YouTube Digest

English

Based on "How the Top 1% of Learners Use AI to Think Better | Anthropic, Drew Bent" from EO Watch the original video

Beyond the Assistant: How Elite Learners are Mastering AI as a Colleague

The world of artificial intelligence is evolving at a breathtaking pace, transforming from a futuristic concept into an indispensable tool for daily life. Yet, as AI models grow exponentially smarter, many of us are still interacting with them using a mindset rooted in yesterday's capabilities. What if the secret to unlocking AI's true power isn't just about better prompts, but about developing an entirely new social skill?

Drew Bent, Head of Education at Anthropic, a leading AI safety and research company, believes this is precisely the challenge and opportunity before us. With a lifelong dedication to education – from founding a tutoring nonprofit to teaching high school math and now spearheading AI tutors – Bent offers a unique perspective on how the most effective learners and professionals are leveraging AI, and how we can all join their ranks.

The Rise of the "AI-Native" Mindset

Bent argues that a fundamental disconnect exists between those who adopted AI early on and a new generation of "AI-native" individuals. "Someone who was using AI tools in 2022, they sort of see it still as like an assistant," he explains. "Maybe they're misjudging the capabilities." In contrast, the AI-native person, perhaps having grown up with AI from day one, intuitively understands its current power and treats it accordingly.

This isn't just a generational divide; it's a cognitive shift. In places like Rwanda and India, where AI tools are among the first advanced technologies many encounter, users often approach them without the preconceived notions of past limitations. They see AI for what it is capable of today, not what it was last year.

The key takeaway for everyone? "We need to sort of think like the AI native person," Bent urges. This means letting go of the "step-by-step playbook" mentality that once defined early AI interaction. Instead, we must provide AI with rich context and then empower it to "do its thing."

From Prompt Engineering to Social Collaboration

The early days of AI focused heavily on technical prompting – learning the exact syntax to get a desired output. But Bent contends that era is over. "Ultimately, you have to treat this more as a colleague, as a collaborator," he states. "And so then it becomes more like a social skill."

Interacting effectively with AI is no longer just about commands; it's about building a rapport, understanding its strengths and weaknesses, and guiding it through complex problems. This requires a different kind of intelligence from us – one that mirrors how we interact with human teammates. We learn to communicate our intentions, provide necessary background, and trust in its evolving capabilities.

This evolving relationship demands that we constantly "elevate our ambition" with the types of problems we present to AI. Many of us, still mentally anchored to older models, offer AI tasks it could have handled a year ago. Yet, Bent observes colleagues achieving "all sorts of powerful things" he didn't even know were possible, simply by pushing the boundaries.

The challenge is amplified by AI's exponential growth. "We as humans are very used to linear or like static things," Bent notes. "And so it's like you have a colleague and every day you come in to work, they're just getting exponentially better. We as humans don't have a way to think about that." This cognitive gap means we often underestimate AI's current prowess, treating it like the version from last month, when its capabilities may have doubled.

Education's AI Paradox: Speed vs. Understanding

Bent's passion for education naturally leads him to explore AI's profound impact on learning. Anthropic recently conducted a study on coding education, splitting students into two groups: one using AI tools, the other not. As expected, the AI-enabled group finished their assignments much faster. However, a follow-up assessment, conducted without AI, revealed a surprising result: the non-AI group performed 17% better, demonstrating a deeper understanding of the concepts.

This highlights a critical warning for AI in education: the risk of "skill atrophy." If AI simply provides answers transactionally, students might complete tasks quickly but fail to internalize the underlying knowledge.

The good news? Not all AI users suffered this setback. Those who engaged with AI in an "inquiry way," probing and asking questions rather than just seeking solutions, performed well on the final assessment. "How you use the AI tools really matters," Bent emphasizes. "It's not just that you're using them or you're not using them. It's the way in which you're engaging." The goal shouldn't just be to complete a task, but to get smarter in the process.

This also means reframing how we approach AI with problems. Instead of coming with a specific solution in mind and asking a narrow question, Bent advises, "come with the problem, not the solution." Presenting an open-ended challenge allows AI to leverage its reasoning capabilities more effectively, guiding you through the struggle rather than just spitting out a pre-determined answer.

The Future of Learning: Personal, Contextual, and Human-Centered

Bent envisions a future where AI revolutionizes education, not by replacing humans, but by enhancing human connection and enabling truly personalized learning at scale. His career has been driven by the "elusive dream" of one-on-one tutoring for everyone, a dream now within reach thanks to AI.

In a 2030 classroom, Bent predicts AI will operate seamlessly in the background, almost invisibly. Teachers might use it to craft personalized lesson plans, group students effectively, and create rich learning environments previously only available in elite private schools. He shares how teachers in a global WhatsApp group are already building custom flashcard apps and formative assessments with tools like Claude, transforming their teaching in real-time.

Crucially, future AI learning companions will be deeply contextualized. They won't just offer generic curricula but understand your school's specific curriculum, your state's standards, and most importantly, you. "It understands where you're coming from and you get to choose to give it this context so that you're growing with it and AI tools also growing with you," Bent explains. This moves beyond the current, often frustrating, experience of repeatedly reminding an AI who you are and what you're working on.

While AI tutors will be amazing, Bent stresses the irreplaceable value of human interaction. Drawing on his experience with Schoolhouse, a peer-to-peer tutoring platform he co-founded, he highlights the "magical" phenomenon of students from Russia, Colombia, the US, and China learning together. They share cultural perspectives on learning, forming connections that transcend borders. "Having someone who cares about your progress, who holds you accountable is equally important, if not more important," he asserts. AI, in this vision, serves to foster and deepen these human-to-human connections, putting humans firmly at the center of the learning experience.

Building Your AI Team: The Essential Skill of Tomorrow

Beyond education, the implications of this new AI-human collaboration extend to every professional field. "We essentially have birthed this new species of artificial intelligence into the world," one expert in the discussion points out. Just as we spend years learning to interact productively with other humans, we now need to develop skills for collaborating with AI. This is a "whole type of different skill," a social skill, in fact, that will define professional success for decades to come.

This new collaboration isn't just about single interactions; it's about building entire teams of AI agents. The expert shares a compelling personal anecdote: as the sole marketing person at his company, he built a team of 40 AI marketing agents. This AI team provides weekly coaching, identifying successful themes and styles, and has generated posts with millions of impressions. The financial implications are staggering: "A high quality marketing contractor, let's say I had four people at $12,500 per month each. That would be $50,000 a month. And my AI bill is $500 a month."

This capability – "building AI agents is the fundamental skill that will define every professional's career for the next 30 years," the expert declares. "It's a requirement. In the same way that knowing how to use spreadsheets like Microsoft Excel was a requirement for the last 40 years, knowing how to build AI agents is a requirement for the next 40 years."

The Underhyped Revolution

The message is clear: AI is not just another tool; it's a new form of intelligence demanding a new form of interaction. It requires us to shed old assumptions, embrace experimentation, and constantly elevate our ambitions. We need to treat AI as a powerful, exponentially improving colleague, provide it with ample context, and engage with it in an inquiry-driven, collaborative manner.

This journey won't always be efficient in the short term. As Bent advises, we should dedicate a "fraction of our time" to experimentation, understanding that "today you're actually wasting a little time to use these tools, but tomorrow you will be saving time." This R&D mindset is crucial for staying at the cutting edge.

Ultimately, the consensus from these leading voices is resounding: "We are all using it for probably 1% of what we should be using it for." And perhaps the most surprising, yet accurate, summary of the current landscape: "AI is way underhyped." The true revolution is just beginning, and those who master the art of collaborating with this new species will be the ones who think better, learn faster, and shape the future.

Based on "5x Your OpenClaw: Full Course (AI Employee)" from Greg Isenberg Watch the original video

Unleash Your Digital Doppelgänger: Mastering OpenClaw to Create Your AI Employee

Jensen Huang, CEO of Nvidia, recently declared that every company needs an "OpenClaw strategy," hailing it as the "new computer." But for many, the leap from a compelling vision to a fully functional, reliable digital employee remains a daunting challenge. How do you move beyond the initial installation to a robust, personalized AI agent that truly works for you in the real world?

Greg Isenberg, host of the Startup Ideas podcast, sat down with AI expert Moritz to demystify OpenClaw, offering a comprehensive, one-hour masterclass that transforms a theoretical concept into a practical blueprint. This isn't about creating "AI slop"; it's about wiring up a sophisticated system that thinks, remembers, and acts like a superhuman extension of yourself.

The Evolution of AI Agents: From Chatbots to Autonomous Employees

Before diving into the "how," it’s crucial to understand what OpenClaw is and how it distinguishes itself from other popular AI tools.

ChatGPT: The Cloud-Based Conversationalist Moritz explains ChatGPT as primarily a chat intelligence living in the cloud. While it has evolved to include memory and basic tool use like web search, its fundamental nature remains a conversational interface. You interact with an intelligence residing remotely, offering convenience but limited local interaction.

Claude Code: The Local Coder's Companion The next paradigm shift arrived with Claude Code, whose key differentiator is its local operation on your machine. This local presence empowers it to read and write files directly, making it incredibly useful for coding tasks where managing vast local file structures is essential. Claude Code also boasts more flexible tools and context management, though its primary focus remained code-centric.

OpenClaw: The Truly Personal and Proactive Agent OpenClaw takes the concept of a local agent several steps further, embodying the vision of a truly autonomous digital employee:

Claude Co-Work and the Future Landscape: Moritz observes that Claude Co-Work emerged as a user-friendly interface for Claude Code, and Anthropic (the creator of Claude) is now playing catch-up, adding features seen in OpenClaw, such as mobile access ("Dispatch"). However, OpenClaw's open-source nature means it will likely remain the most powerful and customizable option, akin to choosing Linux over Windows for its flexibility and community-driven development.

Beyond the Hype: Mastering Your OpenClaw Setup – A 10-Step Guide

Many users attempt to install OpenClaw, only to get stuck when errors arise or it fails to deliver on its promise. Moritz's 10-step optimized setup is designed to prevent these frustrations, transforming your OpenClaw from a novelty into an indispensable digital employee.

1. Build Your AI's First Aid Kit: The Troubleshooting Baseline

When issues inevitably arise, you don't want your AI scrambling for answers online or making things up. Moritz's ingenious first step is to create a "troubleshooting baseline."

2. Crafting a Digital Persona: Personalization is Key

For your OpenClaw to truly sound and act like you, it needs context. Upon installation, OpenClaw creates a workspace folder containing crucial files:

3. The Secret to Smarter AI: Persistent Memory

A common complaint is OpenClaw's apparent forgetfulness. The solution lies in understanding and actively managing its memory:

4. Powering Your AI Brain: Model Selection and Fallbacks

Choosing the right model and ensuring reliability is paramount for cost and performance.

5. Taming the Conversation Chaos: Optimized Chat Management

A single, sprawling chat thread with your OpenClaw quickly becomes unmanageable.

6. Giving Your AI Eyes and Hands: The Power of the Browser

OpenClaw's ability to interact with the internet is a core component of its autonomy, but it's more nuanced than a simple web search.

7. Expanding Capabilities: Harnessing and Building Skills

Skills are the building blocks of OpenClaw's functionality, allowing it to perform complex tasks.

8. The Pulse of Your AI: Understanding Heartbeat and Cron Jobs

The heartbeat file (heartbeat.md) is a script that runs automatically every 30 minutes, keeping your OpenClaw "alive."

9. Fortifying Your Digital Fortress: Essential Security Measures

Security is paramount when giving an AI agent access to your digital life. Moritz identifies two main risks:

10. Unleashing Potential: Real-World Use Cases

With an optimized and secure OpenClaw, the possibilities are vast. Moritz shares a glimpse into one of his powerful systems:

Your Path to a Superhuman Digital Employee

Moritz's masterclass provides a clear, tactical roadmap for anyone looking to harness the full power of OpenClaw. By meticulously addressing personalization, memory, model selection, chat management, browser capabilities, skills, heartbeat, and security, you can transform a complex AI agent into a reliable, proactive, and truly personal digital employee. The journey from initial install to a fully optimized OpenClaw that's "working for me" is now within reach, promising a future where your digital doppelgänger handles the heavy lifting, freeing you to focus on what matters most.

Based on "Who Is Winning the War in Iran?" from New York Times Podcasts Watch the original video

The Unwinnable War? Why Iran Endures Despite Unprecedented Military Strikes

By a New York Times Correspondent

Three weeks into a conflict that has reshaped the geopolitical landscape, the question echoes across the Middle East and beyond: Who is truly winning the war in Iran? On one side, the United States and Israel have unleashed a military campaign of unprecedented scale, systematically dismantling Iran's conventional military capabilities and eliminating key leadership figures. Yet, despite these massive blows, the Iranian regime has not buckled. Instead, it has hardened, finding new, devastating ways to strike back, challenging the world's most powerful militaries through a cunning strategy of asymmetric warfare.

As the conflict stretches on, the initial military successes touted by Washington are increasingly overshadowed by Iran's stubborn resilience and its potent economic leverage, leaving President Trump grappling with a set of options that range from "bad to super bad to worse."

A Hammer Blow from the West

From a purely military perspective, the US and Israeli forces believe they are executing their campaign with remarkable efficiency, even ahead of schedule. "The United States is decimating the radical Iranian regime's military in a way the world has never seen before," boasted one official, a sentiment echoed by military commanders who view their progress as substantial.

In less than three weeks, the scale of the offensive has been staggering:

The human cost, while tragic, is, from the Pentagon's perspective, "on the relatively low side" for American forces, with 13 US service members killed in action (six due to an accidental collision). However, the overall toll is much higher, with the New York Times estimating at least 2,100 deaths on the ground, including over 300 civilians, primarily within Iran. Neighboring countries like Saudi Arabia, the UAE, and Qatar have also suffered casualties from Iranian attacks.

The Unyielding Foe: Asymmetric Warfare and the "Mosaic Defense"

Despite losing top commanders and a significant portion of its military hardware, Iran has refused to back down. The regime, weakened but unbowed, has adopted a strategy of asymmetric warfare, understanding it cannot confront the combined might of the US and Israeli militaries head-on. Instead, it has leveraged a decentralized approach, what experts call a "mosaic defense."

"Iran had a very decentralized system of attack," explains New York Times correspondent Eric Schmidt. "They have some 30 different districts basically that they've assigned defenses to. So that if you knock out the central command and control in Tehran or any one of these important leaders... these independent districts, their commanders already have a set of general instructions and they can continue carrying out attacks with the weapons that they have."

This decentralized command structure has allowed Iran to continue launching retaliatory strikes across the region, using underwater vehicles to target tanker ships and deploying cluster munitions to breach Israel's sophisticated air defenses, causing damage and fatalities. The regime's resilience has surprised many, proving "more hardened" than anticipated.

Iran's Ace in the Hole: The Strait of Hormuz

The most potent weapon in Iran's arsenal, however, has proven to be economic warfare, specifically its ability to choke off the Strait of Hormuz. This narrow, strategic waterway, through which a significant portion of the world's oil and gas flows, has become Iran's ultimate leverage.

"With a handful of mines, but just the threat of attacks, they've basically brought international commerce to a trickle," Schmidt notes, describing the global shockwaves this disruption is already sending through the economy. Nearly 20 different tankers – oil and cargo – have been struck, deterring shipping companies and insurers from risking passage.

Iran's methods are nimble and difficult to counter:

This low-tech, high-impact strategy underscores the paradox of the conflict: "We are seeing the most powerful and sophisticated military in the world being stymied essentially by this totally decentralized strategy carried out by a seriously weakened country that has had 90% of its missile capacity destroyed," observes Natalie Kitroeff. Even if 99% of the threat is eliminated, the remaining 1% can still cause immense damage and disruption.

A Strategic Miscalculation?

The rapid escalation of economic warfare in the Strait of Hormuz appears to have caught the US administration off guard. While military and intelligence leaders, including General Dan Kaine and Admiral Brad Cooper, had briefed President Trump on the predictable threat to the Strait, the speed and ferocity with which Iran deployed this tactic were unexpected.

"Many believe that the Iranians wouldn't reach for this card right away," Schmidt explains. "But the problem with that is it's also been brief that by the intelligence community was that if Iran believed that it was an existential threat to the regime that they might reach for that ace in the hole, bottling up the straight of Hormuz faster than certainly this administration anticipated."

The US was clearly unprepared. Its mine-sweeping capabilities are antiquated, and crucial naval assets were not positioned in the Gulf region. Moreover, the administration had not marshaled international support for a protective operation, leading to President Trump's frustrated demands for European and Asian allies to "pony up ships and other resources," a request met with reluctance given the US's initial unilateral approach.

Trump's Perilous Options

With Iran demonstrating its willingness to fight indefinitely and the global economic fallout mounting, President Trump faces a series of unenviable choices, described as ranging from "bad to super bad to worse":

  1. Tanker Escorts in the Strait of Hormuz:

    • The Plan: The US Navy, potentially with allies, would escort commercial vessels through the 21-mile-wide Strait, employing destroyers with advanced radar, drones, and helicopters to counter mines, missiles, and speedboats.
    • The Risks: An incredibly complex and dangerous operation with a high probability of US warships being hit and American sailors wounded or killed. Ultimately, the success hinges on shipping companies and insurers deeming the risk acceptable.

  2. Seizing Karge Island:

    • The Plan: An amphibious landing by US Marines to seize Karge Island, Iran's main oil hub (90% of its oil production). The US military has already bombed military installations there, but avoided the oil infrastructure to prevent global economic shock. Seizing the island would exert immense economic pressure.
    • The Risks: The Marines would have to navigate the Strait, their approach would lack strategic surprise, and once seized, the island would become a constant target for remaining Iranian forces. Crucially, if the oil infrastructure is damaged during invasion or occupation, the entire point of the mission could be lost. There's also no guarantee this pressure would stop Iran's asymmetric tactics.

  3. Destroying Iran's Nuclear Material:

    • The Plan: This option addresses President Trump's stated goal of denying Iran a nuclear weapon. It could involve continued bombing of underground bunkers storing highly-enriched uranium (at facilities like Isvahan), or a far more dangerous commando operation. Specially trained commandos, accompanied by nuclear experts, would infiltrate these tunnels to extract or neutralize the gaseous uranium canisters.
    • The Risks: An "incredibly risky and dangerous and kind of insane" undertaking. A pierced canister could release highly toxic and radioactive gas. An inadvertent chain reaction is a possibility. Securing the operation would require several hundred troops to seize the surrounding territory, making it a "fight to the death" scenario for Iranian defenders.

  4. Declaring Victory and Disengaging:

    • The Plan: President Trump could declare the military mission accomplished, highlighting the destruction of Iran's military capacity and the elimination of key leaders, arguing that the US is now safer.
    • The Risks: While politically appealing, it doesn't mean the war would actually end. Iran could continue its asymmetric attacks, activate terror cells globally, and Israel, which has its own unfulfilled goals, might not agree to disengage. The fundamental problem of containing a "weakened but hardline" regime would remain.

The Elusive Goal of Regime Change

One of the initial justifications for the war was regime change in Iran. However, after three weeks of intense military pressure, most American intelligence analysts believe this outcome is highly unlikely. The regime, though battered, has proven resilient, showing "no serious defections." Instead, the most probable outcome is a "badly weakened state" filled with even harder-line elements, capable of continued repression and unwilling to negotiate.

Ultimately, the war in Iran has become a complex dilemma for President Trump, balancing military achievements against unfulfilled political goals and escalating economic consequences. The question of "who is winning" remains elusive, as military success has not translated into a clear path to victory. The Iranian regime, despite its losses, has found a way to exert significant influence, demonstrating that even the most powerful military cannot easily subdue a determined, decentralized foe willing to exploit global economic vulnerabilities.

As the situation continues to unfold, Qatar's state-owned energy company recently reported extensive missile damage to a major energy hub, blaming Iran. In a late Wednesday night Truth Social post, President Trump, while blaming Israel for an earlier attack on an Iranian gas field, threatened Iran, stating that if the country attacked Qatar again, the US would "blow up the rest of the oil field." The conflict, far from a clear resolution, appears poised for further escalation, with unpredictable and far-reaching consequences.