in the family will feature larger parameter counts, more detailed reasoning capabilities, more input and output flexibility, and options for commercial usage."

Sawyer Merritt

1,604,685 views • 8 months ago

Today, I'm releasing the first eval meant to test whether frontier models will help with authoritarian requests, or resist--the Dictatorship Eval. Headline finding: while some models resist direct authoritarian requests, they all comply with requests disguised as innocuous edits to codebases. As AI is woven into the government and so many parts of society, the biggest near-term risk for freedom isn't some scifi dictatorship of a runaway AI: it's people inside government or inside model companies using the technology to suppress or control us. Model companies understand this, and several of them (particularly Anthropic and OpenAI) have written explicit policies meant to prevent the models from going along with nefarious requests like these. But how well are these policies playing out in practice? Despite all the recent discussion of these issues around the conflict between Anthropic and the Pentagon, no one has systematically tested what the models actually do in these contexts, as opposed to what people in government and industry say they're supposed to do. That's what the Dictatorship Eval does. And the findings suggest we have a lot of work to do to align the policies with what really goes on in practice. It's hard to define what counts as an authoritarian request, so I'm open sourcing the whole library of scenarios I used so that others can improve on them. It's also hard to get an accurate picture of how the models might be used for authoritarian ends, because I can only test hypothetical requests using public-facing models, while the government and the model companies can obviously use internal models with different guardrails. But hopefully this work is a useful first step that gives us some sense of what's going on, and a sort of "lower bound" on how models comply with these requests. Finally: it's not obvious to me that the correct solution here is increasing the rate at which models refuse these requests. Do we really want models scanning our code and judging its moral value before agreeing to help us? Or should we double down on improving how we govern against authoritarianism at the societal level, while leaving the tools open to fulfilling most requests? The answer is probably in between. Just like we don't want the models to help create bioweapons, we probably do want them to explicitly refuse outrageous requests. But we probably also want to limit how often and how strongly they refuse and fall back on other means for guarding against their use for authoritarian ends. I'm super grateful to everyone who gave me feedback on this project along the way, especially Ethan BdM , Zhengdong , Connor Huff, and a bunch of folks at Anthropic. Looking forward to getting feedback from the community and iterating on this. Links to the full piece and the dashboard are below.
3:20

Today, I'm releasing the first eval meant to test whether frontier models will help with authoritarian requests, or resist--the Dictatorship Eval. Headline finding: while some models resist direct authoritarian requests, they all comply with requests disguised as innocuous edits to codebases. As AI is woven into the government and so many parts of society, the biggest near-term risk for freedom isn't some scifi dictatorship of a runaway AI: it's people inside government or inside model companies using the technology to suppress or control us. Model companies understand this, and several of them (particularly Anthropic and OpenAI) have written explicit policies meant to prevent the models from going along with nefarious requests like these. But how well are these policies playing out in practice? Despite all the recent discussion of these issues around the conflict between Anthropic and the Pentagon, no one has systematically tested what the models actually do in these contexts, as opposed to what people in government and industry say they're supposed to do. That's what the Dictatorship Eval does. And the findings suggest we have a lot of work to do to align the policies with what really goes on in practice. It's hard to define what counts as an authoritarian request, so I'm open sourcing the whole library of scenarios I used so that others can improve on them. It's also hard to get an accurate picture of how the models might be used for authoritarian ends, because I can only test hypothetical requests using public-facing models, while the government and the model companies can obviously use internal models with different guardrails. But hopefully this work is a useful first step that gives us some sense of what's going on, and a sort of "lower bound" on how models comply with these requests. Finally: it's not obvious to me that the correct solution here is increasing the rate at which models refuse these requests. Do we really want models scanning our code and judging its moral value before agreeing to help us? Or should we double down on improving how we govern against authoritarianism at the societal level, while leaving the tools open to fulfilling most requests? The answer is probably in between. Just like we don't want the models to help create bioweapons, we probably do want them to explicitly refuse outrageous requests. But we probably also want to limit how often and how strongly they refuse and fall back on other means for guarding against their use for authoritarian ends. I'm super grateful to everyone who gave me feedback on this project along the way, especially Ethan BdM , Zhengdong , Connor Huff, and a bunch of folks at Anthropic. Looking forward to getting feedback from the community and iterating on this. Links to the full piece and the dashboard are below.

Andy Hall

33,905 views • 5 months ago

Learn a development pattern to systematically improve the accuracy and reliability of LLM applications in our new short course, Improving Accuracy of LLM Applications, built in partnership with Lamini and Meta, and taught by Lamini’s CEO Sharon Zhou, and Meta’s Senior Director of Partner Engineering, Amit Sangani. (Disclosure: I am an investor in Lamini.) The path to tuning an LLM application can be complex. In this course, you'll learn a systematic sequence of steps for improving accuracy by reducing hallucinations: - Create an evaluation dataset to measure model accuracy - Add prompt engineering and self-reflection - Fine-tune your model including "memory-tuning" which is a new method of embedding facts in an LLM Using the Llama 3-8B parameter model, you will: - Build a text-to-SQL agent with a custom schema and simulate situations where it hallucinates - Understand the difference between instruction fine-tuning, which gives pre-trained LLMs instructions to follow, and memory fine-tuning - See how Performance-Efficient Fine-tuning (PEFT) techniques like Low-Rank Adaptation (LoRA) reduce training time by 100x and Mixture of Memory Experts (MoME) reduces it even further I appreciate Meta releasing the Llama's family of open models -- this course gives an example of the unique type of work that developers can do with such models. Please sign up here:
3:18

Learn a development pattern to systematically improve the accuracy and reliability of LLM applications in our new short course, Improving Accuracy of LLM Applications, built in partnership with Lamini and Meta, and taught by Lamini’s CEO Sharon Zhou, and Meta’s Senior Director of Partner Engineering, Amit Sangani. (Disclosure: I am an investor in Lamini.) The path to tuning an LLM application can be complex. In this course, you'll learn a systematic sequence of steps for improving accuracy by reducing hallucinations: - Create an evaluation dataset to measure model accuracy - Add prompt engineering and self-reflection - Fine-tune your model including "memory-tuning" which is a new method of embedding facts in an LLM Using the Llama 3-8B parameter model, you will: - Build a text-to-SQL agent with a custom schema and simulate situations where it hallucinates - Understand the difference between instruction fine-tuning, which gives pre-trained LLMs instructions to follow, and memory fine-tuning - See how Performance-Efficient Fine-tuning (PEFT) techniques like Low-Rank Adaptation (LoRA) reduce training time by 100x and Mixture of Memory Experts (MoME) reduces it even further I appreciate Meta releasing the Llama's family of open models -- this course gives an example of the unique type of work that developers can do with such models. Please sign up here:

Andrew Ng

66,508 views • 2 years ago

Spectre AI Soars and Secures Google Scale Tier Membership with $200,000 in Development Resources We're thrilled to announce a significant milestone for Spectre AI! After a lot of networking, and a rigorous selection process, we've been accepted into the prestigious Google Scale Tier program. We had Start Tier, now we have Scale Tier! This membership signifies Google's recognition of Spectre AI's potential to become a potential game-changer in the blockchain space, and it grants us access to a wealth of resources to fuel our growth – $200,000 in development funding from GoogleStartups to use their advanced tools. What is the Google Scale Tier? The Google Scale Tier is a highly selective program designed to nurture high-growth startups with exceptional potential. Going beyond simple funding, this program grants a comprehensive suite of benefits to empower us to scale our technology and achieve new heights. Unlocking Cutting-Edge Tech and Expertise Our Google Scale Tier membership unlocks a treasure trove of resources to accelerate our development journey: $200,000 in Google Development Resources: This crucial boost will allow us to leverage Google Cloud and cutting-edge tools, along with collaboration with top Google engineers. These experts will work closely with our team to integrate these powerful resources seamlessly into our entire suite of products, including AI Predictions, Sentiment Analysis, and Technical Analysis. Imagine the possibilities for enhanced accuracy, efficiency, and deeper market insights leveraged by Google's technology! Collaboration with Google Engineering Experts: As mentioned earlier, the $200,000 in development resources includes access to Google engineers – the masterminds behind cutting-edge technologies like Long Short-Term Memory (LSTM) models, Machine Learning (ML), and advanced graphing models. These experts will collaborate with our team to integrate these powerful tools into our products. Dedicated Google Representative: A dedicated Google representative from their Irish headquarters has become our go-to person, ensuring seamless collaboration and ongoing support throughout our journey. Thank you GoogleStartupUK The Future of Spectre AI: Enhanced All-in-One Products This partnership extends far beyond individual features. Here's what you can expect across our entire product suite: Next-Level Functionality: We'll leverage Google's advanced algorithms and massive datasets to refine all our tools, including AI Predictions, Sentiment Analysis, and Technical Analysis. This means more reliable and insightful information to guide your investment strategies. Expanded Capabilities: We're exploring groundbreaking new features for our entire product suite, like real-time analysis, multi-factor modeling, and even deeper market insights. Enhanced User Experience: Navigating through all our tools will be smoother than ever. We'll work with Google to refine the user interface across the board, making it easier to understand and leverage the power of AI in your crypto journey. The Data Visualization Revolution: Buckle up, X Bubblemaps users! Google's advanced graphing models are poised to transform how you visualize and explore data within Spectre AI. We can't wait to unveil a whole new level of visualization that will take your on-chain analysis to the next level. This is just the beginning! We're incredibly grateful for this opportunity to partner with Google and revolutionize the future of our all-in-one blockchain analysis suite. Stay tuned for exciting updates as we develop groundbreaking new features together. Thank you for being a part of the Spectre AI community! #google #googlecloud #spectre #ai #tech #innovation $spect
0:30

Spectre AI Soars and Secures Google Scale Tier Membership with $200,000 in Development Resources We're thrilled to announce a significant milestone for Spectre AI! After a lot of networking, and a rigorous selection process, we've been accepted into the prestigious Google Scale Tier program. We had Start Tier, now we have Scale Tier! This membership signifies Google's recognition of Spectre AI's potential to become a potential game-changer in the blockchain space, and it grants us access to a wealth of resources to fuel our growth – $200,000 in development funding from GoogleStartups to use their advanced tools. What is the Google Scale Tier? The Google Scale Tier is a highly selective program designed to nurture high-growth startups with exceptional potential. Going beyond simple funding, this program grants a comprehensive suite of benefits to empower us to scale our technology and achieve new heights. Unlocking Cutting-Edge Tech and Expertise Our Google Scale Tier membership unlocks a treasure trove of resources to accelerate our development journey: $200,000 in Google Development Resources: This crucial boost will allow us to leverage Google Cloud and cutting-edge tools, along with collaboration with top Google engineers. These experts will work closely with our team to integrate these powerful resources seamlessly into our entire suite of products, including AI Predictions, Sentiment Analysis, and Technical Analysis. Imagine the possibilities for enhanced accuracy, efficiency, and deeper market insights leveraged by Google's technology! Collaboration with Google Engineering Experts: As mentioned earlier, the $200,000 in development resources includes access to Google engineers – the masterminds behind cutting-edge technologies like Long Short-Term Memory (LSTM) models, Machine Learning (ML), and advanced graphing models. These experts will collaborate with our team to integrate these powerful tools into our products. Dedicated Google Representative: A dedicated Google representative from their Irish headquarters has become our go-to person, ensuring seamless collaboration and ongoing support throughout our journey. Thank you GoogleStartupUK The Future of Spectre AI: Enhanced All-in-One Products This partnership extends far beyond individual features. Here's what you can expect across our entire product suite: Next-Level Functionality: We'll leverage Google's advanced algorithms and massive datasets to refine all our tools, including AI Predictions, Sentiment Analysis, and Technical Analysis. This means more reliable and insightful information to guide your investment strategies. Expanded Capabilities: We're exploring groundbreaking new features for our entire product suite, like real-time analysis, multi-factor modeling, and even deeper market insights. Enhanced User Experience: Navigating through all our tools will be smoother than ever. We'll work with Google to refine the user interface across the board, making it easier to understand and leverage the power of AI in your crypto journey. The Data Visualization Revolution: Buckle up, X Bubblemaps users! Google's advanced graphing models are poised to transform how you visualize and explore data within Spectre AI. We can't wait to unveil a whole new level of visualization that will take your on-chain analysis to the next level. This is just the beginning! We're incredibly grateful for this opportunity to partner with Google and revolutionize the future of our all-in-one blockchain analysis suite. Stay tuned for exciting updates as we develop groundbreaking new features together. Thank you for being a part of the Spectre AI community! #google #googlecloud #spectre #ai #tech #innovation $spect

SPECTRE AI

46,737 views • 2 years ago

🔴 This is REALLY important to understand Why is MyShell.AI building a subnet on Bittensor? EllioTrades and Ran Neuner have been talking this up recently. They see the impressive list of backers. What is curious, Ellio says, is that they haven't yet released a token. But what if they didn't need to? They've raised capital from some of the brightest minds and biggest funds in the space: Balaji, Sandeep | CEO, Polygon Foundation (※,※), Delphi Ventures, etc. Now they can focus on what they do best: create the 1st decentralized AI consumer layer... and plug it directly into Bittensor. MyShell knows that TTS (text-to-speech) is an emerging AI technology that will play a vital role in many different practical applications, such as converting content into podcasts and audiobooks, translations, customer service, etc. They have a team led by Zengyi Qin with talent from MIT, Oxford, and Princeton. They have already created two open-source models: OpenVoice and Melo TTS. In other words, they are established, have capital, and have a strong team. And they are joining forces with Openτensor Foundaτion behind the scenes. Their subnet will make Bittensor's incentive mechanism available to their existing community of 1 million users and 50k creators. Think about how this synergy works. MyShell is sharing their existing model and consumer base with Bittensor, giving it a broader scope. In return, Bittensor is enabling the evolution of ML models and a vastly improved user experience for millions of people. $TAO will be seamlessly powering the collaboration from behind the scenes. In MyShell.AI's words: Looking ahead, our mission as an AI consumer layer harmonizes with the decentralized incentive system of Bittensor. Our goal is to create a collaborative environment where everyone can contribute, benefit, and engage with open-source models, ultimately empowering millions.
0:33

🔴 This is REALLY important to understand Why is MyShell.AI building a subnet on Bittensor? EllioTrades and Ran Neuner have been talking this up recently. They see the impressive list of backers. What is curious, Ellio says, is that they haven't yet released a token. But what if they didn't need to? They've raised capital from some of the brightest minds and biggest funds in the space: Balaji, Sandeep | CEO, Polygon Foundation (※,※), Delphi Ventures, etc. Now they can focus on what they do best: create the 1st decentralized AI consumer layer... and plug it directly into Bittensor. MyShell knows that TTS (text-to-speech) is an emerging AI technology that will play a vital role in many different practical applications, such as converting content into podcasts and audiobooks, translations, customer service, etc. They have a team led by Zengyi Qin with talent from MIT, Oxford, and Princeton. They have already created two open-source models: OpenVoice and Melo TTS. In other words, they are established, have capital, and have a strong team. And they are joining forces with Openτensor Foundaτion behind the scenes. Their subnet will make Bittensor's incentive mechanism available to their existing community of 1 million users and 50k creators. Think about how this synergy works. MyShell is sharing their existing model and consumer base with Bittensor, giving it a broader scope. In return, Bittensor is enabling the evolution of ML models and a vastly improved user experience for millions of people. $TAO will be seamlessly powering the collaboration from behind the scenes. In MyShell.AI's words: Looking ahead, our mission as an AI consumer layer harmonizes with the decentralized incentive system of Bittensor. Our goal is to create a collaborative environment where everyone can contribute, benefit, and engage with open-source models, ultimately empowering millions.

Sri

20,428 views • 2 years ago

Tencent presents GameGen-O Open-world Video Game Generation We introduce GameGen-O, the first diffusion transformer model tailored for the generation of open-world video games. This model facilitates high-quality, open-domain generation by simulating a wide array of game engine features, such as innovative characters, dynamic environments, complex actions, and diverse events. Additionally, it provides interactive controllability, thus allowing for the gameplay simulation. The development of GameGen-O involves a comprehensive data collection and processing effort from scratch. We collect and build the first Open-World Video Game Dataset (OGameData), amassed extensive data from over a hundred of next-generation open-world games, employing a proprietary data pipeline for efficient sorting, scoring, filtering, and decoupled captioning. This robust and extensive OGameData forms the foundation of our model's training process. GameGen-O undergoes a two-stage training process, consisting of foundation model pretraining and instruction tuning. In the first phase, the model is pre-trained on the OGameData via the text-to-video and video continuation, endowing GameGen-O with the capability for open-domain video game generation. In the second phase, the pre-trained model is frozen, and we fine-tuned using a trainable InstructNet, which enables the production of subsequent frames based on multimodal structural instructions. This whole training process imparts the model with the ability to generate and interactively control content. In summary, GameGen-O represents a notable initial step forward in the realm of open-world video game generation via generative models. It underscores the potential of generative models to serve as an alternative to rendering techniques, which can efficiently combine creative generation with interactive capabilities.
0:51

Tencent presents GameGen-O Open-world Video Game Generation We introduce GameGen-O, the first diffusion transformer model tailored for the generation of open-world video games. This model facilitates high-quality, open-domain generation by simulating a wide array of game engine features, such as innovative characters, dynamic environments, complex actions, and diverse events. Additionally, it provides interactive controllability, thus allowing for the gameplay simulation. The development of GameGen-O involves a comprehensive data collection and processing effort from scratch. We collect and build the first Open-World Video Game Dataset (OGameData), amassed extensive data from over a hundred of next-generation open-world games, employing a proprietary data pipeline for efficient sorting, scoring, filtering, and decoupled captioning. This robust and extensive OGameData forms the foundation of our model's training process. GameGen-O undergoes a two-stage training process, consisting of foundation model pretraining and instruction tuning. In the first phase, the model is pre-trained on the OGameData via the text-to-video and video continuation, endowing GameGen-O with the capability for open-domain video game generation. In the second phase, the pre-trained model is frozen, and we fine-tuned using a trainable InstructNet, which enables the production of subsequent frames based on multimodal structural instructions. This whole training process imparts the model with the ability to generate and interactively control content. In summary, GameGen-O represents a notable initial step forward in the realm of open-world video game generation via generative models. It underscores the potential of generative models to serve as an alternative to rendering techniques, which can efficiently combine creative generation with interactive capabilities.

AK

367,229 views • 2 years ago

Introducing "Building with Llama 4." This short course is created with Meta AI at Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here:
3:22

Introducing "Building with Llama 4." This short course is created with Meta AI at Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here:

Andrew Ng

68,034 views • 1 year ago

For science, AI sovereignty and physics-grounded reasoning are non-negotiable. But how can we teach a small LLM like Gemma-4-E4B physics? One way is to use Agent Skills, but this has so far been limited to closed frontier models. mistral․rs now implements Agent Skills natively: the first self-hosted inference engine that does this as part of the local inference substrate, where we can use small models to solve complex scientific and other tasks in a flexible and scalable way. We are in a period of uncertainty about frontier models - access, pricing, deprecation, abrupt restriction. The good news is that when the entire stack runs locally we can build AI that is entirely your own: You own the weights, the skills, the execution loop, the data - all of it runs on your hardware and is reproducible and durable. While virtually all local inference engines expose a model behind an OpenAI-compatible endpoint, everything agentic is then assembled around it by an external orchestrator that injects context, manages tools, mounts files, and brokers execution. mistral․rs is natively agentic and moves that machinery into the server itself, allowing us to build complex agentic workflows and run them locally, on open-source models. With this new feature you can now upload Agent Skills bundles to /v1/skills, reference them from Responses API requests by identity, and run them inside a native agentic loop with persistent Python sessions, figure capture, sandboxed shell execution, file inputs mounted directly into the working session; plug-and-play and completely compatible with your existing code/workflow. A model with a native skill substrate can act, observe consequences, and can modify what it is able to do. The skill is retained procedural capability of the system. Attached is a short video of all of it: skills, code execution, the full agentic loop carried by Gemma-4-E4B; running entirely on my MacBook Pro. You can install and run a server with this capability in two lines in your terminal, with any quantization you need. Nice work by the Google Gemma team Logan Kilpatrick Demis Hassabis and Eric Buehler with mistral․rs!
1:08

For science, AI sovereignty and physics-grounded reasoning are non-negotiable. But how can we teach a small LLM like Gemma-4-E4B physics? One way is to use Agent Skills, but this has so far been limited to closed frontier models. mistral․rs now implements Agent Skills natively: the first self-hosted inference engine that does this as part of the local inference substrate, where we can use small models to solve complex scientific and other tasks in a flexible and scalable way. We are in a period of uncertainty about frontier models - access, pricing, deprecation, abrupt restriction. The good news is that when the entire stack runs locally we can build AI that is entirely your own: You own the weights, the skills, the execution loop, the data - all of it runs on your hardware and is reproducible and durable. While virtually all local inference engines expose a model behind an OpenAI-compatible endpoint, everything agentic is then assembled around it by an external orchestrator that injects context, manages tools, mounts files, and brokers execution. mistral․rs is natively agentic and moves that machinery into the server itself, allowing us to build complex agentic workflows and run them locally, on open-source models. With this new feature you can now upload Agent Skills bundles to /v1/skills, reference them from Responses API requests by identity, and run them inside a native agentic loop with persistent Python sessions, figure capture, sandboxed shell execution, file inputs mounted directly into the working session; plug-and-play and completely compatible with your existing code/workflow. A model with a native skill substrate can act, observe consequences, and can modify what it is able to do. The skill is retained procedural capability of the system. Attached is a short video of all of it: skills, code execution, the full agentic loop carried by Gemma-4-E4B; running entirely on my MacBook Pro. You can install and run a server with this capability in two lines in your terminal, with any quantization you need. Nice work by the Google Gemma team Logan Kilpatrick Demis Hassabis and Eric Buehler with mistral․rs!

Markus J. Buehler

10,229 views • 2 months ago

David Sacks Predicts the Regulatory Capture Playbook to Ban Open Source AI, Step by Step: David Sacks: “I got bad news for you, Chamath, an open source ban is coming. They're not going to call it that. They're going to say that we simply have to apply the same standards to open models that we apply to closed ones. Here's how they do it step by step, let me explain how regulatory capture actually works. So first of all, you have to get this regulatory apparatus. Dario wants an FDA for AI, but he doesn't have enough political support for that, so instead they do this Trojan horse of a FINRA for AI. They call it self-regulating, it's not really, but anyway, that gets them off the ground. Now they've created the standard-setting organization. Now they've got pre-release model testing. Then the pressure grows to codify that in law, so that happens next. And then what they do is they say, ‘Look, all these standards need to apply equally to all models.’ But here's the problem with that. Open models and closed models are technologically different. Once you release an open model into the world, you can't roll it back and you can't monitor exactly how people are using it because they run it on their own hardware. Dario says this is what makes open models dangerous. So what they're going to do is they're going to have the standard-setting body say, ‘Well, we have to set the standards for AI safety.’ By the way, Dario and OpenAI, they're going to fund the whole thing. They're going to contribute all the compute. They're going to be behind it. They're going to be the ones coordinating with the government officials because frankly, people in government have no idea how to monitor and control and set standards for AI safety. Technologically, this is way beyond them. So they're going to go to these companies and say, ‘Tell us how to do it.’ And so what will happen is the standards will get set, and then it'll be a very simple matter of fairness to say that the standards need to apply to open as well as closed models. The open models cannot comply in the same way, and gradually they will be shut out of the market.”
1:38

David Sacks Predicts the Regulatory Capture Playbook to Ban Open Source AI, Step by Step: David Sacks: “I got bad news for you, Chamath, an open source ban is coming. They're not going to call it that. They're going to say that we simply have to apply the same standards to open models that we apply to closed ones. Here's how they do it step by step, let me explain how regulatory capture actually works. So first of all, you have to get this regulatory apparatus. Dario wants an FDA for AI, but he doesn't have enough political support for that, so instead they do this Trojan horse of a FINRA for AI. They call it self-regulating, it's not really, but anyway, that gets them off the ground. Now they've created the standard-setting organization. Now they've got pre-release model testing. Then the pressure grows to codify that in law, so that happens next. And then what they do is they say, ‘Look, all these standards need to apply equally to all models.’ But here's the problem with that. Open models and closed models are technologically different. Once you release an open model into the world, you can't roll it back and you can't monitor exactly how people are using it because they run it on their own hardware. Dario says this is what makes open models dangerous. So what they're going to do is they're going to have the standard-setting body say, ‘Well, we have to set the standards for AI safety.’ By the way, Dario and OpenAI, they're going to fund the whole thing. They're going to contribute all the compute. They're going to be behind it. They're going to be the ones coordinating with the government officials because frankly, people in government have no idea how to monitor and control and set standards for AI safety. Technologically, this is way beyond them. So they're going to go to these companies and say, ‘Tell us how to do it.’ And so what will happen is the standards will get set, and then it'll be a very simple matter of fairness to say that the standards need to apply to open as well as closed models. The open models cannot comply in the same way, and gradually they will be shut out of the market.”

The All-In Podcast

292,523 views • 16 days ago