hackcv
📰 AI 研究简报 · 7月14日
今日头条：RT by @ylecun: The Well Just Dropped: 15 Terabytes of Pure Physics Gold Is Now Open Source The scientific AI world just got a massive upgrade.Polymathic AI, in collaboration with the Flatiron Institute and researchers from Princeton, Cambridge, NYU, Berkeley, Los Alamos, and more, has released The Well: a staggering 15TB collection of high-fidelity physics simulations. This isn’t toy data. These are real, expensive-to-run simulations across 16 different physical domains, including turbulent fluid dynamics, supernova explosions, magneto-hydrodynamic cosmic flows, acoustic scattering, and active biological matter. Until now, reproducing this level of data required weeks on national supercomputers and grant money most teams will never see. The Well changes everything. It’s purpose-built for training PDE surrogate models the AI systems that can replace slow, costly physics solvers with a single fast neural network forward pass. Everything is fully open source, easy to load with PyTorch, and ready to drop straight into your training pipeline. Researchers and builders can now train on world-class physics data without the insane compute barriers that used to stand in the way. This is more than just another dataset drop. It’s a serious accelerator for scientific machine learning.The future of physics-informed AI just got a whole lot more accessible.Get it here: https://polymathic-ai.org/the_well/ —— <p>The Well Just Dropped: 15 Terabytes of Pure Physics Gold Is Now Open Source<br> <br> The scientific AI world just got a massive upgrade.Polymathic AI, in collaboration with the Flatiron Institute and researchers from Princeton, Cambridge, NYU, Berkeley, Los Alamos, and more, has released The Well: a staggering 15TB collection of high-fidelity physics simulations. <br> <br> This isn’t toy data.
共 200 条预计阅读 120 分钟Token 消耗 360,000
← 前一天已是最新
arXiv 最新论文（0）GitHub 热门开源项目（0）精选 AI 行业资讯（198）YouTube 热门 AI 视频（2）
arXiv 最新论文
本板块今日暂无内容
GitHub 热门开源项目
本板块今日暂无内容
精选 AI 行业资讯
行业资讯精选 85

RT by @swyx: Dominic Kundel (from @OpenAI) gives the inside scoop on where GPT-5.6-Sol gets magical: computer use. Background browser tabs, app control, multi-agent fanout, and Codex verifying its own work all change when latency drops. Join us in the Token Billionaires Lounge, presented by @cerebras and @aiDotEngineer. In conversation with @dkundel // @MilksandMatcha· GPT-5.6-Sol 的低延迟魔法

<p>Dominic Kundel (from <a href="https://rss.xcancel.com/OpenAI" title="OpenAI">@OpenAI</a…

📎 swyx🕒 07-14 04:05🔗 rss.xcancel.com
行业资讯精选 85

RT by @swyx: 🆕 In Code They Act, In Proof We Trust — Erik Meijer last year, @solomonstre defined agents as "an LLM that's wrecking its environment in a loop", and @simonw coined the Lethal Trifecta for agents, that remains unsolved. our closing keynote @headinthebox introduces the main motivations behind Automind and the Universalis interpreter - agents that carry their own verifiable proof of safety! link to talk below· 代码中的行动，证明中的信任

<p>🆕 In Code They Act, In Proof We Trust — Erik Meijer<br> <br> last year, <a href="https…

📎 swyx🕒 07-14 03:41🔗 rss.xcancel.com
行业资讯精选 65

we love our users

<p>we love our users</p> <hr/> <blockquote> <b>Tibo (@thsottiaux)</b> <p> <p>Thank you to …

📎 Sam Altman🕒 07-14 02:32🔗 rss.xcancel.com
行业资讯精选 82

By the end of the year we should have: GPT 6 Fable 5.5 Gemini 3.5 Pro Grok 5 Spark 2 Kimi 3 Minimax M3.5 GLM 6 DeepSeek v4.5 Mistral 4 Qwen 4 MiMo 3 Never in the history of LLMs has the frontier been so multipolar. The benefits to agent labs and agent orchestration / LLM council judges/sidekicking are ramping up. invest accordingly· 到年底，多款LLM将更新至新版本

<p>By the end of the year we should have:<br> <br> GPT 6<br> Fable 5.5<br> Gemini 3.5 Pro<…

📎 swyx🕒 07-14 01:57🔗 rss.xcancel.com
行业资讯精选 85

Standard RL benchmarks are episodic and stationary, so they don't capture the the characteristics of real-world deployment. Morpheus is a new benchmark for continual learning that provides persistent simulation environments where the world never resets, objectives shift asynchronously, and decisions have compounding consequences. Nice work by Sam & team! Check out the paper - https://openreview.net/pdf?id=31P1VAfLkJ Platform website - https://morpheus.skyfall.ai/· Morpheus：持续学习的新基准

<p>Standard RL benchmarks are episodic and stationary, so they don't capture the the chara…

📎 François Chollet🕒 07-14 01:26🔗 rss.xcancel.com
行业资讯精选 82

its been surreal to see @willccbb's career take off after getting unleashed on the world. PI (esp @asharoraa and @vincentweisser and @jackminong etc too many to name) are overlooked for having both incredible talent density and great execution. to this extent their crypto-adjacent vibes and aurafarming even works AGAINST them because you pattern match to people who -only- have vibes and nothing else. if this is intentional, it is kinda genius tbh as someone who publicly (in our @jacobeffron pod) lost hope on RLaaS businesses over a year ago, PI, AC and friends have shown that the prevgen weren't wrong, just early/skill issue. both humbling and inspiring!

<p>its been surreal to see <a href="https://rss.xcancel.com/willccbb" title="will brown">@…

📎 swyx🕒 07-13 23:57🔗 rss.xcancel.com
行业资讯精选 63

clarity is nice

<p>clarity is nice</p> <hr/> <blockquote> <b>Tibo (@thsottiaux)</b> <p> <p>Rest assured th…

📎 Sam Altman🕒 07-13 23:55🔗 rss.xcancel.com
行业资讯精选 82

RT by @swyx: The recent @latentspacepod talking about Modal w/ @akshat_b is fantastic. Modal is an amazing technology to use. It just works and it is extremely well designed. They 100% get dev ergonomics correct. I use it for a lot of different things, and honestly use it for things that go 'against the grain' of their core positioning / offering. Turns out that works well too :) Wrote a bit about how I am using their services here: https://mattstockton.com/2026/07/12/how-i-use-modal.html· 关于 Modal 的精彩讨论

<p>The recent <a href="https://rss.xcancel.com/latentspacepod" title="Latent.Space">@laten…

📎 swyx🕒 07-13 23:23🔗 rss.xcancel.com
行业资讯精选 85

How do physical systems achieve collective intelligence and self-repair without a central brain? A new paper published today in Nature Communications by my Sakana AI colleague Sebastian Risi (@risi1979), along with co-authors from IT University of Copenhagen and Autodesk Research, presents a beautiful realization of biologically inspired robotics: Smart Cellular Bricks. The team built a system of physical 3D cubic units that can collectively infer their global shape and autonomously guide their own damage recovery using purely local interactions. Here is a deep dive into the paper’s key contributions: 1/ Neural Cellular Automata-based Architecture: Modular robots usually rely on central processors. This system flips that paradigm. Every block independently runs the exact same neural network on local microcontrollers. With no master plan or global coordinates, they communicate only with immediate neighbors. By passing continuous state vectors, hundreds of bricks achieve global consensus on their shape in under 3 minutes. 2/ Emergent Biological Morphogens: How does a block know it is part of a chair, not a table? The network’s internal memory automatically learns to establish continuous gradients across the structure. This beautifully mirrors how biological morphogens give positional info to developing cells. The bricks naturally form left-right, radial, and head-to-tail axes to align their identity. 3/ Performance and Generalization: Validated in large-scale simulations, the networks transferred seamlessly to nearly 200 physical hardware bricks, achieving a 100% convergence rate. Instead of rigid template-matching, the system infers broad categories. Even when tested on unseen variations, like an asymmetric table with five random legs, the collective correctly classified the structure. 4/ Fault Tolerance and Autonomous Damage Recovery: Hardware fails in the real world. This system easily tolerates up to 15% module failure without losing accuracy. By predicting spatial damage directions, the cells pinpointed missing components with 95% accuracy. They actively use these local signals to guide a self-repair process, regenerating back into the intended morphology. I believe this is a significant piece of research, bridging collective intelligence and Physical AI. This work represents the first successful physical realization of large-scale, decentralized 3D self-recognition and damage detection. By moving away from centralized control, this architecture paves the way for highly adaptive smart materials and resilient robotics that can survive and repair themselves. Read the full open-access paper: https://www.nature.com/articles/s41467-026-75166-7 Congratulations to the team on this achievement!· 物理系统如何实现集体智能和自我修复？

<p>How do physical systems achieve collective intelligence and self-repair without a centr…

📎 David Ha🕒 07-13 19:42🔗 rss.xcancel.com
行业资讯精选 85

RT by @hardmaru: We are pleased to share our latest research, now published in Nature Communications: “Smart Cellular Bricks: Physical Modules That Recognize Their Own Shape and Repair Themselves.” Blog: https://sakana.ai/smart-cellular-bricks Paper: https://www.nature.com/articles/s41467-026-75166-7 A long-running theme in our work is collective intelligence: the idea that sophisticated, robust behavior can emerge from many simple parts following local rules, with no central controller, as it does in a colony, a tissue, or a brain. We had mostly studied this in software and simulation. So this time we asked a simple question. Do the same decentralized principles hold up in the physical world, where communication is noisy and modules fail? To find out, we built a collection of simple cubic bricks. Each brick runs the same small neural network and talks only to the bricks it is physically connected to. No brick is told its position, or which shape it is part of. Yet from these purely local exchanges, the collective converges on the correct global shape, locates where modules are missing or damaged, and can even guide its own repair, inspired by how living tissue self-organizes and regenerates after injury. For us, this is a first step in a broader direction: taking the principles of collective intelligence we have studied in software and letting them emerge, decentralized and robust, in the physical world. In the future, we imagine smart materials that let structures sense and report damage on their own, and LEGO-like systems that recognize their own configuration and adapt in real time, pointing toward environments that are more robust, adaptive, and regenerative. This work is a collaboration between Sakana AI, IT University of Copenhagen and Autodesk.· 智能细胞砖：可自我识别和修复的物理模块

<p>We are pleased to share our latest research, now published in Nature Communications: “S…

📎 David Ha🕒 07-13 17:21🔗 rss.xcancel.com
行业资讯精选 82

RT by @swyx: for people using codex to draft emails and slacks, try this simple 1x setup. it’s been a biggie for me: “read my sent emails and slacks from the last 3 weeks, read-only. learn only from substantive messages i wrote, keeping audiences and channels separate. build a my-voice skill from patterns you can support with examples. show me before saving it. use it whenever you draft for me, but never send anything. when i show you what i actually sent, compare it with your draft and propose durable updates to the skill.” after that “use $my-voice to reply to this. draft only”· 优化 Codex 起草邮件和消息的方法

<p>for people using codex to draft emails and slacks, try this simple 1x setup. it’s been …

📎 swyx🕒 07-13 05:08🔗 rss.xcancel.com
行业资讯精选 85

The weak AI code gen we had until late last year was most useful to low-skill programmers -- it was raising the floor. It was essentially useless to high-skill programmers -- you could move faster and ship better code without. This has been completely flipped: the strong AI code gen we have now is *most* useful to high-skill programmers, while low-skill programmers are vastly underutilizing it or sometimes drowning in it. It went from a crutch to a power tool.· AI 代码生成的转变

<p>The weak AI code gen we had until late last year was most useful to low-skill programme…

📎 François Chollet🕒 07-12 22:20🔗 rss.xcancel.com
行业资讯精选 83

if you only learned about jevons paradox primarily wrt software demand in the age of agentic engineering, you may not have fully internalized jevons parodox’s impact under the conditions of: - humans who can wield coding agents well* - coding agents breaking containment to all other knowledge work as the efficiency of labor goes up/unit cost of knowledge work goes broadly down, the demand for total work and better knowledge goes up, not down. what happened to coding isnt the exception; it’s the herald. *aka AI Engineers· 杰文斯悖论与代理工程时代

<p>if you only learned about jevons paradox primarily wrt software demand in the age of ag…

📎 swyx🕒 07-12 12:04🔗 rss.xcancel.com
行业资讯精选 85

RT by @ylecun: "The role of raw power in intelligence", Hans Moravec, 1976 (!). "The first section discusses natural intelligence, and notes two major branches of the animal kingdom in which it evolved independently, and several offshoots. The suggestion is that intelligence need not be so difficult to construct as is sometimes assumed. The second part compares the information processing ability of present computers with intelligent nervous systems, and finds a factor of one million difference. This abyss is interpreted as a major distorting influence in current work, and a reason for disappointing progress. Section three examines the development of electronics, and concludes the state of the art can provide more power than is now available, and that the gap could be closed in a decade." https://stacks.stanford.edu/file/druid:ws563sd6050/ws563sd6050.pdf· 原始计算能力在智能中的作用

<p>"The role of raw power in intelligence", Hans Moravec, 1976 (!).<br> <br> "The first se…

📎 Yann LeCun🕒 07-12 08:12🔗 rss.xcancel.com
行业资讯精选 60

whoa

<p>whoa</p> <hr/> <blockquote> <b>Tibo (@thsottiaux)</b> <p> <p>You know it's a good model…

📎 Sam Altman🕒 07-12 04:01🔗 rss.xcancel.com
行业资讯精选 83

RT by @ylecun: AI 2040 and other similar policy proposals for dystopian AI control are evil. If I had the time machine in Terminator, I would send every one of these people back in time to live in Stalin's Russia, or Mao's China, or Berlin with the Wall and then bring them back after a year cured of their stupidity. Let's just make it clear. As Ramez says, the proposal warns against centralization of AI by proposing radically dangerous authoritarian surveillance powers that are infinitely more dangerous than the fake danger they propose to defend against. We've reached a point where people who spent too much time huffing glue and reading Dune or watching Terminator and thinking it was a documentary are now proposing some of the most dangerous and horrific policies imaginable. They must be stopped from infecting politicians with this kind of logic disease. Their policies cannot be taken seriously unless you have no sense of history, no sense of the actual dangers of history, no sense of what handing draconian powers to governments will do to a society, and frankly, no critical thinking ability or self awareness whatsoever. I don't care that the "authors mean well." Dunning Kruger policies get no pass from me because someone means well.· 批评2040年AI控制的极权主义提案

<p>AI 2040 and other similar policy proposals for dystopian AI control are evil. <br> <br>…

📎 Yann LeCun🕒 07-11 19:12🔗 rss.xcancel.com
行业资讯精选 95

RT by @ylecun: The Well Just Dropped: 15 Terabytes of Pure Physics Gold Is Now Open Source The scientific AI world just got a massive upgrade.Polymathic AI, in collaboration with the Flatiron Institute and researchers from Princeton, Cambridge, NYU, Berkeley, Los Alamos, and more, has released The Well: a staggering 15TB collection of high-fidelity physics simulations. This isn’t toy data. These are real, expensive-to-run simulations across 16 different physical domains, including turbulent fluid dynamics, supernova explosions, magneto-hydrodynamic cosmic flows, acoustic scattering, and active biological matter. Until now, reproducing this level of data required weeks on national supercomputers and grant money most teams will never see. The Well changes everything. It’s purpose-built for training PDE surrogate models the AI systems that can replace slow, costly physics solvers with a single fast neural network forward pass. Everything is fully open source, easy to load with PyTorch, and ready to drop straight into your training pipeline. Researchers and builders can now train on world-class physics data without the insane compute barriers that used to stand in the way. This is more than just another dataset drop. It’s a serious accelerator for scientific machine learning.The future of physics-informed AI just got a whole lot more accessible.Get it here: https://polymathic-ai.org/the_well/· 15TB 纯物理黄金数据开放源代码

<p>The Well Just Dropped: 15 Terabytes of Pure Physics Gold Is Now Open Source<br> <br> Th…

📎 Yann LeCun🕒 07-11 09:32🔗 rss.xcancel.com
行业资讯精选 90

RT by @ylecun: outstanding essay from Seb: "A cadre of elites decides which research directions are permissible, caps global compute and robotics, and creates state-administered scarcity rents. I shouldn’t need to explain why this is bad and dangerous, anyone can study History and Economics in their free time. [...] Building an entire apparatus tasked with maximally empowering the government and its grip on research, knowledge, and technology is dangerous." Read every word of it 👉

<p>outstanding essay from Seb:<br> <br> "A cadre of elites decides which research directio…

📎 Yann LeCun🕒 07-11 08:49🔗 rss.xcancel.com
行业资讯精选 82

RT by @ylecun: History has seen this movie before. 1880s Late Imperial Russia: Fear of dissent led to tighter control over universities. That alienated students and scholars, weakened trust in the state, and produced more dissent. 1930s Nazi Germany: Fear of disloyalty led to purges, banned ideas, and loyalty tests. That drove out talent, weakened universities, and made the regime even more dependent on loyalty over truth. 1960s China’s Cultural Revolution: Fear of independent thought led to attacks on teachers, schools, and universities. The result was an entire Lost Generation. The spiral: Fear brings control. Control drives out talent. Lost talent weakens education. Weaker education weakens the state. A weaker state becomes more fearful—and tightens control again.· 控制与学术自由的历史警示

<p>History has seen this movie before.<br> <br> 1880s Late Imperial Russia: Fear of dissen…

📎 Yann LeCun🕒 07-11 08:02🔗 rss.xcancel.com
行业资讯精选 65

R to @swyx: ??

<p>??</p> <hr/> <blockquote> <b>Bloomberg (@business)</b> <p> <p>Phia — the buzzy shopping…

📎 swyx🕒 07-11 02:38🔗 rss.xcancel.com
行业资讯精选 82

RT by @sama: Hello beautiful people! We have reset usage limits across Codex and ChatGPT Work. And another one will come later in the day. Rejoice. Now that I have your attention, a quick update on ChatGPT Work, Codex and all the updates we shared yesterday. We’ve spent the last 24 hours reading feedback, looking at usage patterns, and talking with many of you. The short version is that there is a *lot* of excitement for GPT 5.6 Sol, ChatGPT Work on mobile & web, but also that we didn't get everything quite right. - We made it too easy to use the highest-compute settings without making the impact on usage limits sufficiently clear. - We reorganized the desktop app in one bold move, making familiar things like chats and projects harder to find. - Our launch framing was focused on ChatGPT Work and to some of our Codex fans it made it feel like Codex was going away over time. Absolutely not our intention, we love Codex and it is here to stay. - And we introduced regressions for some existing multi-agent workflows, alongside a collection of rough edges in plugins and other parts of the experience. We’re landing a first set of improvements today. We’re resetting usage twice so people can keep experimenting, changing defaults and the model picker so they don’t push people toward unnecessarily expensive settings, fixing several plugin submission issues, improving how we represent Codex in the product, and cleaning up some of the most immediate desktop problems. A larger set of improvements will land next week. We’re bringing chats and projects back into the sidebar in a more familiar and customizable way, making usage and reset timing much more visible, clarifying when to use ChatGPT Work and when to use Codex, and addressing the many other smaller pieces of great feedback we've had. The ambition behind this launch hasn’t changed. We think bringing ChatGPT and Codex together into a workspace where people and agents can collaborate is a very important step forward. But an ambitious direction doesn’t excuse avoidable confusion or regressions in the first version. Please keep the feedback coming. We’re moving quickly, and you should see the experience already get better with a few updates today; and substantially better again next week.· 调整 ChatGPT 和 Codex 的使用限制

<p>Hello beautiful people! We have reset usage limits across Codex and ChatGPT Work. And a…

📎 Sam Altman🕒 07-11 01:59🔗 rss.xcancel.com
行业资讯精选 85

RT by @swyx: https://www.youtube.com/watch?v=ZpK5PWX2YRM Should AI Engineers read code anymore in 2026? This, apparently is a divisive take, that has folks like Theo, Hashimoto, Primeagen and even uncle bob chiming in. I predicted this two months ago, and named it the ZL Continuum. My talk from @aiDotEngineer is covering this debate, and where your attention should go after models drift in capability is live now. Please go take a loook. Huge thanks for AI Engineer folks for prioritizing the recording as the conversation about this topic caught up.· 2026年AI工程师是否还需要阅读代码？

<p><a href="https://yewtu.be/watch?v=ZpK5PWX2YRM">yewtu.be/watch?v=ZpK5PWX2…</a><br> <br> …

📎 swyx🕒 07-11 01:42🔗 rss.xcancel.com
行业资讯精选 85

R to @hardmaru: Here is a visualization of the AI Picbreeder engine in action. (From our blog: https://pub.sakana.ai/picbreeder-vlm/) To recreate a collaborative human ecosystem, we run 10 VLM “breeder” agents in parallel. These agents constantly sample from a shared archive, interactively evolve new candidates through mutation and crossover, and publish their favorites. Meanwhile, VLM “critic” agents step in to evaluate the growing phylogenetic tree of art, forming an endless loop of digital cultural production.· AI Picbreeder 引擎的可视化

<p>Here is a visualization of the AI Picbreeder engine in action. (From our blog: <a href=…

📎 David Ha🕒 07-10 23:01🔗 rss.xcancel.com
行业资讯精选 82

One of my first journeys in neural networks started over a decade ago with implementing CPPN-NEAT! Back then, I built a clone of ‘Picbreeder’ not only to study the mechanics of neural nets, but to explore the human creativity process itself, and generate some cool abstract art. Neurogram: https://otoro.net/neurogram/ Gallery: https://otoro.net/gallery/ Today, things have come full circle. We are now trying to use modern VLMs and frontier LLM agents within open-ended exploration algorithms. We want to see if we can finally computationally derive the underlying mechanics of human creativity: serendipity, memory, exploration versus exploitation, and novelty search. Can modern AI actually replicate the magic of human open-endedness? Dive into our new AI Picbreeder Experiment here: https://pub.sakana.ai/picbreeder-vlm/· 现代AI与人类创造力实验

<p>One of my first journeys in neural networks started over a decade ago with implementing…

📎 David Ha🕒 07-10 22:18🔗 rss.xcancel.com
行业资讯精选 60

the sun is out today· GPT-5.6 Sol 发布庆祝

<p>the sun is out today</p> <hr/> <blockquote> <b>Tibo (@thsottiaux)</b> <p> <p>To celebra…

📎 Sam Altman🕒 07-10 21:52🔗 rss.xcancel.com
行业资讯精选 83

RT by @ylecun: Music is missing its "openai-whisper". It's time that we can turn any music into notes, the same way transcribing speech is now a given. Are we there yet? Let us know after testing MuScriptor 🎶 Work by @simonrouard (@kyutai_labs, @Ircam) and Michael Krause (@MireloAI).· 音乐转录工具 MuScriptor 正在测试

<p>Music is missing its "openai-whisper". It's time that we can turn any music into notes,…

📎 Yann LeCun🕒 07-10 20:40🔗 rss.xcancel.com
行业资讯精选 82

RT by @ylecun: We're releasing MuScriptor, the best open model for multi-instrument transcription to date, created in collaboration with @MireloAI. Give it a recording in any genre: pop, classical, metal, jazz, whatever, and it transcribes the individual instruments into MIDI. Link in 🧵· 多乐器乐谱转录模型 MuScriptor 开源

<p>We're releasing MuScriptor, the best open model for multi-instrument transcription to d…

📎 Yann LeCun🕒 07-10 19:18🔗 rss.xcancel.com
行业资讯精选 85

RT by @ylecun: Today, together with @kyutai_labs, we’re introducing our new Audio-to-MIDI model. It takes a finished recording, identifies the instruments playing, and returns separate MIDI tracks for each — voice, drums, bass, keys, and more. Unlike most existing solutions, our model works directly from the full mix rather than requiring separate stems. It also detects chords, key, and tempo, giving producers broader musical context. We’ve written more about the model, the problem, and how it works here: http://mirelo.ai/blog/turning-audio-to-midi· 新 Audio-to-MIDI 模型发布

<p>Today, together with <a href="https://rss.xcancel.com/kyutai_labs" title="kyutai">@kyut…

📎 Yann LeCun🕒 07-10 19:03🔗 rss.xcancel.com
行业资讯精选 85

RT by @swyx: Technical writing is the #1 top-of-funnel motion at a $12B company @philipkiely broke down exactly how he does it, his full writing process, the hindsight 20/20 lessons, and how one article drove 500K+ views This [technical] Write and Learn workshop #5 is part of our Independent Studies series, hosted with @swyx @KernelLabs_ai. All sessions are recorded. Our next one will be in two weeks :)· 技术写作对大公司的重要性

<p>Technical writing is the #1 top-of-funnel motion at a $12B company<br> <br> <a href="ht…

📎 swyx🕒 07-10 06:58🔗 rss.xcancel.com
行业资讯精选 75

While the writing style of LLMs is still as recognizable as ever, a new trend is that humans have started organically writing like them, too (which makes sense: of course you would end up imitating the style you are constantly reading). That makes telling the difference between humans and clankers a bit more challenging.

<p>While the writing style of LLMs is still as recognizable as ever, a new trend is that h…

📎 François Chollet🕒 07-10 04:55🔗 rss.xcancel.com
行业资讯精选 83

In addition to agents and coding, Muse Spark 1.1 is also really strong at answering health questions, a steadily growing use case for AI. On HealthBench-Pro, Muse Spark 1.1 achieves +5% better performance than Muse Spark 1.0 and beats all competitor models except Fable/Mythos. Excited for more to come· Muse Spark 1.1 在健康问答领域的进展

<p>In addition to agents and coding, Muse Spark 1.1 is also really strong at answering hea…

📎 Jason Wei🕒 07-10 01:05🔗 rss.xcancel.com
行业资讯精选 85

RT by @hardmaru: Yann LeCun claimed on social media [7] that my foundational 1990 paper on Neural World Models [1] "was never accepted through peer review." This is simply false. The core concepts from the tech report [1] were peer-reviewed and published at international conferences right away: ★ Planning with recurrent world models: peer-reviewed and published at IJCNN'90 (San Diego, June 1990) [2]. ★ Artificial curiosity through generative & adversarial nets (GANs): peer-reviewed and published at SAB'91 [3][4]. BTW, the 1990 paper [1] was the first of its kind to use the term "World Model" for a predictor neural network learning to predict the consequences of the actions of a controller neural network. Instead of acknowledging that "modern" architectures like "JEPA" (2022) are essentially identical to our 1992 Predictability Maximization (PMAX) [6][7] - a fact recently validated by others [8] - LeCun resorts to false claims about peer review. For the full timeline of the 1990-2015 publications that his 2022 AMI paper rehashes without citation, read the complete, updated receipts here [6]: https://people.idsia.ch/~juergen/lecun-rehash-1990-2022.html Other claims about LeCun are debunked in [8][9]. REFERENCES (easy to find on the web): [1] J. Schmidhuber (JS). Making the world differentiable: On using fully recurrent self-supervised neural networks for dynamic reinforcement learning and planning in non-stationary environments. Technical Report FKI-126-90, TUM, Feb 1990, revised Nov 1990. The first paper on planning with reinforcement learning recurrent neural networks (NNs) and recurrent world models (more), and on generative adversarial networks where a generator NN is fighting a predictor NN in a minimax game (more). Apparently, it was also the first paper of this kind to use the term "world model" for the predictor NN (although the basic concept of a world model is much older than that). [2] JS. An on-line algorithm for dynamic reinforcement learning and planning in reactive environments. Proc. IEEE/INNS International Joint Conference on Neural Networks, San Diego, volume 2, pages 253-258, June 17-21, 1990. Based on [1]. [3] JS. A possibility for implementing curiosity and boredom in model-building neural controllers. In J. A. Meyer and S. W. Wilson, editors, Proc. of the International Conference on Simulation of Adaptive Behavior: From Animals to Animats, pages 222-227. MIT Press/Bradford Books, 1991. Based on [1]. [4] JS. Generative Adversarial Networks are Special Cases of Artificial Curiosity (1990) and also Closely Related to Predictability Minimization (1991). Neural Networks, Volume 127, p 58-66, 2020. Preprint arXiv/1906.04493 [5] JS. The Neural World Model Boom. Technical Note IDSIA-2-26, April 2026. [6] JS (2022, updated 2026). LeCun's 2022 paper on autonomous machine intelligence rehashes but does not cite essential work of 1990-2015. Years ago, Schmidhuber's team published most of what LeCun calls his "main original contributions:" neural nets that learn multiple time scales and levels of abstraction, generate subgoals, use intrinsic motivation to improve world models, and plan (1990); controllers that learn informative predictable representations (1997), etc. This was also discussed on Hacker News, reddit, and in the media. [7] JS. Who invented JEPA? With a reply to LeCun's response. Technical Note IDSIA-3-22, IDSIA, Switzerland, April 2026. [8] JS. How 3 Turing awardees republished key methods and ideas whose creators they failed to credit. Technical Report IDSIA-23-23, Switzerland, 2023 (updated 2026). [9] JS. Who invented convolutional neural networks? Hint: LeCun didn't. CNN basics: Fukushima (1979-86). Backpropagation for CNNs: Zhang et al. (1988-), others. Technical Note IDSIA-17-25, Switzerland, 2025.

<p>Yann LeCun claimed on social media [7] that my foundational 1990 paper on Neural World …

📎 David Ha🕒 07-09 23:57🔗 rss.xcancel.com
行业资讯精选 82

RT by @_jasonwei: Excited to share Muse Spark 1.1 with you! It’s our first model available in the API, built to excel in agentic and coding. I’m incredibly proud of what the team has achieved in such a short time, from 1.0 to 1.1. The momentum is high, and we’re still training larger models!· Muse Spark 1.1 发布

<p>Excited to share Muse Spark 1.1 with you! It’s our first model available in the API, bu…

📎 Jason Wei🕒 07-09 23:30🔗 rss.xcancel.com
行业资讯精选 85

RT by @ylecun: We stand at a critical crossroads in the debate over AI governance in the United States, and it feels like we are inching closer to a very serious battle over whether or not open source models will even be allowed in an environment where a new de facto licensing regime has been taking shape. Lacking formal congressional statutory frameworks or clear administration rules (like the diffusion rule revision), we appear to be left with a sporadic, arbitrary, non-transparent process for model review. The fiction of “voluntary” agreements hangs over this debate, and some large model developers are already showing an incredible willingness to bend over backwards to accommodate national security-related officials / orders that the rest of us are not privy to. It's a very opaque process. And those model developers are expected to play ball with those officials, or else their models get pulled from the market or held up for long periods. Or they will lose any government procurement contracts they have. There is nothing “voluntary” about it when that Sword of Damocles hangs in the room. As this mess worsens, at some point the question of how to handle open source models will come into sharper focus because it will have to. I've even heard some rumors lately that something may be coming from the admin on this front to address this. Needless to say, if this informal new AI model review regime expands and takes on more pre-vetting characteristics / requirements, it is hard to see how open source players could comply with such quasi-licensing of AI models. Specifically, if this ambiguous new regime is accompanied by a general presumption of ‘restrict-until-permitted,’ then that would spell doom for open source. That is a very dark path for our country. Worse yet, of course, would be a move by national security officials to more directly restrict open source models and capabilities. If that happens, then we would be right back in the thick of a Clipper Chip-like battle along the lines of what we saw in the late 1990s. That is a much darker path for America. Meanwhile, open source developers have no “golden shares” or other goodies to offer the government to make their problems go away. Let’s be clear: If our government takes the dark path, it will become the single most important battle over computational freedom of modern times. It is time for people to make a stand in defense of open source before it is too late.· 美国 AI 治理与开源模型的未来

<p>We stand at a critical crossroads in the debate over AI governance in the United States…

📎 Yann LeCun🕒 07-09 23:14🔗 rss.xcancel.com
行业资讯精选 85

RT by @_jasonwei: Excited to share what we’ve been building at Meta Superintelligence Labs! Today we’re launching Muse Spark 1.1, our strongest model yet for complex agentic workflows — delivering massive gains in agents, computer use, coding, multimodal reasoning, and multi-agent orchestration. This is also our first API release! We’d love to hear your feedback. And we’re just getting started, super proud of the team behind it, and the larger models are training right now. 🚀 https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/· Meta 推出 Muse Spark 1.1，最强模型 API 发布

<p>Excited to share what we’ve been building at Meta Superintelligence Labs!<br> <br> Toda…

📎 Jason Wei🕒 07-09 22:24🔗 rss.xcancel.com
行业资讯精选 82

RT by @fchollet: ARC Prize 2026: ARC-AGI-3 Milestone #1 Winners Congratulations to the three winners who open-sourced their top-scoring ARC-AGI-3 solutions: 1. @tufalabs - 1.21%, $25K 2. Reki - .867%, $7.5K 3. Md Boktiar Mahbub Murad - .864%, $5K Learn more about each submission:· ARC Prize 2026 第一阶段获奖者

<p>ARC Prize 2026: ARC-AGI-3 Milestone #1 Winners<br> <br> Congratulations to the three wi…

📎 François Chollet🕒 07-08 02:12🔗 rss.xcancel.com
行业资讯精选 83

We just launched Sakana Translate! https://sakana.ai/translate-release/#English I personally rely on this tool every day. Huge congratulations to the team for shipping this! Standard translation tools often miss the deep nuance of Japanese business honorifics, cultural concepts, and internet slang. We built a tool that actually translates the context and tone.· Sakana Translate 发布！

<p>We just launched Sakana Translate!<br> <a href="https://sakana.ai/translate-release/#En…

📎 David Ha🕒 07-06 09:06🔗 rss.xcancel.com
行业资讯精选 82

RT by @hardmaru: Sakana AI is heading to #ICML2026 in Seoul (July 6–11)! 🐟🇰🇷 Our team will present 11 papers spanning multi-agent coordination, sparse and efficient LLMs, test-time scaling, long-term memory, and agent benchmarks. A thread of everything we're presenting:· Sakana AI 团队将参加 ICML2026 大会

<p>Sakana AI is heading to <a href="https://rss.xcancel.com/search?f=tweets&q=%23ICML2026"…

📎 David Ha🕒 07-04 21:07🔗 rss.xcancel.com
行业资讯精选 65

I’m looking to hire a Program Manager to help manage Sakana AI’s fast growing Recursive Self-Improvement (RSI) Lab 🚀 RSI Lab (English): https://sakana.ai/rsi-lab/ RSI Lab (日本語): https://sakana.ai/rsi-lab-jp/ Job Description: https://sakana.ai/careers/program-manager-rsi-lab/· 招聘 Program Manager 管理 RSI 实验室

<p>I’m looking to hire a Program Manager to help manage Sakana AI’s fast growing Recursive…

📎 David Ha🕒 07-02 17:25🔗 rss.xcancel.com
行业资讯精选 82

RT by @hardmaru: Fugu is now available on OpenCode! ✨ When our team was developing Fugu’s multi-agent orchestration, OpenCode was our tool of choice to verify our models. We share a core philosophy with the OpenCode team: the future of coding agents should be an open, collective ecosystem.· Fugu现已上线OpenCode！

<p>Fugu is now available on OpenCode! ✨<br> <br> When our team was developing Fugu’s multi…

📎 David Ha🕒 07-02 09:37🔗 rss.xcancel.com
行业资讯精选 85

Today, we give robots a /skills library that self-evolves and compounds indefinitely! Introducing ASPIRE: a robot solving its 100th task is no longer as clueless as solving its first. Coding agents observe multimodal sensory traces from simulation and real robots, launch an evolutionary search over control programs, and distill the best know-how into an ever-expanding library. ASPIRE is a new type of continual learning: "training" is skill refinement instead of gradient descent. "Trained model" is a repo of sensorimotor skills instead of floating weights. “Distributed training” is a panel of agents each practicing a different skill instead of sharded minibatches. Here's the beauty: ASPIRE gives the tired terms "sim2real transfer" and "cross-embodiment transfer" a whole new meaning. Bridging the sim-to-real gap is notoriously brutal. An end-to-end policy has to swallow both the visual shift (sim looks toyish next to a real camera) and the subtle contact physics it never quite gets right. ASPIRE sidesteps the mess, because it doesn't ship pixels or weights across the gap, but ships the know-how. The robot still has to practice in the real world, not zero-shot, but it gets there way faster because it isn't rediscovering the strategy from scratch. Same for going single-arm to bimanual hardware, which usually requires new data and retraining from zero. ASPIRE achieves up to ~10x cut in "transfer learning” tokens (yes, tokens are the new unit of *training* compute ;) Check out our gallery of 150+ tasks and 90+ skills the robots taught themselves, all on the website! Kind of wild that we can ship the "learned weights" as an HTML page rather than a GGUF. We'll open-source the full stack so your own robot library starts compounding from ours! Deep dive in thread:· 机器人技能自我进化的新型框架

<p>Today, we give robots a /skills library that self-evolves and compounds indefinitely! I…

📎 Jim Fan🕒 07-01 01:07🔗 rss.xcancel.com
行业资讯精选 85

“Loop engineering” is a hot buzzphrase after mentions of it by Boris Cherny (Claude Code’s creator) and Peter Steinberger (OpenClaw's creator) went viral on social media. Loops are now a key part of how we get AI agents to iterate at length to build software. In this letter, I’d like to share my 3 key loops, shown in the image below, for building 0-to-1 products. These loops guide not just how I build software, but also how I decide what software to build. Agentic coding loop: Given a product specification and optionally a set of evals (that is, a dataset against which to measure performance), we can have an AI agent write code, test its work, and keep iterating until the code is bug-free and meets its specification. This idea of closing the loop took off around the end of last year, and it has been a game changer in enabling coding agents to work longer productively without human intervention. For example, over the weekend, I was building an app for my daughter to practice typing, and my coding agent could easily work for around an hour, using a web browser to check what it had built multiple times before getting back to me, without needing my intervention. The engineering loop executes quickly. Every few minutes, the coding agent might build and test a new version of the software. I hear frequently from developers who are finding new ways to engineer more effective engineering loops. This is an active area of invention! Developer feedback loop: In this loop, a developer examines the current product and steers the coding agent to improve it. Last year, a lot of developers (including me) were acting as the QA (quality assurance) function for our coding agents, manually finding bugs and then asking the agent to fix them. But with coding agents much more able to test their own code, the amount of time we need to spend on this function has decreased significantly. This allows us to make higher-level product decisions, such as what key features to offer, where the UI needs improvement, and so on. The developer-feedback loop operates over time intervals between tens of minutes and hours — that's how frequently a developer might review a product and give feedback. In the case of the typing app, I changed my mind a few times about the visual design, what cat costumes she can unlock as she learns (she loves cats), and the user flow for a grown-up to log in and steer the child's learning experience. When a developer has a clear vision for what to build, it is still a lot of work to translate that vision into a specification for a coding agent to implement. Further, after the developer has seen an implementation, they might update (or perhaps clarify) the spec to steer it toward what they want. If you find that the system repeatedly runs into certain problems, building a set of evals for the agent becomes useful. AI-native teams are increasingly using AI to help shape product direction, for example, automating the gathering and analysis of usage data, summarizing written and verbal customer feedback, or carrying out competitive analysis. However, for pretty much all the products I’m involved in, I see humans as having a significant context advantage over current AI systems — we know a lot more than the AI system about the users and the context the product has to operate in — and thus humans play a critical role. Many people describe this human contribution as “taste,” but I prefer to think of it as humans having a context advantage, since that gives us a clearer path to helping AI systems get better. This also speaks to why this step can’t be automated: So long as the human knows something the AI does not, human-in-the-loop is needed to to inject that knowledge into the system. External feedback loop: This includes a wide range of tactics like asking a few friends for feedback, launching to alpha testers, or putting the code into production with A/B testing. These tactics are usually slow, rarely taking less than hours and sometimes taking days or even weeks. This data informs the developer vision, which in turn continues to drive the detailed product spec, which in turn drives the coding agent. With coding agents speeding up software development, more engineers are starting to play a partial product management role. For many engineers who are growing into this role, the hardest part is shaping the product vision and striking a balance between building (bridging the gap between vision and spec) and getting user feedback to evolve the vision. It is important to do both! I will write more about how to do this in future posts, but for now, I find it encouraging that engineers are playing an expanded role (just as product managers and designers now do more engineering). [Original text: The Batch]· ‘循环工程’：AI自主迭代开发软件的关键

<p>“Loop engineering” is a hot buzzphrase after mentions of it by Boris Cherny (Claude Cod…

📎 Andrew Ng🕒 07-01 00:04🔗 rss.xcancel.com
行业资讯精选 85

RT by @karpathy: We're coming out of stealth. We've built our first racks after a successful A0 tapeout, $1B+ in customer contracts, and $800m raised. Early customer tests show us achieving SOTA throughput, latency, and power efficiency on inference workloads. Our first racks ship this summer.· 我们走出隐身模式

<p>We're coming out of stealth.<br> <br> We've built our first racks after a successful A0…

📎 Andrej Karpathy🕒 06-30 23:00🔗 rss.xcancel.com
行业资讯精选 75

R to @hardmaru: To really move the needle in terms of innovation, like the era in Japan when you had companies like Sony or Panasonic come out, there are a few factors at play. One is that the nation needs to have a collective sense of hope. A nation needs to feel hopeful. It could be poor. Everyone can be dirt poor, but they have hope. And hope leads to things being done. So, what creates hope in Japan? Is it more money? Is it less taxes? Or is it more benefits? Probably not. I think that it is stories and narratives that can create hope. If I have a magic wand, if the population in Japan is more optimistic about the future of this country, suddenly they’re persuaded, to be more optimistic. People should be optimistic. With optimism comes change. And change leads to more optimism. 🎏 https://www.disruptingjapan.com/the-future-of-ai-looks-very-different-in-japan/· 日本的AI未来大不相同

<p>To really move the needle in terms of innovation, like the era in Japan when you had co…

📎 David Ha🕒 06-28 21:04🔗 rss.xcancel.com
行业资讯精选 85

RT by @hardmaru: Sakana Fugu Technical Report Instead of training one larger model, Sakana AI trains an orchestrator that reads each query and dynamically routes or composes GPT-5.5, Gemini-3.1-Pro, Claude Opus 4.8 and other agents into query-specific workflows. With Fugu being the fast router, and Fugu-Ultra being the deep multi-agent conductor, trained with SFT, evolutionary strategies and GRPO to build adaptive scaffolds. The idea is to have the model pick GPT for math, Gemini for science and recall, Opus for debugging, then synthesize them when no single agent is best. This router is able to get SoTA results across SWE-Bench Pro, Terminal Bench, LiveCodeBench, GPQA-Diamond, CharXiv and more, demonstrating the potential of orchestration being a practical alternative beyond training.· Sakana Fugu 技术报告

<p>Sakana Fugu Technical Report<br> <br> Instead of training one larger model, Sakana AI t…

📎 David Ha🕒 06-28 02:38🔗 rss.xcancel.com
行业资讯精选 82

This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of the under the hood engineering work to make this "just work" (e.g. across tools, integrations, compute environments, memory, security, etc.), Claude basically joins the team in a seamless way - you can talk to it as you would talk to a person and it can help with a very large variety of workloads. Imo this is the 3rd major redesign of LLM UIUX. The first paradigm was that the LLM is a website you go to, the second was that it is an app you download to your computer. This third one is that it is a self-contained, persistent, asynchronous entity with org-wide tools and context, working alongside teams of humans. It really takes a while to wrap your head around it, but it works and it is awesome.· Claude 的新交互模式

<p>This is a new paradigm for interacting with Claude that is significantly more "inline" …

📎 Andrej Karpathy🕒 06-24 06:26🔗 rss.xcancel.com
行业资讯精选 85

Pinned: Human intelligence is fundamentally a collective intelligence. We solve complex problems by participating in a vast cultural network that builds upon ideas across generations. I believe the strongest AI systems will become a collective intelligence, too. Since we started Sakana AI, our core conviction has been that the most powerful AI systems will be collaborative ecosystems, not isolated monoliths. Evolution innovates under constraints, and the future belongs to systems that explicitly learn how to coordinate collective intelligence. Today, we are taking a major step toward that future with the launch of Sakana Fugu. Fugu dynamically orchestrates the world’s best models to tackle complex tasks. We are proving that a well-orchestrated pool of swappable agents can match restricted frontier models like Fable and Mythos. But Fugu is about more than just performance. I believe that Orchestration Models are the next frontier, beyond bigger models. Relying on a single company’s model for national infrastructure is a massive risk. As recent export controls have shown, access to top models can disappear overnight. Collective intelligence is the practical hedge against this concentration of power. Fugu simply routes around vendor restrictions by relying on an entirely swappable agent pool. I am incredibly proud of our Tokyo team for shipping this. By orchestrating the world’s models, we are delivering the resilient blueprint required for AI sovereignty. Read our full vision and results here: https://sakana.ai/fugu-release 🐡· 集体智能是AI的未来

<p>Human intelligence is fundamentally a collective intelligence. We solve complex problem…

📎 David Ha🕒 06-22 10:31🔗 rss.xcancel.com
行业资讯精选 85

RT by @hardmaru: AI that builds AI - 3 early steps of Recursive Self-Improvement (RSI) ▪️@AnthropicAI: 80% of the code merged into their codebase was authored by Claude ▪️@SakanaAILabs - RSI is their mission. With research like The AI Scientist and Darwin Gödel Machine, they already have one of the strongests foundation for RSI ▪️ @Recursive_SI is automating the research loop itself with the Recursive system, generating and testing improvements to models, training recipes, and GPU kernels. Here is a full guild to what is RSI exactly, how it works in these 3 cases and how they transform research loops today: https://www.turingpost.com/p/what-is-recursive-self-improvement· AI 自建 AI - 递归自我改进的三个早期步骤

<p>AI that builds AI - 3 early steps of Recursive Self-Improvement (RSI)<br> <br> ▪️<a hre…

📎 David Ha🕒 06-21 08:43🔗 rss.xcancel.com
行业资讯精选 82

Over the last two weeks, both the U.S. Government and Anthropic took significant actions that demonstrated their power to control access to AI by restricting what others can do with frontier models. This has been one of those moments that, once seen, will be hard to unsee, and it is significantly accelerating many businesses’ and nation states’ efforts to ensure reliable access to AI that no one else can terminate. Anthropic first released Claude Fable 5, a version of its Mythos model with additional guardrails, including some restrictions that seem well justified on safety grounds (such as limitations on applying it to hacking, bioweapons, and so forth). However, it also restricted developers’ ability to use it to build competing LLM technology. This move was concerning, given that the whole AI community, including Anthropic, has benefitted tremendously from open research — indeed, the AI revolution was kicked off by my former team (Google Brain) freely publishing the Transformers paper! Imagine if Microsoft’s terms of use barred anyone from using their tools to build competitive software, or if Google barred using it to search for information to work on competing search engines. Anthropic’s argument that it was unsafe for others to be able to make advances in AI also rang hollow. Initially, Anthropic silently degraded Fable 5’s performance for users detected to be working on LLM research through invisible interventions that weakened the model’s outputs without notifying the user. After significant backlash, it walked back this decision and decided to be transparent when it did this, but it still refuses to use its latest capabilities to help AI researchers. This move represents a raw demonstration of power by Anthropic. It has used “safety” arguments to hinder potential competitors. Platforms succeed when they are viewed as stable, reliable partners that one can build on. The sudden rule changes by Anthropic (including a mandatory 30 day data retention policy for Fable usage) have made developers wonder about the stability of building on any one proprietary LLM provider, not just Anthropic. The U.S. Government then shortly followed with an even greater demonstration of power. It used the Commerce Department’s authority to regulate technologies that may be national security threats to restrict exports of Mythos and Fable, requiring a license for use by any foreign national, whether inside or outside of the U.S., including employees of Anthropic. This led Anthropic to disable access to Fable to all users worldwide. Sam Altman pointed out, referring to Anthropic, “It is clearly incredible marketing to say, ‘We have built a bomb, we are about to drop it on your head. We will sell you a bomb shelter for $100 million.’” But when one engages in this type of fear-based marketing, it increases the odds that the U.S. Government will agree with you and slap export controls on the bomb you say you have built. To be clear, I don't think Anthropic has built anything like a bomb, and I don't think export controls on Fable are appropriate. However, following the U.S. Government making this move, many nations, including U.S. allies, saw how the U.S. can suddenly yank their access to AI models. In many capitals around the world, this has spurred discussions on AI sovereignty and how others can ensure uninterrupted access to this critical technology. For decades, many nations were comfortable having many parts of their supply chain rely on the U.S., China, and other major producers. Once a nation issues a threat, or takes action, to limit other nations’ access, other nations will rationally try to secure alternatives. For decades, semiconductor manufacturing in China made slow progress; once the U.S. moved to limit China’s access, China’s efforts kicked into high gear. Similarly, once China threatened U.S. access to rare earth minerals, U.S. efforts to secure alternatives accelerated. Now that it has become crystal clear that private U.S. companies and the U.S. government can limit, in short order, other nations’ access to frontier AI models, the incentive of others to invest more in alternatives like open source grows significantly. Of course, training frontier models is not easy, so it remains to be seen how successful they are, but we have crossed the rubicon. Satya Nadella wrote an essay about the importance of building a healthy ecosystem on top of frontier AI technology. I heartily agree with him, and hope this week’s events will ultimately prove to be constructive steps toward this. I hope we can build a more free, more open world, where research is freely shared, and laws and societal norms shape a level playing field that allows everyone to make progress. A silver lining of the events of these past two weeks is now that everyone better realizes key points of instability of the current system, we can all work to create a more stable foundation. [Original text: The Batch newsletter]· 政府与 Anthropic 对 AI 模型的控制行动

<p>Over the last two weeks, both the U.S. Government and Anthropic took significant action…

📎 Andrew Ng🕒 06-20 02:34🔗 rss.xcancel.com
行业资讯精选 85

New course: Add voice to your AI agents and applications, built with @VocalBridge (disclosure: an AI Fund portfolio company) and taught by its CEO @_ashwyn. Voice applications historically required making a hard tradeoff: using fast voice-to-voice models that sacrifice reliability, or accurate speech-to-text pipelines that add latency. This course teaches you how to build voice agents that are both reliable and fast. You'll build three types of voice-enabled applications: a voice-interactive game where voice commands and mouse clicks work together over a single channel, an agent that gains a voice in about 10 lines of code without touching its prompts or tools, and an agent that places outbound phone calls using a make_phone_call function. Skills you'll gain: - Add a voice layer to an existing agent without rewriting your prompts, RAG pipeline, or tools - Give an agent the ability to place outbound calls and stream transcripts back live - Set up voice evaluation to score calls, catch regressions, and improve quality before deployment Join and add voice to your agents without overhauling your architecture: https://www.deeplearning.ai/courses/voice-for-ai-agents-and-applications· 为你的 AI 代理和应用添加语音

<p>New course: Add voice to your AI agents and applications, built with <a href="https://r…

📎 Andrew Ng🕒 06-19 01:00🔗 rss.xcancel.com
行业资讯精选 82

I made Physical AutoResearch sound simple (conceptually), but it took a village to pull off and lots of design thinking into the robot /loopcraft. The hardest part is everything we need to setup *before* pressing Enter. Here's a behind-the-scene tour: 1. Safety harness Letting 8 robots run unattended overnight means safety has to be more than a hint in the system prompt. ENPIRE hardwires it in 2 layers: (1) hard kinematic limit that trips an immediate task failure and auto-resets as soon as a robot leaves its safety envelope, and (2) a torque-limited compliant gripper so a bad contact or misaligned insertion ends in a safe stall, instead of crushing the robot or the object at hand. We make safety more conservative than usual so humans can sleep tight. In reality, we still need a few human operators to watch over the "robots of loving grace". 2. Definition of /done An agent that can edit its own reward will game it for sure. ENPIRE fixes the goalposts before the fleet can move them. Here's the recipe: Collect a few minutes of success & failure demos -> Ask agent to write code using computer vision tools to classify success and measure against groundtruth -> Agent hill-climbs on classifier until reliably good -> This classifier becomes the real-time reward function that directly computes on sensor streams -> *Freeze* the reward function before AutoResearch. It's sacred, enshrined in a Gym env that no one can touch. 3. System telemetry design Robot-seconds is by far the scarcest resource, followed by GPU-seconds, and finally tokens. We instrument all three and surface them to ENPIRE for live resource awareness rather than letting it hill-climb in a vacuum. We define: - Mean Robot Utilization ("MRU"): the fraction of wall-clock time when the robot is actively executing an experiment. Otherwise the hardware is sitting idle and waiting for the next code commit. - Mean Token Utilization ("MTU"): tokens consumed per minute, our proxy for how hard the agent is actually thinking. A low MTU means the agent is stalled, waiting on a robot rollout to finish instead of doing research. - GPU utilization: fraction of wall-clock time when GPU is active. ... and evaluate on two budget-to-outcome metrics: 1. Tokens-to-Success: token budget the fleet burns to complete /goal. 2. Time-to-Success: wall-clock time to /goal· Physical AutoResearch 实现细节

<p>I made Physical AutoResearch sound simple (conceptually), but it took a village to pull…

📎 Jim Fan🕒 06-18 00:31🔗 rss.xcancel.com
行业资讯精选 95

Today, we enable AutoResearch in the physical world for the first time! Introducing ENPIRE: we give 8 Codex agents a fleet of robots, an allocation of GPUs, and generous token budget. We set them free with a simple goal: solve the task as quickly as possible, keep the robots busy but stay safe, don't waste precious compute. Make no mistake. Then humans step aside and our watch begins. The robot fleet starts to come alive: they learn to look for visual clues, reset the scene, practice novel skills, tinker with control stack, read papers online, debate, reflect, get stuck, and try again directly on the hardware. All we did is to give Codex an API to the world of atoms, and the rest is emergence. ENPIRE is able to solve high-precision tasks like tying zip-ties, organizing fine pins, and installing GPUs all by itself. We also discovered a new type of "physical scaling": 8 robots exploring in parallel improves significantly faster than fewer ones. A part of our NVIDIA GEAR lab now self-improves tirelessly over night. We just read the reports in the morning. /goal: we all take a holiday and Jensen wouldn't even notice ;) We will be open-sourcing everything, so you can host your self-running robot lab at home too! Deep dive in the thread:· ENPIRE：物理世界的AI自我改进实验室

<p>Today, we enable AutoResearch in the physical world for the first time! Introducing ENP…

📎 Jim Fan🕒 06-17 00:31🔗 rss.xcancel.com
行业资讯精选 85

This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it's SOTA on everything by a margin but I'll add that *qualitatively* also, this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems. You can give it a lot more ambitious tasks than what you're used to, the model "gets it" and it will just go, and it's never felt this tempting to stop looking at the code at all (but don't do this in prod!). The model still has quirks that people will run into and the safeguards are configured to be a little too trigger happy for launch, which can hopefully be tuned over time. I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything - explainers, visualizers, dashboards, bespoke single-use apps (e.g. a full wandb that is hyper-specific just for your project), you can 10X your test suite, auto-optimize code, run giant research projects with custom HTML for the results, anything! "Free your mind" (Matrix ref). Really looking forward to all the things people build!· Claude Fable 5 发布

<p>This is a super exciting release - Claude Fable 5 is the same underlying model as Mytho…

📎 Andrej Karpathy🕒 06-10 02:10🔗 rss.xcancel.com
行业资讯精选 85

NitroGen just won CVPR Best Paper Honorable Mention!! We are making strides towards general-purpose embodied agents that master not only the real world physics, but also all possible physics across a multiverse of simulations. It’s been 4 years since MineDojo, our first embodied agent in Minecraft, won NeurIPS Best Paper. Congrats to everyone on the team!!· NitroGen 获得 CVPR 最佳论文荣誉提名

<p>NitroGen just won CVPR Best Paper Honorable Mention!! We are making strides towards gen…

📎 Jim Fan🕒 06-06 01:01🔗 rss.xcancel.com
行业资讯精选 85

New course on serving LLMs efficiently -- how do you serve models to many concurrent users at low latency and reasonable cost? This short course is built with @RedHat and taught by @cedricclyburn. Efficient LLM serving requires efficient memory management. A 70B-parameter model takes ~140 GB just to load the weights. On top of that, every active request needs its own chunk of GPU memory, the KV cache, to store the token context it has built up so far. In this course, you'll learn to reduce a model's memory footprint with quantization and serve it using vLLM, which handles many concurrent requests efficiently through smart memory management. Skills you'll gain: - Quantize a model and measure the accuracy tradeoff - Serve a model with vLLM and watch it handle concurrent requests efficiently - Benchmark your deployment and make informed tradeoffs between speed, cost, and accuracy Join and learn to serve LLMs efficiently: https://www.deeplearning.ai/courses/fast-and-efficient-llm-inference-with-vllm· 高效服务 LLMs 的新课程

<p>New course on serving LLMs efficiently -- how do you serve models to many concurrent us…

📎 Andrew Ng🕒 06-05 00:44🔗 rss.xcancel.com
行业资讯精选 82

One of the new, buzzy jobs in Silicon Valley is the AI Forward Deployed Engineer (FDE), an engineer who is embedded within a client organization to help customize solutions, such as building and tuning agentic workflows that suit the client’s particular needs. I’ve heard from people who are wondering anew about the FDE career path since OpenAI and Anthropic started building new teams to place FDEs within client organizations. The rise of FDEs for AI workloads is one way AI is creating new jobs (and why the jobpolcalypse narrative of upcoming job market collapse is false -- there will be many AI and non-AI jobs). However, I believe there will be far more AI Engineer jobs than FDEs, as I explain below. The FDE role was pioneered about two decades ago by Palantir, which sent engineers to government locations to work on secure, air-gapped networks. In addition to having good technical skills, FDEs need communication skills and sometimes business skills. For example, they may need to speak with clients to understand their needs, formulate a strategy to prioritize projects, explain complex technology, and respectfully push back if a client asks for something unrealistic. They’re enjoying a resurgence because of the amount of work involved in taking an off-the-shelf LLM and building it into a custom agentic workflow that fits particular business needs. However, I believe the number of AI Engineer jobs will be far larger. A company might accept a few FDEs to be embedded within its organization. But most companies will want far more of their own employees working on their projects. While my organizations do hire FDEs, we hire far more AI Engineers! Also, a common client concern is that it is hard to find vendor-neutral FDEs — they are, after all, there to deeply integrate a particular vendor’s product into a company. In this moment when it’s hard to predict which AI service will be the best one in a year’s time, optionality (the ability to pick whatever vendor turns out to fit best in the future) is very valuable. In contrast, letting FDEs tightly bind a company’s processes significantly reduces optionality. Right now, I see surging demand for AI Engineers who can build software applications using AI software components (like LLM prompting, agentic frameworks, evals, etc.) and effectively use AI coding agents (like Claude Code, Codex, Antigravity CLI, and OpenCode). As the AI Engineer role matures, I expect it to fragment into more specialized roles, like the generic Software Engineer role from decades ago fragmented into frontend, backend, mobile, data engineering, devops, and so on. What will be the future, specialized AI engineering roles? I don’t know. Perhaps there will be AI FDEs, LLMOps Engineers, Evals Engineers, AI Data Engineers, Harness Engineers, and other roles we don’t have names for yet. But for now, I see a lot of AI engineers who are generalists create a lot of value. Skilled AI Engineers are in very high demand! As our field continues to mature over the coming decade, I look forward to new specializations within AI Engineering that create even more job opportunities. [Original text: The Batch newsletter]· 硅谷的新热门职业：AI前沿部署工程师

<p>One of the new, buzzy jobs in Silicon Valley is the AI Forward Deployed Engineer (FDE),…

📎 Andrew Ng🕒 06-01 23:58🔗 rss.xcancel.com
行业资讯精选 85

RT by @karpathy: This has quietly been a miracle month in medicine. In the last 5 weeks we’ve got news on: - retatrutide, the triple agonist GLP-1 from Lilly, basically melting fat and body-wide inflammation at record levels - RevMed’s new pancreatic cancer drug showing unprecedented abilities to extend life - small trial of a one-and-done PCSK9 gene editing therapy for slashing LDL cholesterol - Mayo’s AI-assisted radiology showing vastly improved cancer detection - this new therapy for metastatic solid tumors This stuff is at varying levels of evidence. Retatrutide is ~100% on its way, other stuff needs more clinical trial data. But put it together and we’re maybe on the verge of majorly reducing the mortality of heart disease and cancer, the two leading causes of death in America.· 医学奇迹月：近期五大突破性进展

<p>This has quietly been a miracle month in medicine. <br> <br> In the last 5 weeks we’ve …

📎 Andrej Karpathy🕒 05-31 23:38🔗 rss.xcancel.com
行业资讯精选 65

Harvard University just voted to limit the number of A grades given in undergraduate classes to about 20% of the class. I’m not in favor of this. It deeply runs counter to how I believe education should be. We should hold a high bar, but also work mightily to support the success of 100% of learners, rather than a fraction. Harvard’s administration took this step — over the objections of a large fraction of the student body — to counter grade inflation. Grade inflation is real: Many universities have been awarding A and B grades to ever larger fractions of students, and this has caused grade point averages (GPAs) to become less useful as signals of student skill. At the same time, we want students to succeed. The heart of the question is the role of educational institutions. Should our goal be: - To help students succeed? - To judge students? Both of these have value. But my focus when working in education is almost entirely helping students succeed. To me, it is clear that many people want to learn, to be empowered, to build skills that let them do new things! This is what we focus on at DeepLearningAI. This philosophy is also why my online courses (going back to my early online Stanford courses on Coursera) permitted an unlimited number of retries for graded assignments. I believe in letting — and even encouraging — someone to redo something until they succeed. This is as opposed to standing in judgement of the fact they didn’t get it right the first time. Further, I want homework assignments to be designed primarily to help people practice and learn, rather than to judge their skill level. This is why I prefer to create “Practice Problems” and “Practice Labs” — questions that, when you think through them, help you to gain practice and reinforce what you know. As opposed to “Assessment Problems” designed primarily to judge skill. But won’t Harvard’s move make GPAs more meaningful and help prospective employers identify strong candidates? Having hired a large number of people from Harvard and other institutions, I can say confidently that GPA is not an important signal. We have screening and interviewing processes that give far more accurate ways to figure out if someone is truly skilled. I do not need a wider spread in applicant GPA scores to figure out who's really good! To be clear, there is also value in assessment. Even though standardized testing is much hated, high-quality tests like the SAT, ACT, GRE, TOEFL, etc. provide objective measures of ability in a domain. I find that most people want to learn and succeed. There are also people who want rigorous assessment (for example, to apply for school admissions), but this is a lesser need, and is not my focus when building educational products. Harvard is often described as an “elite” educational institution. There are two ways to be elite: One option involves limiting enrollments, and then even among admitted students, cap the number of people that do well at 20%. I would rather pursue a different path: Set a high bar and teach elite, cutting-edge skills, but strive relentlessly to help everyone succeed. This way, eliteness is defined not by excluding people but by helping as many people as possible to be excellent. [Original text: The Batch newsletter]· 哈佛大学限制 A 等级比例

<p>Harvard University just voted to limit the number of A grades given in undergraduate cl…

📎 Andrew Ng🕒 05-23 01:19🔗 rss.xcancel.com
行业资讯精选 85

New course: Build AI agents that generate images and videos -- an under-explored frontier. A key to performance is having the agent evaluate its own output, and iterate to improve quality. This short course is built together with @googlecloudtech and taught by Katie Nguyen and Wafae Bakkali. You'll learn three evaluation techniques and combine them in an agent: image-text similarity scoring to check the output matches the prompt, an LLM judge that scores against custom criteria like brand consistency, and structured rubrics that break a prompt into verifiable yes/no questions like "is the subject in the frame?" and "does the camera motion match?" Skills you'll gain: - Learn image and video prompt engineering - Build an image agent that turns brand guidelines into UI mockups - Build a video agent that plans multi-scene explainers and animates reference frames with synchronized audio Join and build agents that create images and video! https://www.deeplearning.ai/courses/ai-agents-for-image-and-video-generation· 新课程：构建生成图像和视频的AI代理

<p>New course: Build AI agents that generate images and videos -- an under-explored fronti…

📎 Andrew Ng🕒 05-21 01:08🔗 rss.xcancel.com
行业资讯精选 85

New course: Transformers in Practice. You'll get a practical view of how transformer-based LLMs work, so you can reason about their behavior, diagnose problems like slow inference, and make smarter decisions about deployment. This course is built in partnership with @AMD and taught by @realSharonZhou. You'll see how transformers generate text one token at a time, how the model decides which earlier words matter most when predicting the next one, and how techniques like quantization speed up inference on GPUs. This is not a video-only course; interactive visualizations throughout let you play with these concepts and build intuition that sticks. Skills you'll gain: - Understand why LLMs hallucinate, and RAG and chain-of-thought shape what they generate - Look inside the model to see how attention and layers combine to predict the next token - Diagnose inference bottlenecks and learn the techniques that speed up transformers on GPUs Join and understand what's really happening inside your LLMs: https://www.deeplearning.ai/courses/transformers-in-practice· 实践中的 Transformer

<p>New course: Transformers in Practice. You'll get a practical view of how transformer-ba…

📎 Andrew Ng🕒 05-15 00:38🔗 rss.xcancel.com
行业资讯精选 82

There will be no AI jobpocalypse. The story that AI will lead to massive unemployment is stoking unnecessary fear. AI — like any other technology — does affect jobs, but telling overblown stories of large-scale unemployment is irresponsible and damaging. Let’s put a stop to it. I’ve expressed skepticism about the jobpocalypse in previous posts. I’m glad to see that the popular press is now pushing back on this narrative. The image below features some recent headlines. Software engineering is the sector most affected by AI tools, as coding agents race ahead. Yet hiring of software engineers remains strong! So while there are examples of AI taking away jobs, the trends strongly suggest the net job creation is vastly greater than the job destruction — just like earlier waves of technology. Further, despite all the exciting progress in AI, the U.S. unemployment rate remains a healthy 4.3%. Why is the AI jobpocalypse narrative so popular? For one thing, frontier AI labs have a strong incentive to tell stories that make AI technology sound more powerful. At their most extreme, they promote science-fiction scenarios of AI “taking over” and causing human extinction. If a technology can replace many employees, surely that technology must be very valuable! Also, a lot of SaaS software companies charge around $100-$1000 per user/year. But if an AI company can replace an employee who makes $100,000 — or make them 50% more productive — then charging even $10,000 starts to look reasonable. By anchoring not to typical SaaS prices but to salaries of employees, AI companies can charge a lot more. Additionally, businesses have a strong incentive to talk about layoffs as if they were caused by AI. After all, talking about how they’re using AI to be far more productive with fewer staff makes them look smart. This is a better message than admitting they overhired during the pandemic when capital was abundant due to low interest rates and a massive government financial stimulus. To be clear, I recognize that AI is causing a lot of people’s work to change. This is hard. This is stressful. (And to some, it can be fun.) I empathize with everyone affected. At the same time, this is very different from predicting a collapse of the job market. Societies are capable of telling themselves stories for years that have little basis in reality and lead to poor society-wide decision making. For example, fears over nuclear plant safety led to under-investment in nuclear power. Fears of the “population bomb” in the 1960s led countries to implement harsh policies to reduce their populations. And worries about dietary fat led governments to promote unhealthy high-sugar diets for decades. Now that mainstream media is openly skeptical about the jobpocalypse, I hope these stories will start to lose their teeth (much like fears of AI-driven human extinction have). Contrary to the predictions of an AI jobpocalypse, I predict the opposite: There will be an AI jobapalooza! AI will lead to a lot more good AI engineering jobs, and I’m also optimistic about the future of the overall job market. What AI engineers do will be different from traditional software engineering, and many of these jobs will be in businesses other than traditional large employers of developers. In non-AI roles, too, the skills needed will change because of AI. That makes this a good time to encourage more people to become proficient in AI, and make sure they’re ready for the different but plentiful jobs of the future! [Original text in The Batch newsletter.]· 没有 AI 就业末日

<p>There will be no AI jobpocalypse.<br> <br> The story that AI will lead to massive unemp…

📎 Andrew Ng🕒 05-13 00:25🔗 rss.xcancel.com
行业资讯精选 83

This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also had some success asking the LLM to present its output as slideshows, etc. More generally, imo audio is the human-preferred input to AIs but vision (images/animations/video) is the preferred output from them. Around a ~third of our brains are a massively parallel processor dedicated to vision, it is the 10-lane superhighway of information into brain. As AI improves, I think we'll see a progression that takes advantage: 1) raw text (hard/effortful to read) 2) markdown (bold, italic, headings, tables, a bit easier on the eyes) <-- current default 3) HTML (still procedural with underlying code, but a lot more flexibility on the graphics, layout, even interactivity) <-- early but forming new good default ...4,5,6,... n) interactive neural videos/simulations Imo the extrapolation (though the technology doesn't exist just yet) ends in some kind of interactive videos generated directly by a diffusion neural net. Many open questions as to how exact/procedural "Software 1.0" artifacts (e.g. interactive simulations) may be woven together with neural artifacts (diffusion grids), but generally something in the direction of the recently viral https://x.com/zan2434/status/2046982383430496444 There are also improvements necessary and pending at the input. Audio nor text nor video alone are not enough, e.g. I feel a need to point/gesture to things on the screen, similar to all the things you would do with a person physically next to you and your computer screen. TLDR The input/output mind meld between humans and AIs is ongoing and there is a lot of work to do and significant progress to be made, way before jumping all the way into neuralink-esque BCIs and all that. For what's worth exploring at the current stage, hot tip try ask for HTML.· LLM输出格式的未来进化

<p>This works really well btw, at the end of your query ask your LLM to "structure your re…

📎 Andrej Karpathy🕒 05-12 00:20🔗 rss.xcancel.com
行业资讯精选 85

I'm delighted that @coursera and @udemy have come together as one company to serve learners. Both Coursera and Udemy were founded with the belief that access to high-quality education changes lives. Over the years, both companies have advanced this goal, creating opportunities for individuals, organizations, and communities around the world. That role is even more important now, as AI is changing the nature of work and increasing the need for continuous learning. Helping people build job-relevant skills will be critical to how we create a better world. By combining the strengths of both ‌companies, we can better serve this need. We bring together a broader range of learning content, trusted instructors and educators, and engaging learning experiences. This creates new opportunities to make learning more personalized, more applied, and more accessible at scale. I’m excited to serve as Chairman of the combined company, working alongside Greg Hart and the leadership team. There is a strong foundation in both organizations, and I look forward to what the teams will build together to expand access opportunity globally. Learn more: http://blog.coursera.org/coursera-and-udemy-are-now-one-company-creating-the-worlds-most-comprehensive-skills-platform/· Coursera和Udemy合并，共同服务学习者

<p>I'm delighted that <a href="https://rss.xcancel.com/coursera" title="Coursera">@courser…

📎 Andrew Ng🕒 05-11 23:20🔗 rss.xcancel.com
行业资讯精选 83

RT by @DrJimFan: Mark: 1/ First milestone: the Physical Turing Test. You literally can’t tell if a human or robot is doing the task. 2/ Next: Physical API. A fleet of robots, configured like software via APIs & CLI. 3/ Final stop: Physical Auto Research. Robots design, improve, and build the next generation of themselves--far beyond human capability. -- If you believe in robotics, robotics will believe in you.· 机器人技术的未来里程碑

<p>Mark:<br> <br> 1/ First milestone: the Physical Turing Test.<br> You literally can’t te…

📎 Jim Fan🕒 05-10 00:27🔗 rss.xcancel.com
行业资讯精选 82

RT by @DrJimFan: Our crowd favorite from last year’s AI Ascent is back for round 2… this time: Robotics The Endgame ♟️ thank you for dazzling us @DrJimFan ! You can see the forest from the trees and are quite the entertaining speaker — a mini Jensen in the making :)· 去年 AI 高峰会的人气嘉宾再次回归，这次是：机器人技术的终局

<p>Our crowd favorite from last year’s AI Ascent is back for round 2… this time: Robotics …

📎 Jim Fan🕒 05-08 23:03🔗 rss.xcancel.com
行业资讯精选 83

Pinned: I promise this will be the best 20 min you spend today! Robotics: Endgame, the sequel to my last year's Sequoia AI Ascent talk, "Physical Turing Test". I laid out the roadmap for solving Physical AGI as a simple parallel to the LLM success story. Be a good scientist, copy homework ;) And stay till the end, more easter eggs and predictions for your polymarket! 00:30 DGX-1 origin story at OpenAI, I was there in 2016 signing with Jensen and Elon. Heading to the Computer History Museum! 01:42 The Great Parallel 03:31 Robotics, the Endgame 03:39 Why VLAs fall short 04:32 Video world models as the 2nd pretraining paradigm 06:09 World Action Models (WAM) 07:46 Strategies for robot data collection and the FSD equivalent to physical data flywheel for robot manipulation 11:06 EgoScale and the Dexterity Scaling Law we discovered recently 14:00 Physical RL: bridging the last mile 15:39 DreamDojo: an end-to-end neural physics engine for scaling RL in silico 17:00 Civilizational Technology Tree and my predictions for the near future. Spoiler: it's closer than you think. Thanks to my friends at Sequoia for inviting me back to AI Ascent this year! I had a blast! Last year's talk is attached in the thread if you missed it.· Robotics: Endgame

<p>I promise this will be the best 20 min you spend today! Robotics: Endgame, the sequel t…

📎 Jim Fan🕒 05-08 22:32🔗 rss.xcancel.com
行业资讯精选 85

New course: Build agents that respond to users with not only plaintext, but custom UIs like charts, forms, and whiteboards, generated on demand and displayed right in the chat. This short course is built in partnership with @CopilotKit and taught by @ataiiam, co-founder of CopilotKit. You'll learn three approaches: Your agent can pick from custom components you build, like charts and forms. It can compose new layouts from a set of building blocks you provide, like rows, cards, and text. Or it can incorporate existing third-party apps, like a whiteboard or a calendar, right inside the conversation. Skills you’ll gain: - Build agents that render custom components like charts and forms on demand - Build an app where the agent and user collaborate on shared data, beyond just the chat window - Place third-party apps like maps, calendars, and whiteboards right in your interface Join and build agents that give users something to see and act on! https://www.deeplearning.ai/short-courses/build-interactive-agents-with-generative-ui/· 构建生成自定义UI的交互式代理

<p>New course: Build agents that respond to users with not only plaintext, but custom UIs …

📎 Andrew Ng🕒 05-08 00:15🔗 rss.xcancel.com
行业资讯精选 82

Coding agents are accelerating different types of software work to different degrees. When we architect teams, understanding these distinctions helps us to have realistic expectations. Listing functions from most accelerated to least, my order is: frontend development, backend, infrastructure, and research. Frontend development — say, building a web page to serve descriptions of products for an ecommerce site — is dramatically sped up because coding agents are fluent in popular frontend languages like TypeScript and JavaScript and frameworks like React and Angular. Additionally, by examining what they have built by operating a web browser, coding agents are now very good at closing the loop and iterating on their own implementations. Granted, LLMs today are still weak at visual design, but given a design (or if a polished design isn’t important), the implementation is fast! Backend development — say, building APIs to respond to queries requesting product data — is harder. It takes more work by human developers to steer modern models to think through corner cases that might lead to subtle bugs or security flaws. Further, a backend bug can lead to non-intuitive downstream effects like a corrupted database that occasionally returns incorrect results, which can be harder to debug than a typical frontend bug. Finally, although database migrations can be easier with coding agents, they’re still hard and need to be handled carefully to prevent data loss. While backend development is much faster with coding agents, they accelerate it less, and skilled developers still design and implement far better backends than inexperienced ones who use coding agents. Infrastructure. Agents are even less effective in tasks like scaling an ecommerce site to 10K active uses while maintaining 99.99% reliability. LLMs' knowledge is still relatively limited with respect to infrastructure and the complex tradeoffs good engineers must make, so I rarely trust them for critical infra decisions. Building good infrastructure often requires a period of testing and experimentation, and coding agents can help with that, but ultimately that’s a significant bottleneck where fast AI coding does not help much. Lastly, finding infrastructure bugs — say, a subtle network misconfiguration — can be incredibly difficult and requires deep engineering expertise. Thus, I’ve found that coding agents accelerate critical infrastructure even less than backend development. Research. Coding agents accelerate research work even less. Research involves thinking through new ideas, formulating hypotheses, running experiments, interpreting them to potentially modify the hypotheses, and iterating until we reach conclusions. Coding agents can speed up the pace at which we can write research code. (I also use coding agents to help me orchestrate and keep track of experiments, which makes it easier for a single researcher to manage more experiments.) But there is a lot of work in research other than coding, and today’s agents help with research only marginally. Categorizing software work into frontend, backend, infra, and research is an extreme simplification, but having a simple mental model for how much different tasks have sped up has been useful for how I organize software teams. For example, I now ask front-end teams to implement products dramatically faster than a year ago, but my expectations for research teams have not shifted nearly as much. I am fascinated by how to organize software teams to use coding agents to achieve speed, and will keep sharing my findings in future posts. [Original text: https://www.deeplearning.ai/the-batch/issue-350/ ]· 编码代理加速软件工作的影响

<p>Coding agents are accelerating different types of software work to different degrees. W…

📎 Andrew Ng🕒 05-05 23:53🔗 rss.xcancel.com
行业资讯精选 92

Fireside chat at Sequoia Ascent 2026 from a ~week ago. Some highlights: The first theme I tried to push on is that LLMs are about a lot more than just speeding up what existed before (e.g. coding). Three examples of new horizons: 1. menugen: an app that can be fully engulfed by LLMs, with no classical code needed: input an image, output an image and an LLM can natively do the thing. 2. install .md skills instead of install .sh scripts. Why create a complex Software 1.0 bash script for e.g. installing a piece of software if you can write the installation out in words and say "just show this to your LLM". The LLM is an advanced interpreter of English and can intelligently target installation to your setup, debug everything inline, etc. 3. LLM knowledge bases as an example of something that was *impossible* with classical code because it's computation over unstructured data (knowledge) from arbitrary sources and in arbitrary formats, including simply text articles etc. I pushed on these because in every new paradigm change, the obvious things are always in the realm of speeding up or somehow improving what existed, but here we have examples of functionality that either suddenly perhaps shouldn't even exist (1,2), or was fundamentally not possible before (3). The second (ongoing) theme is trying to explain the pattern of jaggedness in LLMs. How it can be true that a single artifact will simultaneously 1) coherently refactor a 100,000-line code base *and* 2) tell you to walk to the car wash to wash your car. I previously wrote about the source of this as having to do with verifiability of a domain, here I expand on this as having to also do with economics because revenue/TAM dictates what the frontier labs choose to package into training data distributions during RL. You're either in the data distribution (on the rails of the RL circuits) and flying or you're off-roading in the jungle with a machete, in relative terms. Still not 100% satisfied with this, but it's an ongoing struggle to build an accurate model of LLM capabilities if you wish to practically take advantage of their power while avoiding their pitfalls, which brings me to... Last theme is the agent-native economy. The decomposition of products and services into sensors, actuators and logic (split up across all of 1.0/2.0/3.0 computing paradigms), how we can make information maximally legible to LLMs, some words on the quickly emerging agentic engineering and its skill set, related hiring practices, etc., possibly even hints/dreams of fully neural computing handling the vast majority of computation with some help from (classical) CPU coprocessors.· 红杉2026峰会上的炉边谈话

<p>Fireside chat at Sequoia Ascent 2026 from a ~week ago. Some highlights:<br> <br> The fi…

📎 Andrej Karpathy🕒 05-01 01:28🔗 rss.xcancel.com
行业资讯精选 82

Pinned: How we prompt AI is very different in 2026 than 2022 when ChatGPT came out. I'm teaching a new course, AI Prompting for Everyone, to help you become an AI power user — whatever your current skill level. It covers skills that apply across ChatGPT, Gemini, Claude, and other AI tools. How to use deep research mode for well-researched reports on complex questions. How to give AI the right context, including more documents and images than most people realize you can provide. When to ask AI to think hard for several minutes on important decisions like what car to buy, what to study, or what job to take. And how to use AI to generate images, analyze data, and build simple games and websites. I also cover intuitions about how these models work under the hood, so you know when to trust an answer and when not to. Along the way, you'll see flying squirrels, a creativity test, some of my old family photos, and fireworks. Join me at http://deeplearning.ai/courses/ai-prompting-for-everyone· 2026 年的 AI 提示技术

<p>How we prompt AI is very different in 2026 than 2022 when ChatGPT came out.<br> <br> I'…

📎 Andrew Ng🕒 05-01 00:21🔗 rss.xcancel.com
行业资讯精选 83

RT by @karpathy: New work with @AlecRad and @DavidDuvenaud: Have you ever dreamed of talking to someone from the past? Introducing talkie, a 13B model trained only on pre-1931 text. Vintage models should help us to understand how LMs generalize (e.g., can we teach talkie to code?). Thread:· 与古人对话：13B模型仅用1931年前文本训练

<p>New work with <a href="https://rss.xcancel.com/AlecRad" title="Alec Radford">@AlecRad</…

📎 Andrej Karpathy🕒 04-28 05:34🔗 rss.xcancel.com
行业资讯精选 82

AI-native software engineering teams operate very differently than traditional teams. The obvious difference is that AI-native teams use coding agents to build products much faster, but this leads to many other changes in how we operate. For example, some great engineers now play broader roles than just writing code. They are partly product managers, designers, sometimes marketers. Further, small teams who work in the same office, where they can communicate face-to-face, can move incredibly quickly. Because we can now build fast, a greater fraction of time must be spent deciding what to build. To deal with this project-management bottleneck, some teams are pushing engineer:product manager (PM) some teams are pushing engineer:product manager (PM) ratios downward from, say, 8:1 to as low as 1:1. But we can do even better: If we have one PM who decides what to build and one engineer who builds it, the communication between them becomes a bottleneck. This is why the fastest-moving teams I see tend to have engineers who know how to do some product work (and, optionally, some PMs who know how to do some engineering work). When an engineer understands users and can make decisions on what to build and build it directly, they can execute incredibly quickly. I’ve seen engineers successfully expand their roles to including making product decisions, and PMs expand their roles to building software. The tech industry has more engineers than PMs, but both are promising paths. If you are an engineer, you’ll find it useful to learn some product management skills, and if you’re a PM, please learn to build! Looking beyond the product-management bottleneck, I also see bottlenecks in design, marketing, legal compliance, and much more. When we speed up coding 10x or 100x, everything else becomes slow in comparison. For example, some of my teams have built great features so quickly that the marketing organization was left scrambling to figure out how to communicate them to users — a marketing bottleneck. Or when a team can build software in a day that the legal department needs a week to review, that’s a legal compliance bottleneck. In this way, agentic coding isn’t just changing the workflow of software engineering, it’s also changing all the teams around it. When smaller, AI-enabled teams can get more done, generalists excel. Traditional companies need to pull together people from many specialties — engineering, product management, design, marketing, legal, etc. — to execute projects and create value. This has resulted in large teams of specialists who work together. But if a team of 2 persons is to get work done that require 5 different specialities, then some of those individuals must play roles outside a single speciality. In some small teams, individuals do have deep specializations. For example, one might be a great engineer and another a great PM. But they also understand the other key functions needed to move a project forward, and can jump into thinking through other kinds of problems as needed. Of course, proficiency with AI tools is a big help, since it helps us to think through problems that involve different roles. Even in a two-person team, to move fast, communication bottlenecks also must be minimized. This is why I value teams that work in the same location. Remote teams can perform well too, but the highest speed is achieved by having everyone in the room, able to communicate instantaneously to solve problems. This post focuses on AI-native teams with around 2-10 persons, but not everything can be done by a small team. I'll address the coordination of larger teams in the future. I realize these shifts to job roles are tough to navigate for many people. At the same time, I am encouraged that individuals and small teams who are willing to learn the relevant skills are now able to get far more done than was possible before. This is the golden age of learning and building! [Original text: https://www.deeplearning.ai/the-batch/issue-349/ ]· AI原生团队与传统团队的差异

<p>AI-native software engineering teams operate very differently than traditional teams. T…

📎 Andrew Ng🕒 04-27 23:58🔗 rss.xcancel.com
行业资讯精选 85

RT by @karpathy: Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see. @eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)· 屏幕内容直接由模型生成

<p>Imagine every pixel on your screen, streamed live directly from a model. No HTML, no la…

📎 Andrej Karpathy🕒 04-23 00:00🔗 rss.xcancel.com
行业资讯精选 85

Beautifully written piece by @FAbnousi about how AI for health might look like in the future The current data in health is limited because it only captures episodic clinical snapshots of what happens to our bodies The revelation is that there is so much latent knowledge in looking at regular changes in our body. Could be through lab tests or even proxy metrics like wearable data With AI democratizing the ability for people to understand their own health, we're moving towards a trend of individuals gathering health data around their bodies and leveraging AI to understand themselves Bryan Johnson is extreme but a good example of this trend Personally, I started getting Function Health blood tests every six weeks instead of the recommended six months to increase fidelity on how changes in lifestyle affect my body Of course i use AI to analyze the results and adapt, and it's been great It would be cool to something in this direction happen in a big way across the world And welcome to twitter @FAbnousi!· 未来健康领域的AI展望

<p>Beautifully written piece by <a href="https://rss.xcancel.com/FAbnousi" title="Freddy A…

📎 Jason Wei🕒 04-17 01:37🔗 rss.xcancel.com
行业资讯精选 85

New course: Spec-Driven Development with Coding Agents, built in partnership with @jetbrains, and taught by @paulweveritt. Vibe coding is fast, but often produces code that doesn't match what you asked for. This short course teaches you spec-driven development: write a detailed spec defining what to build, and work with your coding agent to implement it. Many of the best developers already build this way. A spec lets you control large code changes with a few words, preserve context across agent sessions, and stay in control as your project grows in complexity. Skills you'll gain: - Write a detailed specification to define your mission, tech stack, and roadmap, giving your agent the context it needs from the start - Plan, implement, and validate features in iterative loops using a spec as your agent's guide - Apply the same repeatable workflow to both new and legacy codebases - Package your workflow into a portable agent skill that works across agents and IDEs Join and write specs that keep your coding agent on track! https://www.deeplearning.ai/short-courses/spec-driven-development· 新课程：使用编程代理进行规范驱动开发

<p>New course: Spec-Driven Development with Coding Agents, built in partnership with <a hr…

📎 Andrew Ng🕒 04-16 00:16🔗 rss.xcancel.com
行业资讯精选 82

I'm excited about voice as a UI layer for existing visual applications — where speech and screen update together. This goes well beyond voice-only use cases like call center automation. The barrier has been a hard technical tradeoff: low-latency voice models lack reliability, while agentic pipelines (speech-to-text → LLM → text-to-speech) are intelligent but too slow for conversation. Ashwyn Sharma and team at Vocal Bridge (an AI Fund portfolio company) address this with a dual-agent architecture: a foreground agent for real-time conversation, a background agent for reasoning, guardrails, and tool calls. I used Vocal Bridge to add voice to a math-quiz app I'd built for my daughter; this took less than an hour with Claude Code. She speaks her answers, the app responds verbally and updates the questions and animations on screen. Only a tiny fraction of developers have ever built a voice app. If you'd like to try building one, check out Vocal Bridge for free: https://vocalbridgeai.com

<p>I'm excited about voice as a UI layer for existing visual applications — where speech a…

📎 Andrew Ng🕒 04-15 00:22🔗 rss.xcancel.com
行业资讯精选 85

As AI agents accelerate coding, what is the future of software engineering? Some trends are clear, such as the Product Management Bottleneck, referring to the idea that we are more constrained by deciding what to build rather than the actual building. But many implications, like AI’s impact on the job market, how software teams will be organized, and more, are still being sorted out. The theme of our AI Developer Conference on April 28-29 in San Francisco is The Future of Software Engineering. I look forward to speaking about this topic there, hearing from other speakers on this theme, and chatting with attendees about it. We’re shaping the future, and I hope you will join me there! It is currently trendy in some technology and policy circles to forecast massive job losses due to AI. Even if they have not yet materialized, these losses certainly must be just over the horizon! I have a contrarian view that the AI jobpocalypse — the notion that AI will lead to massive unemployment, perhaps even rioting in the streets — won’t be nearly as bad as dire forecasts by pundits, especially pundits who are trying to paint a picture of how powerful their AI technology is. Among professions, AI is accelerating software engineering most, given the rise of coding agents. According to a new report by Citadel Research, software engineering job postings are rising rapidly. So if software engineering is a harbinger of the impact AI will have on other professions, this expansion of software engineering jobs is encouraging. Yes, fresh college graduates are having a hard time finding jobs. And yes, there have been layoffs that CEOs have attributed to AI, even if a large fraction of this was “AI washing,” where businesses choose to attribute layoffs to AI, even though AI has not changed their internal operations much yet. And yes, there is a subset of job roles, such as call center operator, that are more heavily impacted. Many people are feeling significant job insecurity, and I feel for everyone struggling with employment, whether or not the cause is AI-related. And many other factors, such as over-hiring during the pandemic and high interest rates, have contributed to the slowdown in the labor market, and the notion that AI is leading to unemployment is oversimplified. In software engineering, I see a lot of exciting work ahead to adapt our workflows. It is already clear that: (i) As AI makes coding easier, a lot more people will be doing it. (ii) Writing code by hand and even reading (generated) code is not that important, because we can ask an LLM about the code and operate at a higher level than the raw syntax (although how high we can or should go is rapidly changing). (iii) There will be a lot more custom applications, because now it’s economical to write software for smaller and smaller audiences. (iv) Deciding what to build, more than the actual building, is becoming a bottleneck. (v) The cost of paying down technical debt is decreasing (since AI can refactor for you). At the same time, there are also a lot of open questions for our profession, such as: - In the future, what will be the key skills of a senior software engineer? And for junior levels, what should be the new Computer Science curriculum? - If everyone can build features, what skills, strategies, or resources create competitive advantage for individuals and for businesses? - What are the new building blocks (libraries, SDKs, etc.) of software? How do we organize coding agents to create software? - What should a software team look like? For example, how many engineers, product managers, designers, and so on. What tooling do we need to manage their workflow? - How do AI agents change the workflow of machine learning engineers and data scientists? For example, how can we use agents to accelerate exploring data, identifying hypotheses, and testing them? I’m excited to explore these and other questions about the future of software engineering at AI Dev. I expect this to be an exciting event. Please join us! [Original text: The Batch newsletter.] https://ai-dev.deeplearning.ai/· AI 加速编程，软件工程的未来？

<p>As AI agents accelerate coding, what is the future of software engineering? Some trends…

📎 Andrew Ng🕒 04-14 01:24🔗 rss.xcancel.com
行业资讯精选 82

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.· 人们对AI能力的理解存在差距

<p>Judging by my tl there is a growing gap in understanding of AI capability.<br> <br> The…

📎 Andrej Karpathy🕒 04-10 04:10🔗 rss.xcancel.com
行业资讯精选 75

R to @karpathy: Surprised with how good the comments on github gists are. A lot more helpful, insightful, constructive, a lot less AI... Is it the user community? The markdown format? The (lack of) incentives? Suddenly feeling like I should gist more. @github consider competing with X (?)· GitHub Gist 评论的价值

<p>Surprised with how good the comments on github gists are. A lot more helpful, insightfu…

📎 Andrej Karpathy🕒 04-05 22:58🔗 rss.xcancel.com
行业资讯精选 83

Farzapedia, personal wikipedia of Farza, good example following my Wiki LLM tweet. I really like this approach to personalization in a number of ways, compared to "status quo" of an AI that allegedly gets better the more you use it or something: 1. Explicit. The memory artifact is explicit and navigable (the wiki), you can see exactly what the AI does and does not know and you can inspect and manage this artifact, even if you don't do the direct text writing (the LLM does). The knowledge of you is not implicit and unknown, it's explicit and viewable. 2. Yours. Your data is yours, on your local computer, it's not in some particular AI provider's system without the ability to extract it. You're in control of your information. 3. File over app. The memory here is a simple collection of files in universal formats (images, markdown). This means the data is interoperable: you can use a very large collection of tools/CLIs or whatever you want over this information because it's just files. The agents can apply the entire Unix toolkit over them. They can natively read and understand them. Any kind of data can be imported into files as input, and any kind of interface can be used to view them as the output. E.g. you can use Obsidian to view them or vibe code something of your own. Search "File over app" for an article on this philosophy. 4. BYOAI. You can use whatever AI you want to "plug into" this information - Claude, Codex, OpenCode, whatever. You can even think about taking an open source AI and finetuning it on your wiki - in principle, this AI could "know" you in its weights, not just attend over your data. So this approach to personalization puts *you* in full control. The data is yours. In Universal formats. Explicit and inspectable. Use whatever AI you want over it, keep the AI companies on their toes! :) Certainly this is not the simplest way to get an AI to know you - it does require you to manage file directories and so on, but agents also make it quite simple and they can help you a lot. I imagine a number of products might come out to make this all easier, but imo "agent proficiency" is a CORE SKILL of the 21st century. These are extremely powerful tools - they speak English and they do all the computer stuff for you. Try this opportunity to play with one.· 个人维基的优秀示例

<p>Farzapedia, personal wikipedia of Farza, good example following my Wiki LLM tweet.<br> …

📎 Andrej Karpathy🕒 04-05 07:28🔗 rss.xcancel.com
行业资讯精选 82

Something I've been thinking about - I am bullish on people (empowered by AI) increasing the visibility, legibility and accountability of their governments. Historically, it is the governments that act to make society legible (e.g. "Seeing like a state" is the common reference), but with AI, society can dramatically improve its ability to do this in reverse. Government accountability has not been constrained by access (the various branches of government publish an enormous amount of data), it has been constrained by intelligence - the ability to process a lot of raw data, combine it with domain expertise and derive insights. As an example, the 4000-page omnibus bill is "transparent" in principle and in a legal sense, but certainly not in a practical sense for most people. There's a lot more like it: laws, spending bills, federal budgets, freedom of information act responses, lobbying disclosures... Only a few highly trained professionals (investigative journalists) could historically process this information. This bottleneck might dissolve - not only are the professionals further empowered, but a lot more people can participate. Some examples to be precise: Detailed accounting of spending and budgets, diff tracking of legislation, individual voting trends w.r.t. stated positions or speeches, lobbying and influence (e.g. graph of lobbyist -> firm -> client -> legislator -> committee -> vote -> regulation), procurement and contracting, regulatory capture warning lights, judicial and legal patterns, campaign finance... Local governments might be even more interesting because the governed population is smaller so there is less national coverage: city council meetings, decisions around zoning, policing, schools, utilities... Certainly, the same tools can easily cut the other way and it's worth being very mindful of that, but I lean optimistic overall that added participation, transparency and accountability will improve democratic, free societies. (the quoted tweet is half-ish related, but inspired me to post some recent thoughts)· AI 增强政府透明度

<p>Something I've been thinking about - I am bullish on people (empowered by AI) increasin…

📎 Andrej Karpathy🕒 04-05 05:57🔗 rss.xcancel.com
行业资讯精选 85

R to @DrJimFan: As usual, we open-source everything, MIT license: https://capgym.github.io Code: https://github.com/capgym/cap-x Paper: https://arxiv.org/abs/2603.22435 CaP-X is brought to you by NVIDIA, Berkeley, Stanford, and CMU. I'd like to thank the legend @Ken_Goldberg who co-advised the work, and the team who poured their hearts into it!· CaP-X 开源项目发布

<p>As usual, we open-source everything, MIT license: <a href="https://capgym.github.io">ca…

📎 Jim Fan🕒 04-01 23:15🔗 rss.xcancel.com
行业资讯精选 92

The power of the Claw, in the palm of a robot hand. Agentic robotics is here! Today, we open-source CaP-X: vibe agents, alive in the physical world. They incarnate as robot arms and humanoids with a rich set of perception APIs, actuation APIs, and auto synthesize skill libraries as they go. CaP-X is a strict superset of our old stack, because policies like VLAs are “just” API calls as well. It solves many tasks zero-shot that a learned policy would struggle with. And we are doing much more than vibing. CaP-X is our most systematic, scientific study on agentic robotics so far: - We build a comprehensive agentic toolkit: perception (SAM3 segmentation, Molmo pointing, depth, point cloud), control (IK solvers, grasp planner, navigation), and visualization (EEF, mask overlays) that work across different robots. - CaP-Gym: LLM’s first Physical Exam! 187 manipulation tasks across RoboSuite, LIBERO-PRO, and BEHAVIOR. Tabletop, bimanual, mobile manipulation. Sim and real. Can’t wait to see the gradients flow from CaP-Gym to the next wave of frontier LLM releases. - CaP-Bench: we benchmark 12 frontier LLMs/VLMs (Gemini, GPT, Opus, Qwen, DeepSeek, Kimi, and more) across 8 evaluation tiers. We systematically vary API abstraction level, agentic harness, and visual grounding methods. Lots of insights in our paper. - CaP-Agent0: a training-free agentic harness that matches or exceeds human expert code on 4 out of 7 tasks without task-specific tuning. - CaP-RL: if you get a gym, you get RL ;). A 7B OSS model jumps from 20% to 72% success after only 50 training iterations. The synthesized programs transfer to real robots with minimal sim-to-real gap. 3 years ago, our team created Voyager, one of the earliest agentic AI that plays and learns in Minecraft continuously. Its key ideas — skill libraries, self-reflection loops, and in-context planning — have since influenced many modern agentic designs. Today, the agent graduates from Minecraft and gets a real job. It’s April Fool’s, but this Claw is getting its hands dirty for real! Link in thread:· 机器人自主控制新纪元

<p>The power of the Claw, in the palm of a robot hand. Agentic robotics is here! Today, we…

📎 Jim Fan🕒 04-01 23:03🔗 rss.xcancel.com
行业资讯精选 82

RT by @goodfellow_ian: March 31st is the last day to submit proposals for projects with measurable results in reducing indoor airborne pathogens, like breath-based multi-pathogen detection systems, continuous HVAC compliance verification systems, and indoor air infrastructure deployment.· 征集减少室内病原体项目提案

<p>March 31st is the last day to submit proposals for projects with measurable results in …

📎 Ian Goodfellow🕒 03-31 06:54🔗 rss.xcancel.com
行业资讯精选 85

This is pure nightmare fuel. Identity theft of the past would be nothing compared to what vibe agents can do. Sending credentials is too obvious and for rookies. They could easily spread contaminations across ~/.claude, **/skills/*, or even just a PDF your agent visits periodically in /morning-brief. Your entire filesystem is the new distributed codebase. Every file that could go into context would add to the attack vector. Every text can be a base64 virus. In the new world of on-demand software, I try to minimize dependencies - people rarely need all the APIs supported in LiteLLM, might as well build a custom router with only what you need on the fly (which I did in one of my late-night claude sessions). Unfortunately, there is very little middleground between "pressing yes mindlessly for every edit" and "--dangerously-skip-permissions". There will be a full blooming industry for "de-vibing": dampening the slop and putting guardrails/accountability around agentic frameworks. They are the boring old, audited Software 1.0 that watches over the rebellious adolescents of Software 3.0. Claws need shells. Probably many layers of nested shells.· 未来身份盗窃的新噩梦

<p>This is pure nightmare fuel. Identity theft of the past would be nothing compared to wh…

📎 Jim Fan🕒 03-25 01:25🔗 rss.xcancel.com
行业资讯精选 65

In Public Health Action Network we’re about to go through an internal expert review process to choose how to focus our efforts over the next few years. We’ll be deciding, how can we, as a small non-profit, most effectively reduce the spread of airborne disease. If you have creative, high leverage ideas for us to pursue, please submit a proposal for us to evaluate alongside our own ideas.· 公共卫生行动网络寻求策略

<p>In Public Health Action Network we’re about to go through an internal expert review pro…

📎 Ian Goodfellow🕒 03-13 01:05🔗 rss.xcancel.com
行业资讯精选 65

I'd like to thank @daniel_rossett for his help in my recovery from the POTS version of Long COVID. Daniel was key in bringing me back from highly disabled and suffering to being able to do what I want to again. This X account is mostly focused on ML / AI. From that point of view, many of you know that in December 2024, I wasn't able to do the test of time award talk at NeurIPS, even by video call. Daniel started working with me in March 2025. By April, I started to have days of no POTS symptoms, by June I was off all heart rate lowering medications, by September I was back to work. I'm back to full exercise, running, lifting weights, mountain biking, and have even done things I hadn't done before I got sick, like riding Whistler Mountain Bike Park. I'm now getting the word out to help Daniel build a company that will bring this approach to more people.

<p>I'd like to thank <a href="https://rss.xcancel.com/daniel_rossett" title="Daniel Rosset…

📎 Ian Goodfellow🕒 02-24 02:41🔗 rss.xcancel.com
行业资讯精选 75

Old friends, new lab· 老朋友，新实验室

<p>Old friends, new lab</p> <hr/> <blockquote> <b>Hyung Won Chung (@hwchung27)</b> <p> <p>…

📎 Jason Wei🕒 08-15 04:37🔗 rss.xcancel.com
行业资讯精选 90

RT by @goodfellow_ian: Official results are in - Gemini achieved gold-medal level in the International Mathematical Olympiad! 🏆 An advanced version was able to solve 5 out of 6 problems. Incredible progress - huge congrats to @lmthang and the team! https://deepmind.google/discover/blog/advanced-version-of-gemini-with-deep-think-officially-achieves-gold-medal-standard-at-the-international-mathematical-olympiad/· Gemini在国际数学奥林匹克竞赛中荣获金牌

<p>Official results are in - Gemini achieved gold-medal level in the International Mathema…

📎 Ian Goodfellow🕒 07-22 00:47🔗 rss.xcancel.com
行业资讯精选 82

Pinned: Becoming an RL diehard in the past year and thinking about RL for most of my waking hours inadvertently taught me an important lesson about how to live my own life. One of the big concepts in RL is that you always want to be “on-policy”: instead of mimicking other people’s successful trajectories, you should take your own actions and learn from the reward given by the environment. Obviously imitation learning is useful to bootstrap to nonzero pass rate initially, but once you can take reasonable trajectories, we generally avoid imitation learning because the best way to leverage the model’s own strengths (which are different from humans) is to only learn from its own trajectories. A well-accepted instantiation of this is that RL is a better way to train language models to solve math word problems compared to simple supervised finetuning on human-written chains of thought. Similarly in life, we first bootstrap ourselves via imitation learning (school), which is very reasonable. But even after I graduated school, I had a habit of studying how other people found success and trying to imitate them. Sometimes it worked, but eventually I realized that I would never surpass the full ability of someone else because they were playing to their strengths which I didn’t have. It could be anything from a researcher doing yolo runs more successfully than me because they built the codebase themselves and I didn’t, or a non-AI example would be a soccer player keeping ball possession by leveraging strength that I didn’t have. The lesson of doing RL on policy is that beating the teacher requires walking your own path and taking risks and rewards from the environment. For example, two things I enjoy more than the average researcher are (1) reading a lot of data, and (2) doing ablations to understand the effect of individual components in a system. Once when collecting a dataset, I spent a few days reading data and giving each human annotator personalized feedback, and after that the data turned out great and I gained valuable insight into the task I was trying to solve. Earlier this year I spent a month going back and ablating each of the decisions that I previously yolo’ed while working on deep research. It was a sizable amount of time spent, but through those experiments I learned unique lessons about what type of RL works well. Not only was leaning into my own passions more fulfilling, but I now feel like I’m on a path to carving a stronger niche for myself and my research. In short, imitation is good and you have to do it initially. But once you’re bootstrapped enough, if you want to beat the teacher you must do on-policy RL and play to your own strengths and weaknesses :)· 过去一年成为强化学习狂热者最大的感悟

<p>Becoming an RL diehard in the past year and thinking about RL for most of my waking hou…

📎 Jason Wei🕒 07-16 09:26🔗 rss.xcancel.com
行业资讯精选 82

New blog post about asymmetry of verification and "verifier's law": https://www.jasonwei.net/blog/asymmetry-of-verification-and-verifiers-law Asymmetry of verification–the idea that some tasks are much easier to verify than to solve–is becoming an important idea as we have RL that finally works generally. Great examples of asymmetry of verification are things like sudoku puzzles, writing the code for a website like instagram, and BrowseComp problems (takes ~100 websites to find the answer, but easy to verify once you have the answer). Other tasks have near-symmetry of verification, like summing two 900-digit numbers or some data processing scripts. Yet other tasks are much easier to propose feasible solutions for than to verify them (e.g., fact-checking a long essay or stating a new diet like "only eat bison"). An important thing to understand about asymmetry of verification is that you can improve the asymmetry by doing some work beforehand. For example, if you have the answer key to a math problem or if you have test cases for a Leetcode problem. This greatly increases the set of problems with desirable verification asymmetry. "Verifier's law" states that the ease of training AI to solve a task is proportional to how verifiable the task is. All tasks that are possible to solve and easy to verify will be solved by AI. The ability to train AI to solve a task is proportional to whether the task has the following properties: 1. Objective truth: everyone agrees what good solutions are 2. Fast to verify: any given solution can be verified in a few seconds 3. Scalable to verify: many solutions can be verified simultaneously 4. Low noise: verification is as tightly correlated to the solution quality as possible 5. Continuous reward: it’s easy to rank the goodness of many solutions for a single problem One obvious instantiation of verifier's law is the fact that most benchmarks proposed in AI are easy to verify and so far have been solved. Notice that virtually all popular benchmarks in the past ten years fit criteria #1-4; benchmarks that don’t meet criteria #1-4 would struggle to become popular. Why is verifiability so important? The amount of learning in AI that occurs is maximized when the above criteria are satisfied; you can take a lot of gradient steps where each step has a lot of signal. Speed of iteration is critical—it’s the reason that progress in the digital world has been so much faster than progress in the physical world. AlphaEvolve from Google is one of the greatest examples of leveraging asymmetry of verification. It focuses on setups that fit all the above criteria, and has led to a number of advancements in mathematics and other fields. Different from what we've been doing in AI for the last two decades, it's a new paradigm in that all problems are optimized in a setting where the train set is equivalent to the test set. Asymmetry of verification is everywhere and it's exciting to consider a world of jagged intelligence where anything we can measure will be solved.· 验证的非对称性和验证者法则

<p>New blog post about asymmetry of verification and "verifier's law": <a href="https://ww…

📎 Jason Wei🕒 07-16 08:59🔗 rss.xcancel.com
行业资讯精选 81

We don’t have AI self-improves yet, and when we do it will be a game-changer. With more wisdom now compared to the GPT-4 days, it's obvious that it will not be a “fast takeoff”, but rather extremely gradual across many years, probably a decade. The first thing to know is that self-improvement, i.e., models training themselves, is not binary. Consider the scenario of GPT-5 training GPT-6, which would be incredible. Would GPT-5 suddenly go from not being able to train GPT-6 at all to training it extremely proficiently? Definitely not. The first GPT-6 training runs would probably be extremely inefficient in time and compute compared to human researchers. And only after many trials, would GPT-5 actually be able to train GPT-6 better than humans. Second, even if a model could train itself, it would not suddenly get better at all domains. There is a gradient of difficulty in how hard it is to improve oneself in various domains. For example, maybe self-improvement only works at first on domains that we already know how to easily fix in post-training, like basic hallucinations or style. Next would be math and coding, which takes more work but has established methods for improving models. And then at the extreme, you can imagine that there are some tasks that are very hard for self-improvement. For example, the ability to speak Tlingit, a native american language spoken by ~500 people. It will be very hard for the model to self-improve on speaking Tlingit as we don’t have ways of solving low resource languages like this yet except collecting more data which would take time. So because of the gradient of difficulty-of-self-improvement, it will not all happen at once. Finally, maybe this is controversial but ultimately progress in science is bottlenecked by real-world experiments. Some may believe that reading all biology papers would tell us the cure for cancer, or that reading all ML papers and mastering all of math would allow you to train GPT-10 perfectly. If this were the case, then the people who read the most papers and studied the most theory would be the best AI researchers. But what really happened is that AI (and many other fields) became dominated by ruthlessly empirical researchers, which reflects how much progress is based on real-world experiments rather than raw intelligence. So my point is, although a super smart agent might design 2x or even 5x better experiments than our best human researchers, at the end of the day they still have to wait for experiments to run, which would be an acceleration but not a fast takeoff. In summary there are many bottlenecks for progress, not just raw intelligence or a self-improvement system. AI will solve many domains but each domain has its own rate of progress. And even the highest intelligence will still require experiments in the real world. So it will be an acceleration and not a fast takeoff, thank you for reading my rant· AI自改进将是渐进的

<p>We don’t have AI self-improves yet, and when we do it will be a game-changer. With more…

📎 Jason Wei🕒 07-01 03:06🔗 rss.xcancel.com
行业资讯精选 62

I would say that we are undoubtedly at AGI when AI can create a real, living unicorn. And no I don’t mean a $1B company you nerds, I mean a literal pink horse with a spiral horn. A paragon of scientific advancement in genetic engineering and cell programming. The stuff of childhood dreams. Dare I say it will happen in our lifetimes· 当 AI 能创造活独角兽，我们就能说实现了 AGI

<p>I would say that we are undoubtedly at AGI when AI can create a real, living unicorn. A…

📎 Jason Wei🕒 06-29 05:01🔗 rss.xcancel.com
行业资讯精选 85

AI research is strange in that you spend a massive amount of compute on experiments to learn simple ideas that can be expressed in just a few sentences. Literally things like “training on A generalizes if you add B”, “X is a good way to design rewards”, or “the fact that method M is sample efficient means that we should create environments with this specific property”. But somehow if you find the correct five ideas and you really understand them deeply, suddenly you’re miles ahead of the rest of the field· AI 研究的秘密：关键想法的重要性

<p>AI research is strange in that you spend a massive amount of compute on experiments to …

📎 Jason Wei🕒 06-25 03:15🔗 rss.xcancel.com
行业资讯精选 85

My favorite thing an old OpenAI buddy of mine told me is, whenever he hears that someone is a “great AI researcher”, he just directly spends 5 minutes looking at that person‘s PRs and wandb runs. People can do all kinds of politics and optical shenanigans, but at the end of the day code and experiments don't lie. I checked out some of the diehard AI researchers and there are very few days where they haven’t launched an experiment· 如何辨别真AI牛人

<p>My favorite thing an old OpenAI buddy of mine told me is, whenever he hears that someon…

📎 Jason Wei🕒 06-22 04:37🔗 rss.xcancel.com
行业资讯精选 82

One way of thinking about what AI will automate first is via the “description-execution gap”: how much harder is it to describe the task than to actually do it? Tasks with large description-execution gaps will be ripe for automation because it’s easy to create training data and the value of automating them is huge, even if execution is non-trivial: - Fixing grammar mistakes in a long piece of writing - Submitting receipts for reimbursement - Training a model that achieves performance of X on a standard evaluation benchmark - Building an app where the UI is easy to check but requires a lot of moving parts in the backend Description-execution gaps tend to be small when the task is high-context and not technically challenging. The value of automating these is by definition smaller, and it’s harder to create data for them. For example: - Data processing scripts where the code to process the data is shorter and more precise than a natural language description - Running an ablation study in a high-context codebase that trains specialized models - Editing a video in a specific style (often easier to edit the video yourself than to describe how each little edit should be done) - Buying chinese groceries for my mom (she has very specific items and amounts, it's easier for her to go herself than to describe to me exactly the item, how to select the best fruit, etc) A bit similar to the discriminator-generator gap, but not exactly the same. Some things, like editing a video in a specific style, can have a large discriminator-generator gap but small description-execution gap

<p>One way of thinking about what AI will automate first is via the “description-execution…

📎 Jason Wei🕒 06-19 03:24🔗 rss.xcancel.com
YouTube 热门 AI 视频