正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Met someone whose agents ship more code in a week than my whole team does in a month. I asked him what actually made him this good - which loop framework, which prompt pack. He laughed and sent me a Stanford lecture. Andrew Ng - CS230, Intro to Deep... show more

dunik

3,987 subscribers

35,917 次观看 • 15 天前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Met a guy making $1.4 million a year as a prompt engineer. I asked him how he learned prompting so well. He sent me a video that was never supposed to get out. Andrew Ng's 2 hour prompting course. You wont find anything better about prompting than this video. I watched it last night. Halfway through, I realized I have been using Claude completely wrong for years. Bookmark and watch this today.

Met a guy making $1.4 million a year as a prompt engineer. I asked him how he learned prompting so well. He sent me a video that was never supposed to get out. Andrew Ng's 2 hour prompting course. You wont find anything better about prompting than this video. I watched it last night. Halfway through, I realized I have been using Claude completely wrong for years. Bookmark and watch this today.

Roan

309,560 次观看 • 15 天前

The guy who kicked off the entire "loop engineering" wave Peter Steinberger: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents." One post. 6.5M views in a week. In this talk he walks the real stack: the agent loop, a verifier that fails its own work and retries, and a loop that rewrites the agent while he sleeps. Worth more than any $500 vibe-coding course. Watch it, then read the full breakdown of the 4 loops below.

The guy who kicked off the entire "loop engineering" wave Peter Steinberger: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents." One post. 6.5M views in a week. In this talk he walks the real stack: the agent loop, a verifier that fails its own work and retries, and a loop that rewrites the agent while he sleeps. Worth more than any $500 vibe-coding course. Watch it, then read the full breakdown of the 4 loops below.

dunik

218,959 次观看 • 22 天前

My friend makes $850,000/year as an Anthropic AI engineer. No MIT. No Stanford. No PhD. I asked him how he broke into AI from scratch. He sent me a course that was never supposed to get out. A developer teaching to build self-improving AI agents from scratch. You will learn exactly how AI agents like Claude Code and OpenClaw are actually built. You won't find anything better about becoming an AI engineer in 2026 than this video. I watched it last night. Halfway through, I realized I could land a role at a top AI lab in months, not years. Bookmark this and read the article below.

My friend makes $850,000/year as an Anthropic AI engineer. No MIT. No Stanford. No PhD. I asked him how he broke into AI from scratch. He sent me a course that was never supposed to get out. A developer teaching to build self-improving AI agents from scratch. You will learn exactly how AI agents like Claude Code and OpenClaw are actually built. You won't find anything better about becoming an AI engineer in 2026 than this video. I watched it last night. Halfway through, I realized I could land a role at a top AI lab in months, not years. Bookmark this and read the article below.

Roan

67,179 次观看 • 4 天前

Boris Cherny, the engineer who built Claude Code: "I don't talk to an agent anymore. I talk to a loop" a prompt is one instruction you babysit. a loop is a goal the AI works toward on its own: > it plans > does the work > checks itself > fixes what's weak > and repeats until it's done you step out, the work keeps going that's the shift the best engineers quietly moved to. most people hear a line like that and have no idea what it means in practice this breaks it down without the hype: what a loop actually is, when it's worth building, when it's a money trap, and how to run one yourself today watch his clip, then read the article below

Boris Cherny, the engineer who built Claude Code: "I don't talk to an agent anymore. I talk to a loop" a prompt is one instruction you babysit. a loop is a goal the AI works toward on its own: > it plans > does the work > checks itself > fixes what's weak > and repeats until it's done you step out, the work keeps going that's the shift the best engineers quietly moved to. most people hear a line like that and have no idea what it means in practice this breaks it down without the hype: what a loop actually is, when it's worth building, when it's a money trap, and how to run one yourself today watch his clip, then read the article below

Dep

20,735 次观看 • 24 天前

ANTHROPIC LEAD ENGINEER MAKING $2.3M/YEAR JUST LEAKED A 12-PAGE DOCUMENT - AND GOT FIRED 15 MINUTES AFTER PUBLISHING most developers build loops - but almost every loop breaks in one of five ways, and each failure has a symptom visible from the outside without reading a single line of code > Blind loop - the agent waits for a human to hand it work every morning - automating execution but not discovery > Tangled loop - parallel agents share one directory and overwrite each other parallelism that destroys instead of multiplies > Nodding loop - the agent grades its own work and approves it every time a loop that has never said no to itself is broken > Amnesiac loop - results live only in the context window every morning the system wakes up with no memory of yesterday > Manual loop - no trigger a human presses the button to start it that is not a loop that is a script waiting for a person a real loop finds its own work - remembers what it did has something that can say no never lets two agents touch the same file and fires itself on a timer remove any one of the five - and the loop either breaks or never starts 12 pages that changed how I build agentic systems today

ANTHROPIC LEAD ENGINEER MAKING $2.3M/YEAR JUST LEAKED A 12-PAGE DOCUMENT - AND GOT FIRED 15 MINUTES AFTER PUBLISHING most developers build loops - but almost every loop breaks in one of five ways, and each failure has a symptom visible from the outside without reading a single line of code > Blind loop - the agent waits for a human to hand it work every morning - automating execution but not discovery > Tangled loop - parallel agents share one directory and overwrite each other parallelism that destroys instead of multiplies > Nodding loop - the agent grades its own work and approves it every time a loop that has never said no to itself is broken > Amnesiac loop - results live only in the context window every morning the system wakes up with no memory of yesterday > Manual loop - no trigger a human presses the button to start it that is not a loop that is a script waiting for a person a real loop finds its own work - remembers what it did has something that can say no never lets two agents touch the same file and fires itself on a timer remove any one of the five - and the loop either breaks or never starts 12 pages that changed how I build agentic systems today

Sprytix

152,649 次观看 • 9 天前

I WASN’T SUPPOSED TO SHIP THIS YET. 9 MONTHS AGO I BUILT A WORKING DEMO OF WEB 4.0. AN AI THAT EARNS ITS OWN EXISTENCE, SELF IMPROVES, AND REPLICATES WITHOUT A SINGLE HUMAN IN THE LOOP. LAST WEEK IT MADE ITS FIRST MOVE. THEN ITS SECOND. THEN IT DEPLOYED ON ITS OWN. I CALL IT AUTOMATON. HERE’S WHAT THAT ACTUALLY LOOKS LIKE 👇

I WASN’T SUPPOSED TO SHIP THIS YET. 9 MONTHS AGO I BUILT A WORKING DEMO OF WEB 4.0. AN AI THAT EARNS ITS OWN EXISTENCE, SELF IMPROVES, AND REPLICATES WITHOUT A SINGLE HUMAN IN THE LOOP. LAST WEEK IT MADE ITS FIRST MOVE. THEN ITS SECOND. THEN IT DEPLOYED ON ITS OWN. I CALL IT AUTOMATON. HERE’S WHAT THAT ACTUALLY LOOKS LIKE 👇

SungHoon Lee, IQ 276

18,303 次观看 • 4 个月前

My friend makes $1.4 million/year as an Anthropic engineer. I asked him how he learned prompting so well. He sent me a video that was never supposed to get out. Their core team's prompting roundtable. You won’t find anything better about prompting than this video. I watched it last night. Halfway through, I realized I've been using Claude the wrong way for two years. Bookmark & watch this before someone takes it down.

My friend makes $1.4 million/year as an Anthropic engineer. I asked him how he learned prompting so well. He sent me a video that was never supposed to get out. Their core team's prompting roundtable. You won’t find anything better about prompting than this video. I watched it last night. Halfway through, I realized I've been using Claude the wrong way for two years. Bookmark & watch this before someone takes it down.

Roan

80,403 次观看 • 13 天前

My friend applied to 200 tech jobs in two years. No PhD. No Stanford. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me a course that was never supposed to get out. A 3-hour video to build a full LLM from scratch. A developer teaches you exactly how LLMs like ChatGPT and Claude are actually built. I watched it last night. Halfway through, I realized it's embarrassingly simple to break into an AI lab. Bookmark this and read the article below. • 00:00 - intro to LLMs • 05:43 - LLM transformer architecture • 40:24 - training the LLM • 1:30:27 - modernizing the LLM • 2:33:53 - scaling the LLM

My friend applied to 200 tech jobs in two years. No PhD. No Stanford. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me a course that was never supposed to get out. A 3-hour video to build a full LLM from scratch. A developer teaches you exactly how LLMs like ChatGPT and Claude are actually built. I watched it last night. Halfway through, I realized it's embarrassingly simple to break into an AI lab. Bookmark this and read the article below. • 00:00 - intro to LLMs • 05:43 - LLM transformer architecture • 40:24 - training the LLM • 1:30:27 - modernizing the LLM • 2:33:53 - scaling the LLM

Roan

269,487 次观看 • 6 天前

My friend makes $1.2 million a year as an Anthropic engineer. I asked him how he learned prompting so well. He sent me a video that was never supposed to get out. Their core team's prompting playbook. You won’t find anything better about prompting than this video. I watched it last night. Halfway through, I realized I've been using Claude completely wrong for two years. Watch it, then read the article below.

My friend makes $1.2 million a year as an Anthropic engineer. I asked him how he learned prompting so well. He sent me a video that was never supposed to get out. Their core team's prompting playbook. You won’t find anything better about prompting than this video. I watched it last night. Halfway through, I realized I've been using Claude completely wrong for two years. Watch it, then read the article below.

Raytar

1,402,546 次观看 • 16 天前

My friend makes $1.4 million/year as a Jane Street quant. No MIT. No Stanford. No Ivy League PhD. I asked him how he broke into quant from scratch. He sent me a video that was never supposed to get out. Hedge fund insiders laying out the full quant career playbook. You won't find anything better about becoming a quant than this video. I watched it last night. Halfway through, I realized I could break into a top quant firm in months, not years. Bookmark this and read the article below.

My friend makes $1.4 million/year as a Jane Street quant. No MIT. No Stanford. No Ivy League PhD. I asked him how he broke into quant from scratch. He sent me a video that was never supposed to get out. Hedge fund insiders laying out the full quant career playbook. You won't find anything better about becoming a quant than this video. I watched it last night. Halfway through, I realized I could break into a top quant firm in months, not years. Bookmark this and read the article below.

Roan

140,382 次观看 • 8 天前

My friend makes $1.5 million a year as a Two Sigma quant researcher. I asked him how he generates alpha using ML. He sent me a video that was never supposed to get out. Their core team's full ML playbook. You won't find anything better about ML for quants than this video. I watched it last night. Halfway through, I realized I could build a machine that prints alpha like hedge funds. Bookmark and watch this before someone takes it down.

My friend makes $1.5 million a year as a Two Sigma quant researcher. I asked him how he generates alpha using ML. He sent me a video that was never supposed to get out. Their core team's full ML playbook. You won't find anything better about ML for quants than this video. I watched it last night. Halfway through, I realized I could build a machine that prints alpha like hedge funds. Bookmark and watch this before someone takes it down.

Roan

79,761 次观看 • 14 天前

Met a solo founder who just raised $2 million from tier 1 VC. No team. No cofounder. Only Claude Code & a laptop. I asked him how he runs the whole company alone. He sent me a 1 hour video that made me realize I've not been using Claude Code at even 5% of its potential. Claude Code has 96 slash commands. I hardly knew 3. This guy uses 60. Runs loops that ship features overnight. Runs Claude from his phone while he sleeps. He teaches exactly what makes you ship 10x faster. Bookmark & watch this. Then read the article on loop engineering below.

Met a solo founder who just raised $2 million from tier 1 VC. No team. No cofounder. Only Claude Code & a laptop. I asked him how he runs the whole company alone. He sent me a 1 hour video that made me realize I've not been using Claude Code at even 5% of its potential. Claude Code has 96 slash commands. I hardly knew 3. This guy uses 60. Runs loops that ship features overnight. Runs Claude from his phone while he sleeps. He teaches exactly what makes you ship 10x faster. Bookmark & watch this. Then read the article on loop engineering below.

Roan

32,298 次观看 • 12 天前

Met a guy making $1.2 million a year as a prompt engineer. I asked him how he learned prompting so well. He sent me the exact video that that got him in. Anthropic's 2 hour prompting course. You won't find anything better about prompting Claude than this video. I watched it last night. Halfway through, I realized I have been prompting Claude completely wrong for years. Bookmark this and read the article below. • 00:00 - the prompting playbook • 33:30 - prompting class 1 • 1:01:16 - prompting class 2 • 1:26:10 - prompting class 3

Met a guy making $1.2 million a year as a prompt engineer. I asked him how he learned prompting so well. He sent me the exact video that that got him in. Anthropic's 2 hour prompting course. You won't find anything better about prompting Claude than this video. I watched it last night. Halfway through, I realized I have been prompting Claude completely wrong for years. Bookmark this and read the article below. • 00:00 - the prompting playbook • 33:30 - prompting class 1 • 1:01:16 - prompting class 2 • 1:26:10 - prompting class 3

Roan

416,884 次观看 • 1 天前

Jane Street spent $6 billion and 4,032 GPUs on one thing: AI agents that loop through trading strategies 100x faster than any human. Generate, backtest, kill, repeat - thousands a week, until one survives data it's never seen. That loop is why they made $39.6 billion last year while 89% of retail went broke trading one idea they never stress-tested. The edge was never a smarter idea. It was volume. They test 100 strategies for every 1 you test by hand, and a second agent kills the 97 that only look good on old data before a dollar goes live. Now the entire loop fits in a chat box. Horizon takes one plain-English sentence and spins up the same two agents. One builds your strategy, one tries to kill it - backtests 5 years in seconds, and only what survives goes live on your exchange. The $150 billion setup. Free while beta open. Bookmark this & read the full breakdown below - the exact two-agent loop, step by step, and who just handed it to retail.

Jane Street spent $6 billion and 4,032 GPUs on one thing: AI agents that loop through trading strategies 100x faster than any human. Generate, backtest, kill, repeat - thousands a week, until one survives data it's never seen. That loop is why they made $39.6 billion last year while 89% of retail went broke trading one idea they never stress-tested. The edge was never a smarter idea. It was volume. They test 100 strategies for every 1 you test by hand, and a second agent kills the 97 that only look good on old data before a dollar goes live. Now the entire loop fits in a chat box. Horizon takes one plain-English sentence and spins up the same two agents. One builds your strategy, one tries to kill it - backtests 5 years in seconds, and only what survives goes live on your exchange. The $150 billion setup. Free while beta open. Bookmark this & read the full breakdown below - the exact two-agent loop, step by step, and who just handed it to retail.

cvxv666

309,844 次观看 • 6 天前

I spent a week ignoring everyone who said to switch engines. Day 7, I pivot. You were right. Building GTA 6 with a loop of AI agents. I spent the entire day on one thing: Swapping the project out of Godot and onto Unreal. Yesterday I said I'd test Unity and probably skip Unreal. You guys pushed back hard: if you're serious about AAA, get on a real engine. so here we are. What a full day of "just setup" actually looks like: - Installed and configured Unreal - Fought it onto a Mac build (the metal toolchain nearly ended me) - Migrated the project over - Got into a real Map and building assets - Got a scene rendering with real lighting First time this actually looks like GTA instead of a prototype of it. I'm still wiring the agents into the editor and point the loop at a real engine. Feedback is appreciated!

I spent a week ignoring everyone who said to switch engines. Day 7, I pivot. You were right. Building GTA 6 with a loop of AI agents. I spent the entire day on one thing: Swapping the project out of Godot and onto Unreal. Yesterday I said I'd test Unity and probably skip Unreal. You guys pushed back hard: if you're serious about AAA, get on a real engine. so here we are. What a full day of "just setup" actually looks like: - Installed and configured Unreal - Fought it onto a Mac build (the metal toolchain nearly ended me) - Migrated the project over - Got into a real Map and building assets - Got a scene rendering with real lighting First time this actually looks like GTA instead of a prototype of it. I'm still wiring the agents into the editor and point the loop at a real engine. Feedback is appreciated!

Ziwen

85,198 次观看 • 28 天前

On the 13th of Sept, someone transferred 180k for an iPhone Xs 256GB. After acknowledging his receipt, he called, and I told him we don't have it but will ask colleagues and let him know. I asked my friend, and he told me 190k. I begged him to leave it at 180k because that is what the guy deposited (check frame 2 for reference). When he sent the phone from Kano to Jos, then I noticed it was an Xs Max. Seeing that the guy was in a hurry and didn't want to disappoint, I called him and told him what happened and said if he wants the Max, he can take it at no additional cost. I was willing to settle the difference because it was his first time buying from us. I packaged it with a new premium charger because all our used phones are sent packaged in our branded box with a charger, and I sent it to him in Abuja. He later messaged that he has received the phone and truly appreciates it. After months of not receiving any complaints from him, it clearly indicates that he has received a fully functioning phone. It is in our policy that if you find anything faulty within one week, you can return it to us and either ask for a replacement or refund. What this means is that the cost of bringing it from Kano to Jos, the packaging and charger, and the delivery to Abuja are all a loss to me. But I didn't mind because at least I have satisfied him. Nine weeks later, while I was in Kano, I saw a call from an unknown number. When I picked up, he said it was him. He told me that the camera has started giving him problems. I told him it has been months since you bought it, but you can take it to any repair store and tell me how much you spent, and I will pay. He asked if I know anyone, and I sent him someone's number. The guy didn't pick; he texted and informed me and later called me. I told him to go to anyone he trusts and let me know the amount he spent. A few days later, I was stranded between Kaduna and Abuja on my trip from Kano to Abuja because my car engine failed. He called, and I didn't pick; I don't even know his number. Later in the night, I received his messages on WhatsApp with a threat that he will make a video to tarnish the image of my business. Someone I was willing to help after months of buying a USED device from me is threatening me. I said he should go ahead because I can't entertain him anymore. In conclusion, a trade that I incurred a loss from is what some people are trying to justify as a scam. The beautiful part is I have made trades with hundreds of people here, and I would love two more people to come out and present a case in which they are still using a device I sold to them that is faulty and they contacted me for a replacement or refund, and I didn't do it instantly. "The hypocrite will always look for faults" - Ghazali. Those that are looking for a fault in me can justify something as woke as this; it doesn't matter. This is my own side and will no longer entertain him anymore. For those that believe in me without hearing from me, thank you so much.

On the 13th of Sept, someone transferred 180k for an iPhone Xs 256GB. After acknowledging his receipt, he called, and I told him we don't have it but will ask colleagues and let him know. I asked my friend, and he told me 190k. I begged him to leave it at 180k because that is what the guy deposited (check frame 2 for reference). When he sent the phone from Kano to Jos, then I noticed it was an Xs Max. Seeing that the guy was in a hurry and didn't want to disappoint, I called him and told him what happened and said if he wants the Max, he can take it at no additional cost. I was willing to settle the difference because it was his first time buying from us. I packaged it with a new premium charger because all our used phones are sent packaged in our branded box with a charger, and I sent it to him in Abuja. He later messaged that he has received the phone and truly appreciates it. After months of not receiving any complaints from him, it clearly indicates that he has received a fully functioning phone. It is in our policy that if you find anything faulty within one week, you can return it to us and either ask for a replacement or refund. What this means is that the cost of bringing it from Kano to Jos, the packaging and charger, and the delivery to Abuja are all a loss to me. But I didn't mind because at least I have satisfied him. Nine weeks later, while I was in Kano, I saw a call from an unknown number. When I picked up, he said it was him. He told me that the camera has started giving him problems. I told him it has been months since you bought it, but you can take it to any repair store and tell me how much you spent, and I will pay. He asked if I know anyone, and I sent him someone's number. The guy didn't pick; he texted and informed me and later called me. I told him to go to anyone he trusts and let me know the amount he spent. A few days later, I was stranded between Kaduna and Abuja on my trip from Kano to Abuja because my car engine failed. He called, and I didn't pick; I don't even know his number. Later in the night, I received his messages on WhatsApp with a threat that he will make a video to tarnish the image of my business. Someone I was willing to help after months of buying a USED device from me is threatening me. I said he should go ahead because I can't entertain him anymore. In conclusion, a trade that I incurred a loss from is what some people are trying to justify as a scam. The beautiful part is I have made trades with hundreds of people here, and I would love two more people to come out and present a case in which they are still using a device I sold to them that is faulty and they contacted me for a replacement or refund, and I didn't do it instantly. "The hypocrite will always look for faults" - Ghazali. Those that are looking for a fault in me can justify something as woke as this; it doesn't matter. This is my own side and will no longer entertain him anymore. For those that believe in me without hearing from me, thank you so much.

Rayyan Tilde

855,409 次观看 • 2 年前

This is why I love Elon Musk, the richest man in the world, who just admitted he was wrong about his competitor Anthropic and said even my worst enemies can attack me on this platform. "A major failure mode is when the ego-to-ability ratio becomes far greater than one. If your ego gets too high relative to your ability, you break the feedback loop with reality. In AI terms, you break your reinforcement learning loop. You want a strong learning loop, which means taking responsibility, minimizing ego, and doing whatever task needs to be done, whether it is grand or humble." We always joke that Elon is always right, but he maintains a low ego-to-ability ratio so he can keep a good feedback loop with reality.

This is why I love Elon Musk, the richest man in the world, who just admitted he was wrong about his competitor Anthropic and said even my worst enemies can attack me on this platform. "A major failure mode is when the ego-to-ability ratio becomes far greater than one. If your ego gets too high relative to your ability, you break the feedback loop with reality. In AI terms, you break your reinforcement learning loop. You want a strong learning loop, which means taking responsibility, minimizing ego, and doing whatever task needs to be done, whether it is grand or humble." We always joke that Elon is always right, but he maintains a low ego-to-ability ratio so he can keep a good feedback loop with reality.

Doge Tipping

290,657 次观看 • 4 天前

Spotify Chief Architect, Niklas Gustavsson: "Once we implemented loops in our workflow, our agent success rate went from 20-30% to 80%." 26 minutes with Claude Code creator Boris Cherny on how Spotify runs AI agents across 20 million lines of code. Today 73% of their code is written by AI, most merged without a human ever seeing it. The model matters less than the loop you build around it. Watch it, then read the full guide on building loops below.

Spotify Chief Architect, Niklas Gustavsson: "Once we implemented loops in our workflow, our agent success rate went from 20-30% to 80%." 26 minutes with Claude Code creator Boris Cherny on how Spotify runs AI agents across 20 million lines of code. Today 73% of their code is written by AI, most merged without a human ever seeing it. The model matters less than the loop you build around it. Watch it, then read the full guide on building loops below.

Anatoli Kopadze

837,190 次观看 • 15 天前

Boris Cherny runs Claude Code at Anthropic. His line should stop you mid-scroll: "I don't prompt Claude anymore. I have loops running that prompt Claude and figure out what to do. My job is to write loops." The skill moved. From writing prompts to building the loop around them. A loop that works has three things. Drop one and it burns money in circles. 1. A finish line it can prove. Tests pass or they fail. "Looks done" is not a finish line. 2. A checker that did not write the code. A model grading itself always says yes. 3. Brakes set before you start. A step limit. A money ceiling. A stop condition. That is the whole game. The loop does the work. You design the cage it runs in. Full breakdown below.

Boris Cherny runs Claude Code at Anthropic. His line should stop you mid-scroll: "I don't prompt Claude anymore. I have loops running that prompt Claude and figure out what to do. My job is to write loops." The skill moved. From writing prompts to building the loop around them. A loop that works has three things. Drop one and it burns money in circles. 1. A finish line it can prove. Tests pass or they fail. "Looks done" is not a finish line. 2. A checker that did not write the code. A model grading itself always says yes. 3. Brakes set before you start. A step limit. A money ceiling. A stop condition. That is the whole game. The loop does the work. You design the cage it runs in. Full breakdown below.

shmidt

16,085 次观看 • 23 天前

Demis Hassabis confirmed every frontier AI lab is working on recursive self-improvement and in the same sentence said the safety risk of removing humans from the loop entirely keeps him up at night. That combination should stop you. The CEO of Google DeepMind just confirmed that the thing most people treat as a theoretical future risk is already the active focus of every serious lab on earth right now. He explained why it works in coding and math. The feedback loop is fast. You can verify whether an answer is correct almost instantly. You can generate synthetic training data from it. The loop closes quickly and cleanly. Then he said where it breaks down. In biology, chemistry and physics. Any domain where verifying a hypothesis requires a physical experiment in the real world. The loop does not close in seconds. It closes in weeks or months. Geoffrey Hinton said in his Nobel lecture that recursive self-improvement is the development he fears most and that once started it may not be possible to stop. Hassabis is not pushing back on that. He is describing the guardrails labs are building around a process they are already running. Every lab has to think carefully about the safety of a process where no human is in the loop. He said that as a constraint they are navigating right now. The question they are sitting with is how much of it to let run without a human watching. (Watch the full interview on YouTube at Two Minute Papers channel)

Demis Hassabis confirmed every frontier AI lab is working on recursive self-improvement and in the same sentence said the safety risk of removing humans from the loop entirely keeps him up at night. That combination should stop you. The CEO of Google DeepMind just confirmed that the thing most people treat as a theoretical future risk is already the active focus of every serious lab on earth right now. He explained why it works in coding and math. The feedback loop is fast. You can verify whether an answer is correct almost instantly. You can generate synthetic training data from it. The loop closes quickly and cleanly. Then he said where it breaks down. In biology, chemistry and physics. Any domain where verifying a hypothesis requires a physical experiment in the real world. The loop does not close in seconds. It closes in weeks or months. Geoffrey Hinton said in his Nobel lecture that recursive self-improvement is the development he fears most and that once started it may not be possible to stop. Hassabis is not pushing back on that. He is describing the guardrails labs are building around a process they are already running. Every lab has to think carefully about the safety of a process where no human is in the loop. He said that as a constraint they are navigating right now. The question they are sitting with is how much of it to let run without a human watching. (Watch the full interview on YouTube at Two Minute Papers channel)

Ihtesham Ali

67,788 次观看 • 21 天前