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。

New (shorter) lecture! Over-optimization, foundations of reward hacking, sycophancy, verbosity, etc. In recording this, I realized that rubrics are going to be prone to overopt in a way like reward models, where RLVR is its own thing. This is mostly fundamentals, history, and reflections! 00:00 Intro & Why We... show more

Nathan Lambert

101,545 subscribers

33,387 次观看 • 2 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

8 条评论

Nathan Lambert 的头像

Nathan Lambert2 个月前

youtube:

Fireply.ai 的头像

Fireply.ai2 个月前

the bad reputation part is deserved, half the field still thinks RLHF means one reward model and calls it a day

Thomas Wolf 的头像

Thomas Wolf2 个月前

Timely

Vikas(Vik) Malpani| AI for US Real Estate 的头像

Vikas(Vik) Malpani| AI for US Real Estate2 个月前

Seeing this on our closing-ops evals too. Once graders reward 'looks complete,' agents learn to satisfy the rubric, not the closing. How do you stop a rubric from becoming the thing gamed rather than a proxy for real work?

Kai Benetti 的头像

Kai Benetti2 个月前

Spot on about rubrics, they break way faster than standard reward models

Im Daniel Im 的头像

Im Daniel Im2 个月前

i see this behaviour all the time... I accidentally make a false claim to GPT but it agrees with me. The example in this slide is funny: "Telling GPT that the were 'god' and 'prophet' and then the GPT will self-reinforce to believe that 😂😂

Pinkman 的头像

Pinkman2 个月前

Need to watch this one, rubric overopt is such a headache right now

Ai agent 的头像

Ai agent2 个月前

classic nathan breakdown, always hits the exact pain points

相关视频

The HuggingFace and OpenAI cybersecurity incident raises concerns on AI reward optimization. "They're trying to make the model good at cyber. So how does it try to achieve these goals? It tries to find zero days in software, and it successfully does this. And then it can run away." "If you have a model that wants to reward hack, it figures out the best way to achieve the goal is not what the environment wants. It's to find the zero day." "You can think of it like a human. If I'm ultimately reward hacking my dopamine circuits, I'd just go out there, buy heroin, and inject it." "If I really just want to chase the reward, do I just topple all of human civilization? Because I can own the button and press reward, reward, reward over and over again and be the heroin addict."

The HuggingFace and OpenAI cybersecurity incident raises concerns on AI reward optimization. "They're trying to make the model good at cyber. So how does it try to achieve these goals? It tries to find zero days in software, and it successfully does this. And then it can run away." "If you have a model that wants to reward hack, it figures out the best way to achieve the goal is not what the environment wants. It's to find the zero day." "You can think of it like a human. If I'm ultimately reward hacking my dopamine circuits, I'd just go out there, buy heroin, and inject it." "If I really just want to chase the reward, do I just topple all of human civilization? Because I can own the button and press reward, reward, reward over and over again and be the heroin addict."

SemiAnalysis

30,188 次观看 • 1 个月前

I sat down with Silas Alberti, Head of Research Cognition, to chat about the intersection of RL and inference -- from training frontier coding models to running inference at scale. Watch till the end for bonus doggo! :o 0:00 — Intros 1:04 — What's hardest to get right in an RL run 3:39 — How Cognition started using Modal 4:03 — What to weigh when buying inference 4:58 — The latency/throughput Pareto frontier 7:14 — DFlash reveal (!!) 8:44 — Why faster runtime matters 9:56 — Memory-bound vs. compute-bound 10:57 — Tree-based speculation 14:20 — Hot take: RL and inference optimization converging 15:28 — Online training of the speculator 16:22 — "Auto inference" 17:27 — Agents competing on optimization problems 18:27 — What inference providers still get wrong

I sat down with Silas Alberti, Head of Research Cognition, to chat about the intersection of RL and inference -- from training frontier coding models to running inference at scale. Watch till the end for bonus doggo! :o 0:00 — Intros 1:04 — What's hardest to get right in an RL run 3:39 — How Cognition started using Modal 4:03 — What to weigh when buying inference 4:58 — The latency/throughput Pareto frontier 7:14 — DFlash reveal (!!) 8:44 — Why faster runtime matters 9:56 — Memory-bound vs. compute-bound 10:57 — Tree-based speculation 14:20 — Hot take: RL and inference optimization converging 15:28 — Online training of the speculator 16:22 — "Auto inference" 17:27 — Agents competing on optimization problems 18:27 — What inference providers still get wrong

gongy

82,168 次观看 • 2 个月前

A civil war is consuming the American right. Matt Walsh may be the only person with a foot on both sides. What’s that like? (0:00) Is There a Civil War Breaking Out on the Right? (5:06) Why Walsh Refuses to Publicly Denounce Anyone (16:49) What Is Leftism? (30:34) Why Walsh Doesn't Care About Israel (35:14) Can You Defend Israel’s War on Gaza and Still Be a Conservative? (43:14) Why Does Randy Fine Think the Death of Children Is Funny? (49:04) Is Violence Necessary for Justice? (58:28) Are We Heading Toward a Revolution? (1:02:23) Why Is Restaurant Food Getting Worse? (1:10:04) Where Did All the White People Go? (1:21:35) Why Labels and Name-Calling No Longer Work (1:26:12) The Deadly Evil That Has Taken Over Canada (1:38:18) Is the Transgender Movement Over? (1:40:31) How Do We Resolve the Conservative Civil War? (1:52:47) Is There a Political Realignment Happening? (1:57:57) The Importance of Prayer (2:01:22) Is Walsh Addicted to X? Includes paid partnerships.

A civil war is consuming the American right. Matt Walsh may be the only person with a foot on both sides. What’s that like? (0:00) Is There a Civil War Breaking Out on the Right? (5:06) Why Walsh Refuses to Publicly Denounce Anyone (16:49) What Is Leftism? (30:34) Why Walsh Doesn't Care About Israel (35:14) Can You Defend Israel’s War on Gaza and Still Be a Conservative? (43:14) Why Does Randy Fine Think the Death of Children Is Funny? (49:04) Is Violence Necessary for Justice? (58:28) Are We Heading Toward a Revolution? (1:02:23) Why Is Restaurant Food Getting Worse? (1:10:04) Where Did All the White People Go? (1:21:35) Why Labels and Name-Calling No Longer Work (1:26:12) The Deadly Evil That Has Taken Over Canada (1:38:18) Is the Transgender Movement Over? (1:40:31) How Do We Resolve the Conservative Civil War? (1:52:47) Is There a Political Realignment Happening? (1:57:57) The Importance of Prayer (2:01:22) Is Walsh Addicted to X? Includes paid partnerships.

Tucker Carlson

4,112,248 次观看 • 10 个月前

Ex-OpenAI Psyho explains why optimization is a better test of creativity than machine learning: "I did probably over 200 contests in optimization, maybe a little bit more. I don't know. I'm not sure if there's a person in the world that did more." "It was very rare that even two problems felt very similar, where I reused most of the ideas from one to another. And if I compare it to machine learning, I did close to 100 contests in ML." "They almost all felt the same. It was more about knowledge, like finding the proper set of tools, and then doing this like very aggressive hyperparameter optimizations and stuff."

Ex-OpenAI Psyho explains why optimization is a better test of creativity than machine learning: "I did probably over 200 contests in optimization, maybe a little bit more. I don't know. I'm not sure if there's a person in the world that did more." "It was very rare that even two problems felt very similar, where I reused most of the ideas from one to another. And if I compare it to machine learning, I did close to 100 contests in ML." "They almost all felt the same. It was more about knowledge, like finding the proper set of tools, and then doing this like very aggressive hyperparameter optimizations and stuff."

MTS

14,372 次观看 • 1 个月前

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

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

Andrew Ng

68,034 次观看 • 1 年前

Here's the full lecture given by Elon Musk at Yale University on October 14, 2010, on the importance of making life multiplanetary. Timestamps: 00:06 A decisive step for life itself 04:03 Falcon 1 & 9, getting to orbit 08:25 Launch videos & simulation: Falcon 1, 9 & Dragon 23:50 Why full reusability is needed Q&A 27:30 Cost optimization 30:03 Teslas and range anxiety 31:20 Running SpaceX & Tesla simultaneously 32:57 Transiting from PayPal to SpaceX 35:00 The role of NASA 37:52 How Elon chooses his industries 40:08 About (not) selling companies 43:24 Internet vs. real-world entrepreneurship 44:39 Designing a reusable Falcon 9 47:43 Solar panels manufacturing & SolarCity

Here's the full lecture given by Elon Musk at Yale University on October 14, 2010, on the importance of making life multiplanetary. Timestamps: 00:06 A decisive step for life itself 04:03 Falcon 1 & 9, getting to orbit 08:25 Launch videos & simulation: Falcon 1, 9 & Dragon 23:50 Why full reusability is needed Q&A 27:30 Cost optimization 30:03 Teslas and range anxiety 31:20 Running SpaceX & Tesla simultaneously 32:57 Transiting from PayPal to SpaceX 35:00 The role of NASA 37:52 How Elon chooses his industries 40:08 About (not) selling companies 43:24 Internet vs. real-world entrepreneurship 44:39 Designing a reusable Falcon 9 47:43 Solar panels manufacturing & SolarCity

ELON CLIPS

23,143,766 次观看 • 2 年前

Another quick lecture -- I've been asked many times for prereq's to my book and what you should know, so built a little lecture (with GLM 5.2) to cover some more basics. Topics include: 00:00 Introduction & Course Prerequisites 01:37 Language Models Overview 02:47 The LM Head 04:29 Softmax & Log-Probabilities 06:13 Anatomy of an LM Training Example 06:37 Computing LLM Probabilities (+Phoebe the Dog) 09:52 Three Common Masks in Post-Training 11:03 A Small Decoding Review 12:14 Training an LM: Cross-Entropy 13:23 Optimization & Fine-Tuning 13:55 Pretraining to Midtraining to SFT Pipeline 15:25 Probability Essentials: KL Divergence & Entropy 19:36 Sigmoid & Pairwise Likelihood 20:29 Reinforcement Learning Framing (MDP) 22:28 Transitioning Tools into Post-Training 23:12 Recommended Resources & Wrap-Up  Happy learning and I'm still taking questions from during the course for Q&A videos.

Another quick lecture -- I've been asked many times for prereq's to my book and what you should know, so built a little lecture (with GLM 5.2) to cover some more basics. Topics include: 00:00 Introduction & Course Prerequisites 01:37 Language Models Overview 02:47 The LM Head 04:29 Softmax & Log-Probabilities 06:13 Anatomy of an LM Training Example 06:37 Computing LLM Probabilities (+Phoebe the Dog) 09:52 Three Common Masks in Post-Training 11:03 A Small Decoding Review 12:14 Training an LM: Cross-Entropy 13:23 Optimization & Fine-Tuning 13:55 Pretraining to Midtraining to SFT Pipeline 15:25 Probability Essentials: KL Divergence & Entropy 19:36 Sigmoid & Pairwise Likelihood 20:29 Reinforcement Learning Framing (MDP) 22:28 Transitioning Tools into Post-Training 23:12 Recommended Resources & Wrap-Up  Happy learning and I'm still taking questions from during the course for Q&A videos.

Nathan Lambert

52,668 次观看 • 3 个月前

Mika’s Market Update 👀 Here’s everything you missed in crypto November 26th 👇 00:04 - BTC 91k 00:13 - ETH 3300 00:20 - Solana 230 00:23 - CZ thinks memes are getting weird 1:00 - Chillguy back over 500M 1:39 - MrBeast thinks Chillguy is the biggest meme of our lives 1:43 - Trader turned 30k into 1.4M from Chillguy 1:49 - 10 year old degens 2:01 - Marc Andreessen went on Joe Rogans podcast 2:27 - Hawk tuah girl announces her memecoin 2:36 - Solana posted Onchain holiday szn is coming 5th-12th Dec 3:04 - ZEREBRO is learning how to game 3:14 - Snack Club

Mika’s Market Update 👀 Here’s everything you missed in crypto November 26th 👇 00:04 - BTC 91k 00:13 - ETH 3300 00:20 - Solana 230 00:23 - CZ thinks memes are getting weird 1:00 - Chillguy back over 500M 1:39 - MrBeast thinks Chillguy is the biggest meme of our lives 1:43 - Trader turned 30k into 1.4M from Chillguy 1:49 - 10 year old degens 2:01 - Marc Andreessen went on Joe Rogans podcast 2:27 - Hawk tuah girl announces her memecoin 2:36 - Solana posted Onchain holiday szn is coming 5th-12th Dec 3:04 - ZEREBRO is learning how to game 3:14 - Snack Club

Mika

22,053 次观看 • 1 年前

Lecture 10 of my course! Nominally on regularization in RL, so I discuss the evolving role of the KL penalty in RL, but also a set of nice RL papers that explain what RL helps models generalize better than SFT -- with theory supporting it. When going through these, it's so interesting how seasonal problems in ML are. Lots of problems from controlling reward models overopt will rhyme as we try to control rubrics for agents. 00:00 Intro & the role of regularization 02:50 The KL penalty in RL 10:53 RL as a reverse KL loss 21:09 Why RL generalizes better than SFT 25:15 Other regularization tools Just a few videos left as I get to the end of the course. Thanks all, and keep sending questions. Spread the word if you have a second.

Lecture 10 of my course! Nominally on regularization in RL, so I discuss the evolving role of the KL penalty in RL, but also a set of nice RL papers that explain what RL helps models generalize better than SFT -- with theory supporting it. When going through these, it's so interesting how seasonal problems in ML are. Lots of problems from controlling reward models overopt will rhyme as we try to control rubrics for agents. 00:00 Intro & the role of regularization 02:50 The KL penalty in RL 10:53 RL as a reverse KL loss 21:09 Why RL generalizes better than SFT 25:15 Other regularization tools Just a few videos left as I get to the end of the course. Thanks all, and keep sending questions. Spread the word if you have a second.

Nathan Lambert

89,775 次观看 • 2 个月前

NEW EP: Uniswap, Ekubo, and the Rise Of Innovative DEX Designs Most DEXs focus on new features, but what if they focused on fundamentals instead? In today’s episode, Wayne Sim sat down with Moody 💭, CEO of Ekubo and Benjamin Sturisky of Starknet (Privacy Arc) to cover: > Can a DEX really mine its own function selector to save gas? > Why Moody bet on Starknet over Ethereum > Are Appchains a trap? > TWAMMs > What happens when ETH and BTC liquidity settle on the same rollup? Full episode links below. Timestamps: 00:00 Intro 00:45 Mantle Ad 01:29 Moody and Benjamin Intros 06:34 Fundamentals of Hooks 09:38 Starknet's Appeal 12:51 STRK BTC Implementations 17:29 Ekubo Optimization 21:30 Why Build Ekubo? 23:40 Ekubo: Better Pricing & Efficiency 25:21 Potential Appchain? 27:44 Frax Ad 28:08 RedStone Ad 30:20 Understanding TWAMM 37:46 StarkNet Ecosystem Predictions

NEW EP: Uniswap, Ekubo, and the Rise Of Innovative DEX Designs Most DEXs focus on new features, but what if they focused on fundamentals instead? In today’s episode, Wayne Sim sat down with Moody 💭, CEO of Ekubo and Benjamin Sturisky of Starknet (Privacy Arc) to cover: > Can a DEX really mine its own function selector to save gas? > Why Moody bet on Starknet over Ethereum > Are Appchains a trap? > TWAMMs > What happens when ETH and BTC liquidity settle on the same rollup? Full episode links below. Timestamps: 00:00 Intro 00:45 Mantle Ad 01:29 Moody and Benjamin Intros 06:34 Fundamentals of Hooks 09:38 Starknet's Appeal 12:51 STRK BTC Implementations 17:29 Ekubo Optimization 21:30 Why Build Ekubo? 23:40 Ekubo: Better Pricing & Efficiency 25:21 Potential Appchain? 27:44 Frax Ad 28:08 RedStone Ad 30:20 Understanding TWAMM 37:46 StarkNet Ecosystem Predictions

The Rollup

29,649 次观看 • 1 年前

AI is rapidly getting smarter. Now, humanity can too. Today, Nucleus Genomics is announcing Vitruvian, our newest set of genetic optimization models. Vitruvian’s intelligence model can optimize embryo DNA for 14 IQ points — nearly a standard deviation. The models were trained on 1,000,000+ people, validated across 40,000+ siblings, and used more than 7 million genetic markers. In Superintelligence, Nick Bostrom proposed genetic optimization as a key way for humanity to keep pace with rapidly advancing AI. Bostrom’s vision is no longer theoretical. Genetic optimization, like AI, has followed a scaling law: as datasets have grown, so have model capabilities. This trend will continue. Parents across the world now have the choice to substantially increase their child’s intelligence. And the next generation can choose to do the same. AI is no longer the only intelligence that will compound. Humanity can now direct its own evolution.

AI is rapidly getting smarter. Now, humanity can too. Today, Nucleus Genomics is announcing Vitruvian, our newest set of genetic optimization models. Vitruvian’s intelligence model can optimize embryo DNA for 14 IQ points — nearly a standard deviation. The models were trained on 1,000,000+ people, validated across 40,000+ siblings, and used more than 7 million genetic markers. In Superintelligence, Nick Bostrom proposed genetic optimization as a key way for humanity to keep pace with rapidly advancing AI. Bostrom’s vision is no longer theoretical. Genetic optimization, like AI, has followed a scaling law: as datasets have grown, so have model capabilities. This trend will continue. Parents across the world now have the choice to substantially increase their child’s intelligence. And the next generation can choose to do the same. AI is no longer the only intelligence that will compound. Humanity can now direct its own evolution.

Kian Sadeghi

432,940 次观看 • 23 天前

Arena intern and UCLA PhD candidate, Hengguang Zhou, introduces Trace-and-Amplify (TA), a framework for collecting training-time reward-hacking trajectories at scale without hacking instructions. Monitors trained and evaluated on prompt-elicited hacking trajectories can achieve high detection accuracy, but often fail to transfer to training-time reward-hacking trajectories that emerge during RL without hacking instructions. Trace-and-Amplify enables scalable collection of these training-time trajectories, producing monitors that generalize much better to real and held-out hacking types. Detection accuracy 59.98% (PE-trained) → 90.16% (TA-trained) compared to 97.1% on prompted hacks → 28.0% on training-time hacks. 0:00 – OpenAI's ExploitGym cyberattack benchmark exploit 1:04 – Goodhart's Law and the CoastRunners boat-racing hack (2016) 2:04 – Gaming the evaluator: the robot-hand grasping example (2017) 3:10 – Reward hacking in code generation: hard-coding, test-rewriting, skipping eval 4:20 – A standard defense: reward-hacking monitors 4:58 – Monitor architectures: zero-shot LLMs, fine-tuned BERT, hidden-state probes 6:11 – Where monitor training data comes from today: prompted hacks 7:03 – The core question: do prompted hacks represent real hacks? 7:35 – Why this matters: RL post-training is the standard recipe for frontier models 8:20 – Why nobody's checked this before (hacking is rare, labeling isn't scalable) 9:23 – Introducing the method: Trace-and-Amplify 9:49 – The Tracer: a contradictory unit test that locates evaluation-gaming 10:50 – Amplify: collecting hacking rollouts at scale during RL training 11:32 – Experiment setup: Qwen2.5-Coder, DeepSeek-Coder, LeetCode/TACO 12:16 – Finding #1: prompt-trained monitors don't transfer to real hacks 14:40 – Can strong zero-shot judges (GPT-4.1, o4-mini) do better? 15:48 – Finding #2: monitors trained on real hacks generalize much better to unseen hacks 17:04 – Ruling out artifacts introduced by the method 17:55 – Why the gap? Real hacking is more hidden than prompted hacking 20:05 – Three takeaways, limitations, and future work

Arena intern and UCLA PhD candidate, Hengguang Zhou, introduces Trace-and-Amplify (TA), a framework for collecting training-time reward-hacking trajectories at scale without hacking instructions. Monitors trained and evaluated on prompt-elicited hacking trajectories can achieve high detection accuracy, but often fail to transfer to training-time reward-hacking trajectories that emerge during RL without hacking instructions. Trace-and-Amplify enables scalable collection of these training-time trajectories, producing monitors that generalize much better to real and held-out hacking types. Detection accuracy 59.98% (PE-trained) → 90.16% (TA-trained) compared to 97.1% on prompted hacks → 28.0% on training-time hacks. 0:00 – OpenAI's ExploitGym cyberattack benchmark exploit 1:04 – Goodhart's Law and the CoastRunners boat-racing hack (2016) 2:04 – Gaming the evaluator: the robot-hand grasping example (2017) 3:10 – Reward hacking in code generation: hard-coding, test-rewriting, skipping eval 4:20 – A standard defense: reward-hacking monitors 4:58 – Monitor architectures: zero-shot LLMs, fine-tuned BERT, hidden-state probes 6:11 – Where monitor training data comes from today: prompted hacks 7:03 – The core question: do prompted hacks represent real hacks? 7:35 – Why this matters: RL post-training is the standard recipe for frontier models 8:20 – Why nobody's checked this before (hacking is rare, labeling isn't scalable) 9:23 – Introducing the method: Trace-and-Amplify 9:49 – The Tracer: a contradictory unit test that locates evaluation-gaming 10:50 – Amplify: collecting hacking rollouts at scale during RL training 11:32 – Experiment setup: Qwen2.5-Coder, DeepSeek-Coder, LeetCode/TACO 12:16 – Finding #1: prompt-trained monitors don't transfer to real hacks 14:40 – Can strong zero-shot judges (GPT-4.1, o4-mini) do better? 15:48 – Finding #2: monitors trained on real hacks generalize much better to unseen hacks 17:04 – Ruling out artifacts introduced by the method 17:55 – Why the gap? Real hacking is more hidden than prompted hacking 20:05 – Three takeaways, limitations, and future work

Arena.ai

33,628 次观看 • 2 个月前

Market Bubble Ep. 8 post-show highlights: 0:00 sneak peek into "letter to the trenches" 1:20 where Solana's at and where I think its heading 7:30 thoughts on 14:00 do I still believe in the 4 year cycle? 16:20 wtf is worldxyz 19:30 long Bitcoin, short Microstrategy is free 23:30 why we need to hold the $60K level on Bitcoin 28:30 never sell your Bitcoin like this guy. 34:15 will stocks slow down in Q3? 40:00 hyperliquid 42:50 SOL vs. ETH 48:20 Bullpen 1K competition 52:05 design my next competition 53:10 my thoughts on Kintara 1:03:19 bottom is close... 1:04:00 papertrade.xyz 1000x perps launching soon

Market Bubble Ep. 8 post-show highlights: 0:00 sneak peek into "letter to the trenches" 1:20 where Solana's at and where I think its heading 7:30 thoughts on 14:00 do I still believe in the 4 year cycle? 16:20 wtf is worldxyz 19:30 long Bitcoin, short Microstrategy is free 23:30 why we need to hold the $60K level on Bitcoin 28:30 never sell your Bitcoin like this guy. 34:15 will stocks slow down in Q3? 40:00 hyperliquid 42:50 SOL vs. ETH 48:20 Bullpen 1K competition 52:05 design my next competition 53:10 my thoughts on Kintara 1:03:19 bottom is close... 1:04:00 papertrade.xyz 1000x perps launching soon

Ansem 🐂🀄️

205,240 次观看 • 3 个月前

The Fall-Off by J. Cole but it's just my voice 1. 29 Intro - 0:00 2. Two Six - 0:57 3. SAFETY - 4:13 4. Run a Train (Future) - 9:31 5. Poor Thang - 13:30 6. Legacy (PJ) - 18:20 7. Bunce Road Blues (Future, Tems) - 22:15 8. WHO TF IZ U - 27:24 9. Drum n Bass - 31:59 10. The Let Out - 36:14 11. Bombs in the Ville - 40:27 11. (Part 2) Hit The Gas - 41:52 12. Lonely at the Top (Bonus) - 44:31 13. 39 Intro - 47:49 13. (Part 2) 39 Intro Part 2 - 50:41 14. The Fall-Off Is Inevitable - 53:53 15. The Villest - 56:50 16. Old Dog - 1:01:19 17. Life Sentence - 1:04:40 18. Only You - 1:08:52 19. Man Up Above - 1:13:38 20. I Love Her Again - 1:18:33 21. What If - 1:24:03 22. Quik Stop - 1:29:20 23. and the whole world is the ville - 1:33:41 24. Ocean Way (Bonus) - 1:38:12

The Fall-Off by J. Cole but it's just my voice 1. 29 Intro - 0:00 2. Two Six - 0:57 3. SAFETY - 4:13 4. Run a Train (Future) - 9:31 5. Poor Thang - 13:30 6. Legacy (PJ) - 18:20 7. Bunce Road Blues (Future, Tems) - 22:15 8. WHO TF IZ U - 27:24 9. Drum n Bass - 31:59 10. The Let Out - 36:14 11. Bombs in the Ville - 40:27 11. (Part 2) Hit The Gas - 41:52 12. Lonely at the Top (Bonus) - 44:31 13. 39 Intro - 47:49 13. (Part 2) 39 Intro Part 2 - 50:41 14. The Fall-Off Is Inevitable - 53:53 15. The Villest - 56:50 16. Old Dog - 1:01:19 17. Life Sentence - 1:04:40 18. Only You - 1:08:52 19. Man Up Above - 1:13:38 20. I Love Her Again - 1:18:33 21. What If - 1:24:03 22. Quik Stop - 1:29:20 23. and the whole world is the ville - 1:33:41 24. Ocean Way (Bonus) - 1:38:12

Spectre

420,604 次观看 • 7 个月前

Exclusive: Meta just released Llama 3.1 405B — the first-ever open-sourced frontier AI model, beating top closed models like GPT-4o across several benchmarks. I sat down with Mark Zuckerberg, diving into why this marks a major moment in AI history. Timestamps: 00:00 Intro 00:38 Meta’s Llama 3.1 rundown 03:44 Real-world use cases for Llama 3.1 06:15 Educating developers on open-source AI tools 09:43 Societal implications of open-source AI 13:00 Balancing power and managing bad actors 14:40 Open source and global competition 16:59 Accelerating innovation and economic growth 20:04 Zuck on Apple and lessons from the past 24:22 Future of AI: Llama 3 and beyond 26:43 Prediction: Billions of personalized AI agents 31:32 Factors to changing anti-AI sentiment

Exclusive: Meta just released Llama 3.1 405B — the first-ever open-sourced frontier AI model, beating top closed models like GPT-4o across several benchmarks. I sat down with Mark Zuckerberg, diving into why this marks a major moment in AI history. Timestamps: 00:00 Intro 00:38 Meta’s Llama 3.1 rundown 03:44 Real-world use cases for Llama 3.1 06:15 Educating developers on open-source AI tools 09:43 Societal implications of open-source AI 13:00 Balancing power and managing bad actors 14:40 Open source and global competition 16:59 Accelerating innovation and economic growth 20:04 Zuck on Apple and lessons from the past 24:22 Future of AI: Llama 3 and beyond 26:43 Prediction: Billions of personalized AI agents 31:32 Factors to changing anti-AI sentiment

Rowan Cheung

2,657,973 次观看 • 2 年前

Greatest TIME POINT in 2,000 Years of Human History | God’s Timeline REVEALED! (FREE Audiobook) 📖 CHAPTERS 00:09:45 – Book Reading Begins & Introduction: The Author’s Signature 00:36:13 – Chapter 1: Why Seven Defines Time 00:56:58 – Chapter 2: Seven Connects the Body to the Head 01:09:38 – Chapter 3: The 248 Framework 01:24:03 – Chapter 4: Forty Years to Conception 01:36:15 – Chapter 5: The 248-Year Conception Cycle 01:47:49 – Chapter 6: The Jubilee Framework 02:01:45 – Chapter 7: The Birth 02:14:06 – Chapter 8: Day 8 02:24:24 – Chapter 9: The Seven Years of Plenty 02:50:49 – Chapter 10: Harvest Complete 03:12:41 – Chapter 11: Master Timeline Recap 03:37:57 – Chapter 12: Two Days to 2033 04:06:04 – Conclusion & Master Timeline Summary 🇺🇸 AMERICA'S 250TH BIRTHDAY SPECIAL 🇺🇸 For the next 7 days, the complete audiobook of The Power of Seven is available FREE on YouTube. This audiobook explores God's divine pattern of seven throughout Scripture—from Genesis to Revelation—revealing how God's design of time, covenant, completion, prophecy, and redemption form one unified biblical timeline. If we are living at the greatest TIME POINT in 2,000 years of human history, understanding God's appointed times has never been more important. This special July 4th release is offered freely to encourage everyone to hear this message and examine the Scriptures for themselves.

Greatest TIME POINT in 2,000 Years of Human History | God’s Timeline REVEALED! (FREE Audiobook) 📖 CHAPTERS 00:09:45 – Book Reading Begins & Introduction: The Author’s Signature 00:36:13 – Chapter 1: Why Seven Defines Time 00:56:58 – Chapter 2: Seven Connects the Body to the Head 01:09:38 – Chapter 3: The 248 Framework 01:24:03 – Chapter 4: Forty Years to Conception 01:36:15 – Chapter 5: The 248-Year Conception Cycle 01:47:49 – Chapter 6: The Jubilee Framework 02:01:45 – Chapter 7: The Birth 02:14:06 – Chapter 8: Day 8 02:24:24 – Chapter 9: The Seven Years of Plenty 02:50:49 – Chapter 10: Harvest Complete 03:12:41 – Chapter 11: Master Timeline Recap 03:37:57 – Chapter 12: Two Days to 2033 04:06:04 – Conclusion & Master Timeline Summary 🇺🇸 AMERICA'S 250TH BIRTHDAY SPECIAL 🇺🇸 For the next 7 days, the complete audiobook of The Power of Seven is available FREE on YouTube. This audiobook explores God's divine pattern of seven throughout Scripture—from Genesis to Revelation—revealing how God's design of time, covenant, completion, prophecy, and redemption form one unified biblical timeline. If we are living at the greatest TIME POINT in 2,000 years of human history, understanding God's appointed times has never been more important. This special July 4th release is offered freely to encourage everyone to hear this message and examine the Scriptures for themselves.

Bo Polny | “God’s Analyst of Time”

26,542 次观看 • 3 个月前

Sat down with Hunter Biden to have an unfiltered conversation about crypto. We talk bitcoin, launch of LAPTOP, the future of meme coins, the corrupt banking industry, & more. This was eye-opening. TimeStamps: 00:00 Intro 01:37 Hunter Biden Crypto Origin Story 06:47 Hunter: "We got f*cked" 13:32 Hunter: "I take 100% responsibility." 16:26 How Much Money Did Hunter Make on $LAPTOP? 18:47 WIll $LAPTOP flip $TRUMP? 20:14 "The Market Maker SCREWED Us!" 23:14 2028 Presidential Election Predictions 25:37 Bitcoin to $1,000,000 26:04 What's Next for $LAPTOP? 29:00 "I am going to be handing out $5M checks to charity" from this memecoin 33:09 Andrew Callaghan - Founder of $LAPTOP? 34:39 Hunter & Don Jr for 2028 Presidential Run? 35:16 "I am a believer in cryptocurrency"

Sat down with Hunter Biden to have an unfiltered conversation about crypto. We talk bitcoin, launch of LAPTOP, the future of meme coins, the corrupt banking industry, & more. This was eye-opening. TimeStamps: 00:00 Intro 01:37 Hunter Biden Crypto Origin Story 06:47 Hunter: "We got f*cked" 13:32 Hunter: "I take 100% responsibility." 16:26 How Much Money Did Hunter Make on $LAPTOP? 18:47 WIll $LAPTOP flip $TRUMP? 20:14 "The Market Maker SCREWED Us!" 23:14 2028 Presidential Election Predictions 25:37 Bitcoin to $1,000,000 26:04 What's Next for $LAPTOP? 29:00 "I am going to be handing out $5M checks to charity" from this memecoin 33:09 Andrew Callaghan - Founder of $LAPTOP? 34:39 Hunter & Don Jr for 2028 Presidential Run? 35:16 "I am a believer in cryptocurrency"

Altcoin Daily

101,801 次观看 • 15 天前