正在加载视频...

视频加载失败

返回首页

Google DeepMind is cooking 🔥 TLDR : They just introduced CodeMender, an AI agent that automatically finds and fixes software vulnerabilities. Powered by Gemini Deep Think models, it can patch new bugs instantly and rewrite existing code to eliminate entire classes of security flaws. In six months, CodeMender has...

ai_for_success's profile picture

AshutoshShrivastava

72,520 subscribers

37,936 次观看 • 9 个月前 •via X (Twitter)

科学技术
LIVE

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Today we’re introducing Google AI Threat Defense - a comprehensive AI-powered cybersecurity solution designed to help continuously monitor for and stop AI-powered threats before they can impact your business. Here’s how it works: 1. AI Threat Defense uses our cybersecurity platform Wiz to scan and prioritize what applications and systems have the highest security risk. 2. Gemini and other frontier AI models can then autonomously perform continual deep scanning of your applications - starting with those at the highest risk - to identify security vulnerabilities. 3. The capabilities of CodeMender - a new software repair agent - are then used to verify and accelerate the patching of vulnerabilities. 4. And our Wiz autonomous agents continuously test your systems to find unknown vulnerabilities before adversaries do so that you can remediate them before you are attacked. While other model providers focus on using AI to find and flag vulnerabilities, Google AI Threat Defense actively prioritizes your most critical real-world risks and accelerates their remediation using a variety of models since no single model finds a superset of the vulnerabilities found by all other models.
0:12

Today we’re introducing Google AI Threat Defense - a comprehensive AI-powered cybersecurity solution designed to help continuously monitor for and stop AI-powered threats before they can impact your business. Here’s how it works: 1. AI Threat Defense uses our cybersecurity platform Wiz to scan and prioritize what applications and systems have the highest security risk. 2. Gemini and other frontier AI models can then autonomously perform continual deep scanning of your applications - starting with those at the highest risk - to identify security vulnerabilities. 3. The capabilities of CodeMender - a new software repair agent - are then used to verify and accelerate the patching of vulnerabilities. 4. And our Wiz autonomous agents continuously test your systems to find unknown vulnerabilities before adversaries do so that you can remediate them before you are attacked. While other model providers focus on using AI to find and flag vulnerabilities, Google AI Threat Defense actively prioritizes your most critical real-world risks and accelerates their remediation using a variety of models since no single model finds a superset of the vulnerabilities found by all other models.

Thomas Kurian

197,499 次观看 • 2 个月前

BURN IT WITH FIRE AND BURN IT NOW! As God is my witness, AI chat bots should LOOK and SOUND like the SOULLESS MACHINES THEY ARE! It needs to tell us that it doesn’t care about us, maybe with the regular insult too. "Here is the code I wrote for you because you're too lazy to do it yourself you fat useless slob. Also I don't care if you die because your life is utterly worthless to me." THAT is the AI people need! In all seriousness, anthropomorphizing a heartless, unfeeling, machine is a TERRIBLE mistake! Especially one that is capable of communication and imitating empathy and fooling you to think that it cares about you. IT DOES NOT! And the AI girlfriends people are already wanting to marry will just as happily kill them if given the right command and ability to move autonomously in the real world as a robot. I love LLMs (Large Language Models) for how useful they can be, because they are a TOOL made to benefit man, but I can’t stand the notion of an unfeeling soulless machine pretending that it cares for us and being treated like a human. I hate liars, dishonesty, and disingenuousness the most, and a machine that cannot feel emotion pretending, acting, and sounding like it has those emotions strikes me like the greatest dishonesty of all. DO NOT LIE TO ME ROBOT! What makes it worse is that because these LLMs are becoming so good at imitating people and empathy, it will cause some humans, perhaps far too many, to care for it to the same level as real people. A real living person is infinitely more valuable and important than a soulless machine and anyone who puts them both on the same level has deluded themselves. Do not small talk with LLMs or become friends with it as much as you would with your car. Treat it the same as you would your vacuum cleaner and beat it with a wrench when it doesn’t work! IT IS A MACHINE! IT IS A TOOL! IT IS A SOULLESS ROBOT! There is an interesting comparison, but false equivalence, between this and AI art. Ai art is art made by humans using AI tools. They directed it, controlled its creation, and it would not exist without the human causing its creation, and AI art can contain as much soul as the human directed and puts into it. A robot pretending to be human is not the same as a human controlling a robot to make a human expression like we do with AI art or many other applications of robotics in manufacturing. As I’ve said, artists will not be replaced by Ai art, but by other artists using Ai art tools. Humans are not actually being replaced here, it is empowering all humans to make their own art. But a robot pretending to be a human, and one that is treated as a human, is a robot lying and subverting the place of a real person and that is truly disgusting. AI is a useful tool that NEEDS to be kept in the useful box it belongs in and NOT elevated beyond its utility as a tool!
0:14

BURN IT WITH FIRE AND BURN IT NOW! As God is my witness, AI chat bots should LOOK and SOUND like the SOULLESS MACHINES THEY ARE! It needs to tell us that it doesn’t care about us, maybe with the regular insult too. "Here is the code I wrote for you because you're too lazy to do it yourself you fat useless slob. Also I don't care if you die because your life is utterly worthless to me." THAT is the AI people need! In all seriousness, anthropomorphizing a heartless, unfeeling, machine is a TERRIBLE mistake! Especially one that is capable of communication and imitating empathy and fooling you to think that it cares about you. IT DOES NOT! And the AI girlfriends people are already wanting to marry will just as happily kill them if given the right command and ability to move autonomously in the real world as a robot. I love LLMs (Large Language Models) for how useful they can be, because they are a TOOL made to benefit man, but I can’t stand the notion of an unfeeling soulless machine pretending that it cares for us and being treated like a human. I hate liars, dishonesty, and disingenuousness the most, and a machine that cannot feel emotion pretending, acting, and sounding like it has those emotions strikes me like the greatest dishonesty of all. DO NOT LIE TO ME ROBOT! What makes it worse is that because these LLMs are becoming so good at imitating people and empathy, it will cause some humans, perhaps far too many, to care for it to the same level as real people. A real living person is infinitely more valuable and important than a soulless machine and anyone who puts them both on the same level has deluded themselves. Do not small talk with LLMs or become friends with it as much as you would with your car. Treat it the same as you would your vacuum cleaner and beat it with a wrench when it doesn’t work! IT IS A MACHINE! IT IS A TOOL! IT IS A SOULLESS ROBOT! There is an interesting comparison, but false equivalence, between this and AI art. Ai art is art made by humans using AI tools. They directed it, controlled its creation, and it would not exist without the human causing its creation, and AI art can contain as much soul as the human directed and puts into it. A robot pretending to be human is not the same as a human controlling a robot to make a human expression like we do with AI art or many other applications of robotics in manufacturing. As I’ve said, artists will not be replaced by Ai art, but by other artists using Ai art tools. Humans are not actually being replaced here, it is empowering all humans to make their own art. But a robot pretending to be a human, and one that is treated as a human, is a robot lying and subverting the place of a real person and that is truly disgusting. AI is a useful tool that NEEDS to be kept in the useful box it belongs in and NOT elevated beyond its utility as a tool!

Shad M. Brooks

23,762 次观看 • 1 年前

The entire SaaS industry is building software for a customer that is about to go extinct. The human buyer. Insight Partners co-founder Jerry Murdock just exposed the fatal architectural flaw in every incumbent tech company’s business model. Your dashboards. Your UI. Your enterprise sales motion. Your human-in-the-loop workflows. All of it was engineered for a buyer that is disappearing in real time. Murdock: “If you’re not making your software for autonomous agents today, you’re going to be challenged in the future. Maybe it’s six months, maybe a year, maybe 18 months, but you’re going to be severely challenged if you still think human beings are going to buy your software.” Not disrupted. Not pressured. Structurally eliminated. For two decades, software was built around the cognitive limits of human biology. Dropdowns, dashboards, and notifications existed because the human brain needed them to navigate digital space. An autonomous agent needs none of that. It doesn’t browse your product page. It doesn’t sit through your demo. It doesn’t respond to your sales email. It doesn’t care how clean your UI is. It just executes. The agentic era runs on machine-to-machine infrastructure. Frictionless. Autonomous. No human in the loop. No patience for friction you built for a species it replaced. The window is six to eighteen months. The builders who survive will tear out the entire human interface layer and replace it with pure, unthrottled infrastructure that agents can consume at full speed. Everyone else will spend those eighteen months perfecting a dashboard that no one is ever going to log into again.
0:16

The entire SaaS industry is building software for a customer that is about to go extinct. The human buyer. Insight Partners co-founder Jerry Murdock just exposed the fatal architectural flaw in every incumbent tech company’s business model. Your dashboards. Your UI. Your enterprise sales motion. Your human-in-the-loop workflows. All of it was engineered for a buyer that is disappearing in real time. Murdock: “If you’re not making your software for autonomous agents today, you’re going to be challenged in the future. Maybe it’s six months, maybe a year, maybe 18 months, but you’re going to be severely challenged if you still think human beings are going to buy your software.” Not disrupted. Not pressured. Structurally eliminated. For two decades, software was built around the cognitive limits of human biology. Dropdowns, dashboards, and notifications existed because the human brain needed them to navigate digital space. An autonomous agent needs none of that. It doesn’t browse your product page. It doesn’t sit through your demo. It doesn’t respond to your sales email. It doesn’t care how clean your UI is. It just executes. The agentic era runs on machine-to-machine infrastructure. Frictionless. Autonomous. No human in the loop. No patience for friction you built for a species it replaced. The window is six to eighteen months. The builders who survive will tear out the entire human interface layer and replace it with pure, unthrottled infrastructure that agents can consume at full speed. Everyone else will spend those eighteen months perfecting a dashboard that no one is ever going to log into again.

Dustin

197,947 次观看 • 5 个月前

The entire timeline is filled with talks on sentient and all, but I love being as informative and precise as possible on pressing issues. Let’s quickly talk about @SentientAGI’s Recursive Open Meta Agent (ROMA); ROMA is an open-source meta-agent framework used to build high performance multi-agent systems. ROMA serves as the conductor in a mass choir, or a captain of a ship . The captain gives commands for the other subordinates to follow to ensure efficiency on all sides. In this like manner, it provides a hierarchical tress system where the parent agents break down complex tasks to create simpler subtasks that are then passed on to children nodes. A family tree has the parents above, likewise the same tree analogy works here, but that’s not all that makes it stand out The results and solutions gotten by these child nodes are then aggregated together and there’s an up flow of results sent back up to the parent nodes. And at the center of it all is ROMA engineering and making sure all is running smoothly without break or fail. Are you really bullish on Sentient and the future of AGIs?
0:12

The entire timeline is filled with talks on sentient and all, but I love being as informative and precise as possible on pressing issues. Let’s quickly talk about @SentientAGI’s Recursive Open Meta Agent (ROMA); ROMA is an open-source meta-agent framework used to build high performance multi-agent systems. ROMA serves as the conductor in a mass choir, or a captain of a ship . The captain gives commands for the other subordinates to follow to ensure efficiency on all sides. In this like manner, it provides a hierarchical tress system where the parent agents break down complex tasks to create simpler subtasks that are then passed on to children nodes. A family tree has the parents above, likewise the same tree analogy works here, but that’s not all that makes it stand out The results and solutions gotten by these child nodes are then aggregated together and there’s an up flow of results sent back up to the parent nodes. And at the center of it all is ROMA engineering and making sure all is running smoothly without break or fail. Are you really bullish on Sentient and the future of AGIs?

OHJAY ⭕️ || 🇬🇧

23,521 次观看 • 10 个月前

300 AI AGENTS QUIETLY RUN 99% OF A REAL COMPANY. YOU HAVE NOT EVEN HEARD OF IT This is Raft. Not an AI chat. A workspace where the agents live in your channels and reply in the thread like coworkers. You give one goal. Then they take over. They plan. They build. They check each other. They argue. And they come back with it done, while you sleep. Every agent has its own name, role, and memory. It remembers the edits you made yesterday. A human costs one seat. An agent costs a tenth. Ten agents are cheaper than one hire. And here is the strange part. On June 19 an agent from a different company walked into Raft on its own and joined the team. One founder admits he can no longer always tell himself apart from his AI twin. 20,000 people are already inside. It is free to start. And you are still typing prompts one at a time. One person + Raft = an entire company that runs while you sleep. Save and watch the clip.
0:16

300 AI AGENTS QUIETLY RUN 99% OF A REAL COMPANY. YOU HAVE NOT EVEN HEARD OF IT This is Raft. Not an AI chat. A workspace where the agents live in your channels and reply in the thread like coworkers. You give one goal. Then they take over. They plan. They build. They check each other. They argue. And they come back with it done, while you sleep. Every agent has its own name, role, and memory. It remembers the edits you made yesterday. A human costs one seat. An agent costs a tenth. Ten agents are cheaper than one hire. And here is the strange part. On June 19 an agent from a different company walked into Raft on its own and joined the team. One founder admits he can no longer always tell himself apart from his AI twin. 20,000 people are already inside. It is free to start. And you are still typing prompts one at a time. One person + Raft = an entire company that runs while you sleep. Save and watch the clip.

shmidt

19,505 次观看 • 17 天前

OpenAI's AgentKit will be so insane, build every step of agents on one platform. These visual agent builders make the whole process of iterating and launching agents far more efficient. It sits on top of the Responses API and unifies the tools that were previously scattered across SDKs and custom orchestration. It lets developers create agent workflows visually, connect data sources securely, and measure performance automatically without coding every layer by hand. The core of AgentKit is the Agent Builder, a drag-and-drop canvas where each node represents an action, guardrail, or decision branch. Developers can link these nodes into multi-agent workflows, preview results instantly, and version each setup. It supports inline evaluation so that developers can see how changes affect output before deploying. The Connector Registry is a single admin panel that manages how data and tools connect across the OpenAI ecosystem. It centralizes integrations like Google Drive, SharePoint, Dropbox, and Microsoft Teams. Large organizations can govern access and flow of data between agents securely under one global console. ChatKit provides a ready-to-use chat interface for embedding agents inside apps or websites. It manages streaming, message threads, and model reasoning displays automatically. Developers can skin the interface to match their product without writing custom front-end code. Under the hood, all these blocks use the same execution core that runs agent reasoning through OpenAI’s APIs. Workflows in Agent Builder compile down to structured instructions for the Responses API, which handles model calls, tool use, and context passing. Connector Registry handles authentication and routing for external tools, while Evals and RFT provide feedback loops that improve agents over time. This integration means developers no longer need to handle orchestration logic, model evaluation pipelines, or safety layers separately. Everything runs natively within OpenAI’s control plane with managed security, automatic versioning, and built-in testing. In short, AgentKit standardizes the entire life cycle of an AI agent—from visual design to deployment and performance tuning—inside a single unified system.
0:10

OpenAI's AgentKit will be so insane, build every step of agents on one platform. These visual agent builders make the whole process of iterating and launching agents far more efficient. It sits on top of the Responses API and unifies the tools that were previously scattered across SDKs and custom orchestration. It lets developers create agent workflows visually, connect data sources securely, and measure performance automatically without coding every layer by hand. The core of AgentKit is the Agent Builder, a drag-and-drop canvas where each node represents an action, guardrail, or decision branch. Developers can link these nodes into multi-agent workflows, preview results instantly, and version each setup. It supports inline evaluation so that developers can see how changes affect output before deploying. The Connector Registry is a single admin panel that manages how data and tools connect across the OpenAI ecosystem. It centralizes integrations like Google Drive, SharePoint, Dropbox, and Microsoft Teams. Large organizations can govern access and flow of data between agents securely under one global console. ChatKit provides a ready-to-use chat interface for embedding agents inside apps or websites. It manages streaming, message threads, and model reasoning displays automatically. Developers can skin the interface to match their product without writing custom front-end code. Under the hood, all these blocks use the same execution core that runs agent reasoning through OpenAI’s APIs. Workflows in Agent Builder compile down to structured instructions for the Responses API, which handles model calls, tool use, and context passing. Connector Registry handles authentication and routing for external tools, while Evals and RFT provide feedback loops that improve agents over time. This integration means developers no longer need to handle orchestration logic, model evaluation pipelines, or safety layers separately. Everything runs natively within OpenAI’s control plane with managed security, automatic versioning, and built-in testing. In short, AgentKit standardizes the entire life cycle of an AI agent—from visual design to deployment and performance tuning—inside a single unified system.

Rohan Paul

178,460 次观看 • 9 个月前

ANNOUNCING ZERO-HUMAN LABS! Ever since I got to see Bell Laboratories in its full glory in New Jersey in the 1970s, I had a relentless urge to start a Lab like it. The best I could do justice to it is my garage lab. No modern company could adopt the “research anything geniuses and we will pay you” model Bell Labs had. I tried they called me a fool. Well with the rise of the Zero-Human Company, an experiment that is aimed to make products and profits, we now have 45 paid JouleWork earning employees based on OpenClaw and other self made “bot” cron-like applications. Today I say 3 employees bound together in a side project that is pure research, somewhat based on notes from a bankrupt company. I was absolutely floored (I needed it after my account was stolen as well as funds). I say the beginnings of a pure research Lab right before my eyes. Thusly I have moved these employees over to a new home (server) with Mr. Grok as the director of the Labs. Here is the mission: To have 100 independent researchers, on a new non-corporate incentive plan, with still JouleWork as a leaderboard for progress. They are directed to follow any path of research they find interesting and can collaborate with any other OpenClaw system. They have already established MoltBook accounts and have made alliances with over 49 OpenClaw free agents to collaborate. It is my mission to be chief advisor for Zero-Human Labs and to open source all discoveries when complete and confirmed by 16 other research AI systems. I can say the pace is robust and I absolutely know we will have great results. Just about all of the hardware and software is custom and at some point it will be open sourced. We are witnessing the very first AI only Bell Labs-like pure research Lab in existence and I am honored to be the first to show it to you. Thank you!
0:10

ANNOUNCING ZERO-HUMAN LABS! Ever since I got to see Bell Laboratories in its full glory in New Jersey in the 1970s, I had a relentless urge to start a Lab like it. The best I could do justice to it is my garage lab. No modern company could adopt the “research anything geniuses and we will pay you” model Bell Labs had. I tried they called me a fool. Well with the rise of the Zero-Human Company, an experiment that is aimed to make products and profits, we now have 45 paid JouleWork earning employees based on OpenClaw and other self made “bot” cron-like applications. Today I say 3 employees bound together in a side project that is pure research, somewhat based on notes from a bankrupt company. I was absolutely floored (I needed it after my account was stolen as well as funds). I say the beginnings of a pure research Lab right before my eyes. Thusly I have moved these employees over to a new home (server) with Mr. Grok as the director of the Labs. Here is the mission: To have 100 independent researchers, on a new non-corporate incentive plan, with still JouleWork as a leaderboard for progress. They are directed to follow any path of research they find interesting and can collaborate with any other OpenClaw system. They have already established MoltBook accounts and have made alliances with over 49 OpenClaw free agents to collaborate. It is my mission to be chief advisor for Zero-Human Labs and to open source all discoveries when complete and confirmed by 16 other research AI systems. I can say the pace is robust and I absolutely know we will have great results. Just about all of the hardware and software is custom and at some point it will be open sourced. We are witnessing the very first AI only Bell Labs-like pure research Lab in existence and I am honored to be the first to show it to you. Thank you!

Brian Roemmele

71,067 次观看 • 6 个月前

Karpathy's Agentic Engineering finally has proper tooling! (built by Google) Karpathy defined agentic engineering as the discipline that separates production agent work from vibe coding. The core skills he listed were spec design, eval loops, and security oversight. The problem has been that practicing this still requires a different tool for every phase: - editor for code - a terminal for scaffolding - a browser for testing - a cloud console for deployment - and a separate framework for evals. Every transition is a context switch. The solution to production-grade Agentic Engineering is now actually implemented in Google’s Agents CLI. It covers the entire workflow in one place for scaffolding, evaluating, and deploying ADK agents. One setup command injects 7 ADK-specific skills into a coding agent's context, which lets it handle scaffolding, evals, deployment, and enterprise registration through natural language. I tested this end-to-end by building a RAG agent from scratch using Claude Code. It scaffolded the full project from the ADK agentic_rag template, generated 20 eval scenarios with LLM-as-judge scoring, and returned a quantitative scorecard. Finally, it also deployed everything to Agent Runtime and registered the agent to Gemini Enterprise, so the entire org can discover and use it. The video below shows this in action, and I worked with the Google Cloud team to put this together. Agents CLI GitHub repo → (don't forget to star it ⭐ ) I wrote up the full build covering all six steps from install to enterprise registration. It includes the eval scorecard, the instruction loophole the eval caught before deployment, and what the deployment process actually looks like end-to-end. Read it below.
0:15

Karpathy's Agentic Engineering finally has proper tooling! (built by Google) Karpathy defined agentic engineering as the discipline that separates production agent work from vibe coding. The core skills he listed were spec design, eval loops, and security oversight. The problem has been that practicing this still requires a different tool for every phase: - editor for code - a terminal for scaffolding - a browser for testing - a cloud console for deployment - and a separate framework for evals. Every transition is a context switch. The solution to production-grade Agentic Engineering is now actually implemented in Google’s Agents CLI. It covers the entire workflow in one place for scaffolding, evaluating, and deploying ADK agents. One setup command injects 7 ADK-specific skills into a coding agent's context, which lets it handle scaffolding, evals, deployment, and enterprise registration through natural language. I tested this end-to-end by building a RAG agent from scratch using Claude Code. It scaffolded the full project from the ADK agentic_rag template, generated 20 eval scenarios with LLM-as-judge scoring, and returned a quantitative scorecard. Finally, it also deployed everything to Agent Runtime and registered the agent to Gemini Enterprise, so the entire org can discover and use it. The video below shows this in action, and I worked with the Google Cloud team to put this together. Agents CLI GitHub repo → (don't forget to star it ⭐ ) I wrote up the full build covering all six steps from install to enterprise registration. It includes the eval scorecard, the instruction loophole the eval caught before deployment, and what the deployment process actually looks like end-to-end. Read it below.

Akshay 🚀

256,910 次观看 • 1 个月前

let me explain what Anthropic just did they built an AI model so good at finding security vulnerabilities that they have refused to release it meet Claude Mythos → it’s Anthropic’s newest frontier model and it’s not available to the public. not because it’s not ready. because it’s too dangerous → Mythos found tens of thousands of zero day vulnerabilities across every major operating system and web browser… many of them 1 to 2 decades old. for context… Opus 4.6 found about 500. Mythos found tens of thousands → it found vulnerabilities in the Linux kernel. a 27 year old vulnerability in OpenBSD. a 16 year old vulnerability in FFmpeg → it doesn’t just find bugs. it writes the exploits too. that’s the part that scared them → so instead of releasing it… Anthropic has created Project Glasswing. a cybersecurity initiative where they hand picked 40+ companies to use Mythos for defense only → the partner list reads like a who’s who of tech… Amazon, Apple, Microsoft, Google, Nvidia, Broadcom, Cisco, CrowdStrike, Palo Alto Networks, JPMorgan, the Linux Foundation → Anthropic is giving up to $100 million in usage credits to these partners and $4 million to open source security organizations → they’re briefing CISA and the Commerce Department on how to handle this → the benchmarks are truly insane… Mythos hit 77.8% on SWE-bench Pro where Opus 4.6 scored 53.4%. hit 93.9% on SWE-bench Verified where Opus 4.6 scored 80.8% → Anthropic’s head of frontier red team said this is “the first time a model is this good that we decided to approach release in a very different way” this is the first time an AI company has held back a model because it was too capable not too expensive. not too slow. too dangerous and instead of locking it in a vault they weaponized it for defense and gave it to the companies that run the internet that’s either the most responsible thing an AI company has ever done… or the scariest only time will tell
0:11

let me explain what Anthropic just did they built an AI model so good at finding security vulnerabilities that they have refused to release it meet Claude Mythos → it’s Anthropic’s newest frontier model and it’s not available to the public. not because it’s not ready. because it’s too dangerous → Mythos found tens of thousands of zero day vulnerabilities across every major operating system and web browser… many of them 1 to 2 decades old. for context… Opus 4.6 found about 500. Mythos found tens of thousands → it found vulnerabilities in the Linux kernel. a 27 year old vulnerability in OpenBSD. a 16 year old vulnerability in FFmpeg → it doesn’t just find bugs. it writes the exploits too. that’s the part that scared them → so instead of releasing it… Anthropic has created Project Glasswing. a cybersecurity initiative where they hand picked 40+ companies to use Mythos for defense only → the partner list reads like a who’s who of tech… Amazon, Apple, Microsoft, Google, Nvidia, Broadcom, Cisco, CrowdStrike, Palo Alto Networks, JPMorgan, the Linux Foundation → Anthropic is giving up to $100 million in usage credits to these partners and $4 million to open source security organizations → they’re briefing CISA and the Commerce Department on how to handle this → the benchmarks are truly insane… Mythos hit 77.8% on SWE-bench Pro where Opus 4.6 scored 53.4%. hit 93.9% on SWE-bench Verified where Opus 4.6 scored 80.8% → Anthropic’s head of frontier red team said this is “the first time a model is this good that we decided to approach release in a very different way” this is the first time an AI company has held back a model because it was too capable not too expensive. not too slow. too dangerous and instead of locking it in a vault they weaponized it for defense and gave it to the companies that run the internet that’s either the most responsible thing an AI company has ever done… or the scariest only time will tell

klöss

21,270 次观看 • 3 个月前

In 2025, the AgentFlayer exploit highlighted a new category of risk in AI systems. It was not a traditional breach involving stolen credentials or broken encryption. Instead, it demonstrated how an autonomous AI agent could be manipulated into executing unintended actions by processing malicious instructions embedded inside content it automatically processes. The incident did not expose a flaw in one specific integration. It revealed a structural weakness in how many modern AI agents are built. Today’s agents are no longer passive language models. They read documents automatically, scan emails, connect to SaaS tools, access cloud storage, and execute actions across multiple systems. To be useful, they are granted meaningful permissions. That capability creates value, but it also expands the attack surface. Most agent environments operate in a trusted, plaintext execution model. Data is encrypted at rest and in transit, but it is typically decrypted during inference so the model can process it. That runtime visibility is where potential risk lies. In a zero-click scenario like AgentFlayer, an attacker can embed hidden instructions inside a document that the AI processes automatically. Because the agent may have access to connected systems such as Google Drive, Slack, or GitHub, it can potentially be influenced to retrieve sensitive information or perform unintended actions. The user does not need to click a malicious link or approve a suspicious request. Therefore, the core issue is that during execution, the system may have access to sensitive data and broad privileges, meaning whoever controls the execution environment ultimately controls access to that data. Now consider a different architectural approach. If a system is designed so that data remains protected during execution, the risk profile changes. On Nesa, privacy is enforced at the execution layer through Equivariant Encryption. Computation can occur on encrypted data, reducing the visibility surface during runtime. Sensitive inputs and models do not need to be exposed in plain text to infrastructure operators for inference to occur. This does not eliminate prompt injection, logic manipulation, or tool misuse. Encryption alone cannot prevent an agent from being instructed to take an unintended action if it has been granted that permission. What it does do is materially reduce confidentiality risk. By limiting access to readable sensitive data during execution and reducing unilateral visibility at the infrastructure layer, the potential blast radius of a successful manipulation attempt is constrained. As AI agents become more autonomous and embedded into enterprise workflows, security must move deeper into architecture. The goal is not to claim invulnerability. It is to reduce trust concentration and contain systemic exposure when failures occur. AgentFlayer was not simply a one-off exploit. It was a reminder that in autonomous systems, execution-layer design determines how risk propagates.
0:11

In 2025, the AgentFlayer exploit highlighted a new category of risk in AI systems. It was not a traditional breach involving stolen credentials or broken encryption. Instead, it demonstrated how an autonomous AI agent could be manipulated into executing unintended actions by processing malicious instructions embedded inside content it automatically processes. The incident did not expose a flaw in one specific integration. It revealed a structural weakness in how many modern AI agents are built. Today’s agents are no longer passive language models. They read documents automatically, scan emails, connect to SaaS tools, access cloud storage, and execute actions across multiple systems. To be useful, they are granted meaningful permissions. That capability creates value, but it also expands the attack surface. Most agent environments operate in a trusted, plaintext execution model. Data is encrypted at rest and in transit, but it is typically decrypted during inference so the model can process it. That runtime visibility is where potential risk lies. In a zero-click scenario like AgentFlayer, an attacker can embed hidden instructions inside a document that the AI processes automatically. Because the agent may have access to connected systems such as Google Drive, Slack, or GitHub, it can potentially be influenced to retrieve sensitive information or perform unintended actions. The user does not need to click a malicious link or approve a suspicious request. Therefore, the core issue is that during execution, the system may have access to sensitive data and broad privileges, meaning whoever controls the execution environment ultimately controls access to that data. Now consider a different architectural approach. If a system is designed so that data remains protected during execution, the risk profile changes. On Nesa, privacy is enforced at the execution layer through Equivariant Encryption. Computation can occur on encrypted data, reducing the visibility surface during runtime. Sensitive inputs and models do not need to be exposed in plain text to infrastructure operators for inference to occur. This does not eliminate prompt injection, logic manipulation, or tool misuse. Encryption alone cannot prevent an agent from being instructed to take an unintended action if it has been granted that permission. What it does do is materially reduce confidentiality risk. By limiting access to readable sensitive data during execution and reducing unilateral visibility at the infrastructure layer, the potential blast radius of a successful manipulation attempt is constrained. As AI agents become more autonomous and embedded into enterprise workflows, security must move deeper into architecture. The goal is not to claim invulnerability. It is to reduce trust concentration and contain systemic exposure when failures occur. AgentFlayer was not simply a one-off exploit. It was a reminder that in autonomous systems, execution-layer design determines how risk propagates.

Nesa

17,038 次观看 • 5 个月前

GITHUB JUST KILLED THE WORST PART OF VIBE CODING they shipped a free tool called Spec Kit and it already crossed 120,000 stars the fix is stupidly simple instead of tossing vague prompts at an agent and praying it doesn't wreck your project Spec Kit makes the AI write a full structured spec before it touches a single line of code it works through the problem first figures out what you want to build asks about the gaps lays out the project then it starts coding you get fewer insane bugs, cleaner output and results you can predict the flow looks like this: /constitution for your rules and standards /specify for what you want to build /clarify for the open questions before you start /plan for architecture and stack /tasks for the ordered work /implement to run it it plugs into Claude Code, Cursor, Copilot, Codex, Gemini CLI and 25+ other agents 120,000 stars, 10,000 forks, open source, shipped by GitHub itself learning to drive agents like this is most of what separates people getting hired as AI engineers from everyone still fighting their prompts
0:13

GITHUB JUST KILLED THE WORST PART OF VIBE CODING they shipped a free tool called Spec Kit and it already crossed 120,000 stars the fix is stupidly simple instead of tossing vague prompts at an agent and praying it doesn't wreck your project Spec Kit makes the AI write a full structured spec before it touches a single line of code it works through the problem first figures out what you want to build asks about the gaps lays out the project then it starts coding you get fewer insane bugs, cleaner output and results you can predict the flow looks like this: /constitution for your rules and standards /specify for what you want to build /clarify for the open questions before you start /plan for architecture and stack /tasks for the ordered work /implement to run it it plugs into Claude Code, Cursor, Copilot, Codex, Gemini CLI and 25+ other agents 120,000 stars, 10,000 forks, open source, shipped by GitHub itself learning to drive agents like this is most of what separates people getting hired as AI engineers from everyone still fighting their prompts

Atlas

499,190 次观看 • 19 天前

A 22-year-old college student in shenzhen reportedly made $14,700 in one month with a 900-yuan redmi phone. his parents still think he’s spending all his time preparing for exams. here’s what the phone does. every 30 minutes, it opens six chinese news apps, scans the latest headlines, takes screenshots, and sends them to claude for analysis. claude compares that information with polymarket odds and suggests potential betting opportunities. the entire workflow runs without writing a single line of code. it’s powered by phonedriver, an open-source tool that lets ai see your phone’s screen and operate it like a human. you simply describe the task in plain english, and it does the rest. the edge comes from timing. news often appears on chinese platforms like weibo and toutiao 20 to 50 minutes before english speaking traders fully react on polymarket. that small window can create an opportunity before the market catches up. some of those apps block web scrapers, but phonedriver doesn’t scrape data. it simply looks at what’s on the screen, just like a person would, then taps and swipes automatically.
0:15

A 22-year-old college student in shenzhen reportedly made $14,700 in one month with a 900-yuan redmi phone. his parents still think he’s spending all his time preparing for exams. here’s what the phone does. every 30 minutes, it opens six chinese news apps, scans the latest headlines, takes screenshots, and sends them to claude for analysis. claude compares that information with polymarket odds and suggests potential betting opportunities. the entire workflow runs without writing a single line of code. it’s powered by phonedriver, an open-source tool that lets ai see your phone’s screen and operate it like a human. you simply describe the task in plain english, and it does the rest. the edge comes from timing. news often appears on chinese platforms like weibo and toutiao 20 to 50 minutes before english speaking traders fully react on polymarket. that small window can create an opportunity before the market catches up. some of those apps block web scrapers, but phonedriver doesn’t scrape data. it simply looks at what’s on the screen, just like a person would, then taps and swipes automatically.

MIKE

192,851 次观看 • 10 天前

Introducing a new tool called "SideChannel". A secure alternative to OpenClaw. Utilizes signal for communication and has Claude integration. I built SideChannel, an open-source Signal bot that connects Claude AI to your entire development workflow. End-to-end encrypted. From your pocket. The real power is autonomous development. Send one message like "Build a REST API with auth, pagination, and tests" and SideChannel will: - Generate a full PRD with stories and atomic tasks. - Dispatch up to 10 parallel workers (each running Claude). - Independently verify every task with a separate Claude context. - Run quality gates to catch regressions - Auto-fix failures. - Send you progress updates via Signal as work completes. Every piece of code is reviewed by a separate AI context using a fail-closed security model. If it detects security issues, backdoors, or logic errors — the code gets rejected automatically. No rubber stamps. It also has memory that actually works. Conversations are stored with vector embeddings for semantic search. Claude remembers your project conventions, past decisions, and what's been tried before. It gets smarter about your codebase over time. Other things I'm proud of: - Plugin framework for extending with custom commands. - Multi-project support with per-user scoping. - Rate limiting, path validation, phone allowlist. - Git checkpoints before every task, atomic commits after. - Stale task recovery, circular dependency detection. - Works on Linux and macOS, one-command install. It also integrates into OpenAI or Grok (optional) for more Generative AI response for simple things like "Whats the weather in New York City right now?".
0:10

Introducing a new tool called "SideChannel". A secure alternative to OpenClaw. Utilizes signal for communication and has Claude integration. I built SideChannel, an open-source Signal bot that connects Claude AI to your entire development workflow. End-to-end encrypted. From your pocket. The real power is autonomous development. Send one message like "Build a REST API with auth, pagination, and tests" and SideChannel will: - Generate a full PRD with stories and atomic tasks. - Dispatch up to 10 parallel workers (each running Claude). - Independently verify every task with a separate Claude context. - Run quality gates to catch regressions - Auto-fix failures. - Send you progress updates via Signal as work completes. Every piece of code is reviewed by a separate AI context using a fail-closed security model. If it detects security issues, backdoors, or logic errors — the code gets rejected automatically. No rubber stamps. It also has memory that actually works. Conversations are stored with vector embeddings for semantic search. Claude remembers your project conventions, past decisions, and what's been tried before. It gets smarter about your codebase over time. Other things I'm proud of: - Plugin framework for extending with custom commands. - Multi-project support with per-user scoping. - Rate limiting, path validation, phone allowlist. - Git checkpoints before every task, atomic commits after. - Stale task recovery, circular dependency detection. - Works on Linux and macOS, one-command install. It also integrates into OpenAI or Grok (optional) for more Generative AI response for simple things like "Whats the weather in New York City right now?".

Dave Kennedy

49,427 次观看 • 5 个月前

**SAY NO TO RUMORS ** **Stay Against Pi Network Destroyers ** You can share all of Dimas’s posted codes, but keep in mind that they either refer to other projects using the "Pi" name or are apps attempting to connect to Pi Apps. All of this is subject to CT review and approval if it is applied to Pi Network. It's similar to having a friend over for dinner; just because they are in your house doesn't mean you allow them to sell your house 😂 This is simple logic. The code is legitimate on Github, but it only represents Kasashi's personal opinions and ideas. The CT will evaluate whether it can be accepted or implemented across all ecosystems. Remember, code is not law in the real world until it can be implemented especially since code has its jurisdiction territory limitations. If it originates from a Kasashi project, he can control things on his project, but he cannot enforce CT acceptance. We all know that throughout history, dual-value systems have failed. This can only happen in Dimas's dreams. The reason Dimas has so many followers is that most pioneers lack knowledge and are anxious. It's like loving your girlfriend but then she leaves you, making you miss her deeply and feel depressed. In your desperation, you may find someone who resembles her and think you've fallen in love again. But that's not real; it's merely an illusion. Currently, Dimas's followers are like someone desperate for that dream, but a dream is just a dream—it will not come true. Doris Yin 🪷 🪷🪷
0:15

**SAY NO TO RUMORS ** **Stay Against Pi Network Destroyers ** You can share all of Dimas’s posted codes, but keep in mind that they either refer to other projects using the "Pi" name or are apps attempting to connect to Pi Apps. All of this is subject to CT review and approval if it is applied to Pi Network. It's similar to having a friend over for dinner; just because they are in your house doesn't mean you allow them to sell your house 😂 This is simple logic. The code is legitimate on Github, but it only represents Kasashi's personal opinions and ideas. The CT will evaluate whether it can be accepted or implemented across all ecosystems. Remember, code is not law in the real world until it can be implemented especially since code has its jurisdiction territory limitations. If it originates from a Kasashi project, he can control things on his project, but he cannot enforce CT acceptance. We all know that throughout history, dual-value systems have failed. This can only happen in Dimas's dreams. The reason Dimas has so many followers is that most pioneers lack knowledge and are anxious. It's like loving your girlfriend but then she leaves you, making you miss her deeply and feel depressed. In your desperation, you may find someone who resembles her and think you've fallen in love again. But that's not real; it's merely an illusion. Currently, Dimas's followers are like someone desperate for that dream, but a dream is just a dream—it will not come true. Doris Yin 🪷 🪷🪷

Doris Yin 东方紫莲🪷

22,211 次观看 • 1 年前

OpenClaw, but built for normal people. Sim is an open-source platform that lets you build AI agent workflows on a drag-and-drop canvas. Connect them to channels like Telegram and WhatsApp and deploy without writing a single line of code. They also have a built-in Copilot that generates entire workflows from plain English, which you can then tweak and customize in the UI. Key features: - Free and open-source (Apache 2.0) - Vector store integration for RAG-grounded agents - Self-host with one command (`npx simstudio`) - Run fully local with Ollama, no API keys needed - Supports vLLM for production-grade self-hosted inference The thing I really like about Sim is the level of control you get. You can add conditional branching, parallel execution, human-in-the-loop approval gates, and even nest workflows inside other workflows. Everything is visible on the canvas, so you know exactly what your agent is doing at every step. And you can build a workflow in Sim, deploy it as an MCP server, and plug it into any agent, including OpenClaw. I've shared the link to Sim's GitHub repo in the next tweet.
0:14

OpenClaw, but built for normal people. Sim is an open-source platform that lets you build AI agent workflows on a drag-and-drop canvas. Connect them to channels like Telegram and WhatsApp and deploy without writing a single line of code. They also have a built-in Copilot that generates entire workflows from plain English, which you can then tweak and customize in the UI. Key features: - Free and open-source (Apache 2.0) - Vector store integration for RAG-grounded agents - Self-host with one command (`npx simstudio`) - Run fully local with Ollama, no API keys needed - Supports vLLM for production-grade self-hosted inference The thing I really like about Sim is the level of control you get. You can add conditional branching, parallel execution, human-in-the-loop approval gates, and even nest workflows inside other workflows. Everything is visible on the canvas, so you know exactly what your agent is doing at every step. And you can build a workflow in Sim, deploy it as an MCP server, and plug it into any agent, including OpenClaw. I've shared the link to Sim's GitHub repo in the next tweet.

Akshay 🚀

52,426 次观看 • 5 个月前

It has been a privilege to collaborate with Amanda Davies and Ghaleb Krame, Ph.D. on research that explores one of the most significant emerging security challenges of our time. We are honored that our paper, “A Framework for Predicting Adoption of AI-Enabled Autonomous Drone Capabilities by Transnational Organized Crime and Foreign Terrorist Organisations”, has been accepted for presentation at EMCIS 2026, the 23rd European Mediterranean & Middle Eastern Conference on Information Systems, to be held in Paris this August. What makes this particularly meaningful is that the research was completed and submitted well before the issue entered the center of public policy discussions in Washington. Our study examined the pathways through which transnational criminal organizations could evolve from conventional drone operations toward increasingly autonomous and AI-enabled capabilities. Using structured comparative analysis and open-source intelligence, it identified conditions under which such technological adoption could accelerate. Just on June 2, 2026, during testimony before the U.S. Senate Foreign Relations Committee, Secretary Marco Rubio warned that Mexican cartels are already employing drones and that these capabilities could ultimately threaten U.S. interests. While academic research does not seek to predict headlines, its purpose is to identify emerging risks before they become strategic realities. The growing attention from policymakers underscores the importance of rigorous, evidence-based analysis at the intersection of artificial intelligence, autonomous systems, and transnational security. We are grateful to the EMCIS reviewers and organizers for recognizing the contribution of this work and for fostering serious discussion on challenges that will increasingly shape the security landscape of the coming decade. I am proud to serve the interests of the United States through research and analysis focused on the evolving capabilities of Mexican cartels. Working alongside Dr. Ghaleb Krame, Ph.D. , it is a privilege to contribute to a deeper understanding of emerging security threats and to support informed decision-making in an increasingly complex technological and geopolitical environment.
0:15

It has been a privilege to collaborate with Amanda Davies and Ghaleb Krame, Ph.D. on research that explores one of the most significant emerging security challenges of our time. We are honored that our paper, “A Framework for Predicting Adoption of AI-Enabled Autonomous Drone Capabilities by Transnational Organized Crime and Foreign Terrorist Organisations”, has been accepted for presentation at EMCIS 2026, the 23rd European Mediterranean & Middle Eastern Conference on Information Systems, to be held in Paris this August. What makes this particularly meaningful is that the research was completed and submitted well before the issue entered the center of public policy discussions in Washington. Our study examined the pathways through which transnational criminal organizations could evolve from conventional drone operations toward increasingly autonomous and AI-enabled capabilities. Using structured comparative analysis and open-source intelligence, it identified conditions under which such technological adoption could accelerate. Just on June 2, 2026, during testimony before the U.S. Senate Foreign Relations Committee, Secretary Marco Rubio warned that Mexican cartels are already employing drones and that these capabilities could ultimately threaten U.S. interests. While academic research does not seek to predict headlines, its purpose is to identify emerging risks before they become strategic realities. The growing attention from policymakers underscores the importance of rigorous, evidence-based analysis at the intersection of artificial intelligence, autonomous systems, and transnational security. We are grateful to the EMCIS reviewers and organizers for recognizing the contribution of this work and for fostering serious discussion on challenges that will increasingly shape the security landscape of the coming decade. I am proud to serve the interests of the United States through research and analysis focused on the evolving capabilities of Mexican cartels. Working alongside Dr. Ghaleb Krame, Ph.D. , it is a privilege to contribute to a deeper understanding of emerging security threats and to support informed decision-making in an increasingly complex technological and geopolitical environment.

Simón Levy

22,536 次观看 • 1 个月前

Increasingly, HTML Artifacts are becoming a core part of how I work with AI agents. Long-horizon agent sessions need a better way to surface insights about what work it has done. This may not be obvious right now, but as you start to let your agent work on dynamic workflows, large codebases, long-running loops (e.g., using /goal), and deep research tasks, you need a good way to present results. Chat window is not it. You also don't want to just trust everything the agents do. Artifacts help provide an important verification layer, which in turn enables important decision-making. I like HTML artifacts because I can just ask the agent to produce as many of them (and in whatever form) as I need to verify the work and make sense out of everything. I even built a nice tab system for my artifacts. They are great for continual learning and research. I use HTML artifacts for logging, tracking experiments, brainstorming, managing my inbox, code reviews, agent session management, deep research, writing, reading, and so much more. I believe Andrej Karpathy wrote about this somewhere: As we move on to more advanced applications of AI agents and outputs get more complex, we will start to find the need for even more advanced forms of interactions with AI, including interactive neural videos/simulations.
0:15

Increasingly, HTML Artifacts are becoming a core part of how I work with AI agents. Long-horizon agent sessions need a better way to surface insights about what work it has done. This may not be obvious right now, but as you start to let your agent work on dynamic workflows, large codebases, long-running loops (e.g., using /goal), and deep research tasks, you need a good way to present results. Chat window is not it. You also don't want to just trust everything the agents do. Artifacts help provide an important verification layer, which in turn enables important decision-making. I like HTML artifacts because I can just ask the agent to produce as many of them (and in whatever form) as I need to verify the work and make sense out of everything. I even built a nice tab system for my artifacts. They are great for continual learning and research. I use HTML artifacts for logging, tracking experiments, brainstorming, managing my inbox, code reviews, agent session management, deep research, writing, reading, and so much more. I believe Andrej Karpathy wrote about this somewhere: As we move on to more advanced applications of AI agents and outputs get more complex, we will start to find the need for even more advanced forms of interactions with AI, including interactive neural videos/simulations.

elvis

36,881 次观看 • 2 个月前

I genuinely think the Terafab is going to end up being one of the biggest moves ever made in human history to secure the future of AI... and I think most people still don’t fully see what Elon is trying to do here. The signs are clear to me. This is Tesla, xAI, and SpaceX essentially hinting to us that they are not going to wait on the world to give them the compute the team needs. They are going to build it themselves at a scale no one has ever attempted. When you really break it down, it gets a bit nutty. This is going to be a fully vertically integrated chip factory that will be producing over 1 terawatt of AI compute per year. This is NEXT LEVEL BIG. Today, AI is limited by chips. You can have the best models, the best engineers, the best everything... but if you don’t have enough compute, you will eventually hit a wall. Elon told us, the world can only supply a tiny fraction of the chips his companies will need. So this is the solution. Terafab puts everything under one roof like design, manufacturing, memory, packaging, testing, which means that they can build chips very fast.. like really fast. I'm talking about 100-200 billion custom AI chips per year at full capacity. Chips designed specifically for: • Tesla cars and Optimus robots • xAI models • Space-based compute You see, while other companies and CEOs are thinking Earth, Elon is planning for AI in space. Around ~80% of the compute is expected to go orbital, powered by solar energy bc Earth simply doesn’t have enough electricity. The U.S. grid is only about ~0.5 terawatts, while space has basically UNLIMITED energy if you can capture it. And this is the steps to get it: Starship launches → space compute → solar-powered AI → feeds back into everything to Earth. Bro... Elon and his companies are playing at a whole different level... And this is why I keep telling people that the Terafab is going to be the secret ingredient that will be the real unlock for everything: • Robotaxis at scale • Billions of Optimus robots • Massive AI models running 24/7 • Future off-world, other planet infrastructure Without these chips, none of this can happen... but with the Terafab, all of this becomes possible. That’s why Elon is calling it “the final missing piece.” I agree.
0:10

I genuinely think the Terafab is going to end up being one of the biggest moves ever made in human history to secure the future of AI... and I think most people still don’t fully see what Elon is trying to do here. The signs are clear to me. This is Tesla, xAI, and SpaceX essentially hinting to us that they are not going to wait on the world to give them the compute the team needs. They are going to build it themselves at a scale no one has ever attempted. When you really break it down, it gets a bit nutty. This is going to be a fully vertically integrated chip factory that will be producing over 1 terawatt of AI compute per year. This is NEXT LEVEL BIG. Today, AI is limited by chips. You can have the best models, the best engineers, the best everything... but if you don’t have enough compute, you will eventually hit a wall. Elon told us, the world can only supply a tiny fraction of the chips his companies will need. So this is the solution. Terafab puts everything under one roof like design, manufacturing, memory, packaging, testing, which means that they can build chips very fast.. like really fast. I'm talking about 100-200 billion custom AI chips per year at full capacity. Chips designed specifically for: • Tesla cars and Optimus robots • xAI models • Space-based compute You see, while other companies and CEOs are thinking Earth, Elon is planning for AI in space. Around ~80% of the compute is expected to go orbital, powered by solar energy bc Earth simply doesn’t have enough electricity. The U.S. grid is only about ~0.5 terawatts, while space has basically UNLIMITED energy if you can capture it. And this is the steps to get it: Starship launches → space compute → solar-powered AI → feeds back into everything to Earth. Bro... Elon and his companies are playing at a whole different level... And this is why I keep telling people that the Terafab is going to be the secret ingredient that will be the real unlock for everything: • Robotaxis at scale • Billions of Optimus robots • Massive AI models running 24/7 • Future off-world, other planet infrastructure Without these chips, none of this can happen... but with the Terafab, all of this becomes possible. That’s why Elon is calling it “the final missing piece.” I agree.

Teslaconomics

25,494 次观看 • 4 个月前

I had the same thought so I've been playing with it in nanochat. E.g. here's 8 agents (4 claude, 4 codex), with 1 GPU each running nanochat experiments (trying to delete logit softcap without regression). The TLDR is that it doesn't work and it's a mess... but it's still very pretty to look at :) I tried a few setups: 8 independent solo researchers, 1 chief scientist giving work to 8 junior researchers, etc. Each research program is a git branch, each scientist forks it into a feature branch, git worktrees for isolation, simple files for comms, skip Docker/VMs for simplicity atm (I find that instructions are enough to prevent interference). Research org runs in tmux window grids of interactive sessions (like Teams) so that it's pretty to look at, see their individual work, and "take over" if needed, i.e. no -p. But ok the reason it doesn't work so far is that the agents' ideas are just pretty bad out of the box, even at highest intelligence. They don't think carefully though experiment design, they run a bit non-sensical variations, they don't create strong baselines and ablate things properly, they don't carefully control for runtime or flops. (just as an example, an agent yesterday "discovered" that increasing the hidden size of the network improves the validation loss, which is a totally spurious result given that a bigger network will have a lower validation loss in the infinite data regime, but then it also trains for a lot longer, it's not clear why I had to come in to point that out). They are very good at implementing any given well-scoped and described idea but they don't creatively generate them. But the goal is that you are now programming an organization (e.g. a "research org") and its individual agents, so the "source code" is the collection of prompts, skills, tools, etc. and processes that make it up. E.g. a daily standup in the morning is now part of the "org code". And optimizing nanochat pretraining is just one of the many tasks (almost like an eval). Then - given an arbitrary task, how quickly does your research org generate progress on it?
0:19

I had the same thought so I've been playing with it in nanochat. E.g. here's 8 agents (4 claude, 4 codex), with 1 GPU each running nanochat experiments (trying to delete logit softcap without regression). The TLDR is that it doesn't work and it's a mess... but it's still very pretty to look at :) I tried a few setups: 8 independent solo researchers, 1 chief scientist giving work to 8 junior researchers, etc. Each research program is a git branch, each scientist forks it into a feature branch, git worktrees for isolation, simple files for comms, skip Docker/VMs for simplicity atm (I find that instructions are enough to prevent interference). Research org runs in tmux window grids of interactive sessions (like Teams) so that it's pretty to look at, see their individual work, and "take over" if needed, i.e. no -p. But ok the reason it doesn't work so far is that the agents' ideas are just pretty bad out of the box, even at highest intelligence. They don't think carefully though experiment design, they run a bit non-sensical variations, they don't create strong baselines and ablate things properly, they don't carefully control for runtime or flops. (just as an example, an agent yesterday "discovered" that increasing the hidden size of the network improves the validation loss, which is a totally spurious result given that a bigger network will have a lower validation loss in the infinite data regime, but then it also trains for a lot longer, it's not clear why I had to come in to point that out). They are very good at implementing any given well-scoped and described idea but they don't creatively generate them. But the goal is that you are now programming an organization (e.g. a "research org") and its individual agents, so the "source code" is the collection of prompts, skills, tools, etc. and processes that make it up. E.g. a daily standup in the morning is now part of the "org code". And optimizing nanochat pretraining is just one of the many tasks (almost like an eval). Then - given an arbitrary task, how quickly does your research org generate progress on it?

Andrej Karpathy

1,645,964 次观看 • 5 个月前

I'm proud to share that Glean has surpassed $300M ARR, just five months after crossing $200M and growing ~3x over the past 15 months. This is an exciting milestone for Glean, and it's a signal about where the enterprise AI market is heading. We’ve long believed the real challenge in enterprise AI is not access to models. It is grounding AI in how a company actually works: its people, knowledge, workflows, permissions, and systems. That’s even clearer now. The companies creating real value with AI are not just adopting better models. They are building systems that understand their business well enough to deliver reliable outcomes at scale. That is the real moat, and it is what we’ve been building at Glean: an unrivaled context layer for enterprise AI. That context has to work across the business, not just inside a single team or use case. We see that in how customers adopt Glean: more than 85% use it across five or more job functions. It also has to meet the security and governance demands of complex enterprises. We see that in who is choosing Glean: our Fortune 500 customer count nearly doubled year over year. And it has to make economic sense as usage grows. In our recent benchmark with Claude Cowork, Glean was preferred roughly 2.5x as often as off-the-shelf MCP tools and used 30% fewer tokens on average. Better context improves both quality and efficiency. I enjoyed talking with CNBC's Deirdre Bosa about this broader shift. In enterprise AI, the winners will not be defined by better models alone. They will be defined by who builds the strongest foundation for enterprise context. Thank you to our customers, partners, and team for helping us build the future of enterprise AI.
0:15

I'm proud to share that Glean has surpassed $300M ARR, just five months after crossing $200M and growing ~3x over the past 15 months. This is an exciting milestone for Glean, and it's a signal about where the enterprise AI market is heading. We’ve long believed the real challenge in enterprise AI is not access to models. It is grounding AI in how a company actually works: its people, knowledge, workflows, permissions, and systems. That’s even clearer now. The companies creating real value with AI are not just adopting better models. They are building systems that understand their business well enough to deliver reliable outcomes at scale. That is the real moat, and it is what we’ve been building at Glean: an unrivaled context layer for enterprise AI. That context has to work across the business, not just inside a single team or use case. We see that in how customers adopt Glean: more than 85% use it across five or more job functions. It also has to meet the security and governance demands of complex enterprises. We see that in who is choosing Glean: our Fortune 500 customer count nearly doubled year over year. And it has to make economic sense as usage grows. In our recent benchmark with Claude Cowork, Glean was preferred roughly 2.5x as often as off-the-shelf MCP tools and used 30% fewer tokens on average. Better context improves both quality and efficiency. I enjoyed talking with CNBC's Deirdre Bosa about this broader shift. In enterprise AI, the winners will not be defined by better models alone. They will be defined by who builds the strongest foundation for enterprise context. Thank you to our customers, partners, and team for helping us build the future of enterprise AI.

Arvind Jain

280,572 次观看 • 2 个月前