<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0"
     xmlns:content="http://purl.org/rss/1.0/modules/content/"
     xmlns:dc="http://purl.org/dc/elements/1.1/"
     xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The Multiplied Self</title>
    <link>https://themultipliedself.com/</link>
    <description>An independent publication on using AI to become more capable in everyday life — practical ideas, experiments, and better ways to think and work.</description>
    <language>en</language>
    <copyright>© 2026 The Multiplied Self</copyright>
    <managingEditor>hello@themultipliedself.com (The Multiplied Self)</managingEditor>
    <webMaster>hello@themultipliedself.com (The Multiplied Self)</webMaster>
    <generator>The Multiplied Self</generator>
    <lastBuildDate>Tue, 17 Nov 2026 09:00:00 +0000</lastBuildDate>
    <ttl>1440</ttl>
    <atom:link href="https://themultipliedself.com/feed.xml" rel="self" type="application/rss+xml" />
    <image>
      <url>https://static.readdy.ai/image/a27c572c077dee3e3f0d75b52d205ce4/a7135c9fe00667140abff06f60415c55.jpeg</url>
      <title>The Multiplied Self</title>
      <link>https://themultipliedself.com/</link>
    </image>

    <item>
      <title>The One-Person Team</title>
      <link>https://themultipliedself.com/editions/the-one-person-team</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/the-one-person-team</guid>
      <pubDate>Tue, 17 Nov 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Working</category>
      <category>Tools</category>
      <category>Experiments</category>
      <description>My team is not human: five AI tools with different jobs, one person deciding. Plus a four-step ladder for testing the idea you keep putting off.</description>
      <content:encoded><![CDATA[
<p><em>Edition 07 — 17 November 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80,width=1200/uploads/asset/file/588510fe-44ca-4819-ac99-5278f457dba6/Tools-Around-an-Idea.png?t=1791532547" alt="Different tools gathered around a single idea, showing that each one has a different strength." width="1200" height="675" />
<figcaption>Different tools, different strengths, one person bringing them together.</figcaption>
</figure>
<p>I have a team now.</p>
<p>The funny thing is, none of them are human.</p>
<p>ChatGPT helps me think through decisions, organize my work and coordinate the other tools.</p>
<p>Gemini and Perplexity help me research topics and find supporting evidence for each weekly edition.</p>
<p>Claude reviews my drafts, challenges my arguments and points out gaps in the reasoning and flow.</p>
<p>Canva helps me create visuals, while Wix and Readdy help me build and improve the website.</p>
<p>Each has a different role. I bring their contributions together, check how they fit, and decide what makes it into the final product.</p>
<p>And I’m still the only person working on The Multiplied Self.</p>

<h2>Why this newsletter exists</h2>
<p>When I started, a lot of the conversations I came across about AI focused on individual tasks. Ask a question. Summarize a document. Find some information.</p>
<p>All useful.</p>
<p>But I kept wondering about a bigger question:</p>
<blockquote>What could one person attempt if trying got cheaper?</blockquote>
<p>Could someone start the thing they’d been putting off? Build something without knowing how to code? Explore an idea without the money to hire experts or the years to learn every skill first?</p>
<p>That question is why The Multiplied Self exists. Every Tuesday, I try one experiment and invite you to try it too.</p>

<h2>How it’s built</h2>
<p>My professional background is in data and analytics. I’m comfortable working with data, technology, and analytical tools, but I’m not a web developer or a graphic designer.</p>
<p>Yet I built <a href="https://themultipliedself.com">The Multiplied Self website</a> without becoming one. I used AI to help me research topics I wanted to explore, create images instead of purchasing stock photography, and bring ideas to life that would once have required skills I didn’t have or people I would have needed to hire.</p>
<p>I didn’t become an expert in all these disciplines. I found a way to extend what I could do with the skills I already had.</p>
<p>Here’s how I think about it. Instead of expecting one tool to do everything, I ask: What kind of help do I need right now? Research, a challenge to my thinking, a second pair of eyes on my writing, a designer, a developer. Different jobs, different tools. None is reliably good at everything, so I check the work.</p>
<p>And there are parts I don’t hand over:</p>
<ul>
<li>Which ideas are worth exploring</li>
<li>Which suggestions actually hold up</li>
<li>What’s worth sharing with you</li>
<li>Where this goes next</li>
</ul>
<p>Direction, judgment, taste and responsibility stay with me. AI changed how far I can get. It didn’t change who decides.</p>

<h2>The pattern behind the editions</h2>
<p>If you’ve been reading from the start, you may have noticed that every edition does the same thing in a different way.</p>
<p>We’ve used AI to surface evidence we might overlook, drawing inspiration from Darwin’s habit of recording observations that challenged his existing beliefs. We’ve made it play the curious beginner to test whether we actually understood something. We’ve used it as a flashlight to find useful passages in documents nobody wants to read. Last week, we gave it a problem from a field we know well to see what it missed.</p>
<p>Each experiment hands AI a job. None hands it the decision.</p>
<p>That’s the idea behind this newsletter, and this edition is where I’d like to say it plainly: <strong>AI can lower the cost of trying. What you do with that is yours.</strong></p>

<h2>The Possibility Ladder</h2>
<p>We often rule out an idea before we’ve explored it. Maybe you want to build an app, write a book, start a small business, or learn about a subject you’ve always found fascinating. But perhaps you lack the coding skills, writing confidence, market knowledge, or connections to make it happen. You look at everything involved and decide it’s too much.</p>
<p>Sometimes that’s the right call. But sometimes we’re judging an idea by what we can do today, rather than what we could do with the right help.</p>
<p>Here’s a way to find out before deciding. Climb one rung at a time:</p>
<ol>
<li><strong>Name the idea.</strong> What would you love to try? For me, it was building a website for this newsletter.</li>
<li><strong>Map the requirements.</strong> What skills, resources and help would it need? In my case, design, code, content and somewhere to host it.</li>
<li><strong>Identify the help.</strong> Where could AI assist, and where would you still need another person?</li>
<li><strong>Run one small test.</strong> What could you try before committing to the whole project? For me, it was generating a single-page HTML mock-up in Claude and opening it in my browser to see if the layout felt right.</li>
</ol>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80,width=1200/uploads/asset/file/3d560710-cb83-49d6-abab-7c2c1ff6e3cc/The-Possibility-Ladder-Diagonal.png?t=1791881470" alt="A diagonal ladder of four steps, each one a smaller ask than committing to the whole idea." width="1200" height="675" />
<figcaption>One rung at a time. Each step asks less of you than committing to the whole idea.</figcaption>
</figure>
<p>You haven’t committed to the whole project. You’ve learned something about what it would take and whether you want to explore it further.</p>
<p>You might discover it isn’t viable. You might need expertise AI can’t provide. Or you might find you can do more than you expected.</p>
<p>The idea doesn’t have to wait until you have the entire team. You can start building the team around the idea.</p>

<h2>Try this: bring your own idea to the table</h2>
<p>Think of something you’ve wanted to do but haven’t pursued because you lacked the time, money, expertise, or help. It doesn’t have to be a business. Maybe it’s learning something new, creating something for your family, or exploring an interest you’ve put aside.</p>
<h3>Copy-paste AI prompt: test your idea before building</h3>
<div class="prompt">
<p>There’s something I’ve wanted to do for a while, but I haven’t pursued it because of the time, money, skills, or resources involved.</p>
<p>Here’s the idea:</p>
<p>[Describe what you want to do and why it matters to you.]</p>
<p>Help me explore whether it’s realistic. First, ask me questions to understand the idea, my goals, and my constraints before offering suggestions. Ask one question at a time.</p>
<p>Then help me identify what I would need to learn, what skills and resources the project requires, and where AI could help me.</p>
<p>Be honest about what AI cannot do, where I might need outside help, and where this idea is unlikely to work. Identify the biggest assumptions and obstacles, and suggest one small, practical step I could take to test the idea before committing too much time or money.</p>
</div>
<p>When you get the response, don’t treat it as a verdict. Challenge the suggestions, check the assumptions, and see what you learn. Some tasks need real expertise, access to resources, or another human being. A confident answer from AI doesn’t make that go away.</p>

<h2>What one person can do</h2>
<p>The Multiplied Self is still a one-person project. I have limited time, there’s a lot I don’t know, and using AI doesn’t guarantee success. But I can explore more ideas and try things I once would have hesitated to attempt.</p>
<p>That’s what an AI team offers: the chance to pursue ideas you might otherwise have left on the shelf.</p>
<p>The point isn’t to replace myself. It’s to multiply myself.</p>
<p><strong>Let’s multiply ourselves!</strong></p>
<p>P.S. What idea would you climb the ladder for? Hit reply and tell me. I’m really curious.</p>
<p>💡 Know someone with an idea they keep putting off? Forward them this post. The first rung is smaller than they think.</p>
<p><em>Next Tuesday: a Second World War statistician, a pile of bullet-ridden bombers, and a question to ask before you commit to any idea you’re excited about.</em></p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

    <item>
      <title>What We Don’t Know About AI (And How to Be Ready Anyway)</title>
      <link>https://themultipliedself.com/editions/what-we-dont-know-about-ai</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/what-we-dont-know-about-ai</guid>
      <pubDate>Tue, 10 Nov 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Thinking</category>
      <category>Uncertainty</category>
      <category>Experiments</category>
      <description>A $25 million deepfake call, a widening AI jobs gap and five questions about AI that nobody can fully answer — plus three habits for deciding anyway.</description>
      <content:encoded><![CDATA[
<p><em>Edition 06 — 10 November 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,width=1200,onerror=redirect/uploads/asset/file/1d77a77b-71e1-43c9-8767-af8df24ba655/image.jpeg?t=1793776639" alt="The cover image for What We Don’t Know About AI — a corporate video call in which every face but one is a deepfake." width="1200" height="675" />
</figure>
<p>A finance employee in Hong Kong received a message that appeared to come from his company’s Chief Financial Officer, asking for a secret payment. He suspected a scam.</p>
<p>Then he joined a video call with the CFO and several colleagues. He recognized their faces and voices. Everything appeared to confirm the request.</p>
<p>Every other person on that call was a deepfake.</p>
<p>Over the following week, he made 15 transfers totaling HK$200 million, about US$25.6 million. He discovered the deception only later, after checking with the company’s head office.</p>
<p>The company was Arup, the engineering firm behind projects such as the Sydney Opera House. The incident happened in 2024, and it points to a much bigger challenge.</p>
<p>AI is changing how we work, what we trust, how information is produced, and how decisions get made. Some consequences are becoming visible. Others are still difficult to see. Each change brings uncertainty.</p>
<p>How do we prepare ourselves to make good decisions when we don’t know what’s coming?</p>
<p>A perfect forecast is impossible. What could help is a set of ways to recognize change, question what we’re being told, and adapt when our assumptions stop holding up.</p>
<p>Here are five questions that can help us do that.</p>

<h2>1. Is AI a bubble?</h2>
<p>Nobody knows yet, and that uncertainty is worth understanding.</p>
<p>The scale of investment is extraordinary. Bloomberg reported that Amazon, Microsoft, Meta, and Alphabet were collectively on track to spend around $750 billion on capital expenditure in 2026, much of it on AI data centers and computing capacity.</p>
<p>At the same time, investors are asking whether the revenue and productivity gains will justify that spending. The question became harder to ignore on October 29, when Alphabet lost about $475 billion in market value in a single day, its largest one-day loss in the company’s history.</p>
<p>One day later, a different picture emerged. On October 30, Apple closed as the world’s most valuable company, ahead of Nvidia. Both stocks fell that day, but Nvidia fell further, as investors looked for the relative safety of a company that has spent far less on AI infrastructure than its rivals.</p>
<p>Apple’s next generation of AI models, including the new Siri, will be built on Google’s Gemini. Bloomberg reported the arrangement could cost Apple about $1 billion a year, though the companies haven’t disclosed the terms.</p>
<p>The title has changed hands since, so the ranking matters less than what it illustrates: companies can find different ways to create value from AI. Some invest heavily in the infrastructure. Others may capture value by building AI into products, services, and customer relationships.</p>
<p>A market correction could happen while AI continues to transform the economy.</p>
<p>The internet did not disappear when the dot-com bubble burst. Many companies failed, valuations collapsed, and investment became more disciplined. The underlying technology continued to change the world.</p>
<p>Something similar could happen with AI. The technology can be genuinely useful, demand can continue to grow, and some companies can still over-invest or struggle to stay afloat.</p>
<h3>What this means for us</h3>
<ul>
<li>Avoid major personal or financial decisions that depend on today’s AI enthusiasm continuing indefinitely.</li>
<li>A technology can be transformative without every company built around it becoming successful.</li>
</ul>

<h2>2. Will AI take our jobs?</h2>
<p>This is probably the question people feel most personally.</p>
<p>The headlines in 2026 have been difficult to ignore. Amazon announced plans to eliminate around 16,000 roles in January. Meta announced plans to cut around 8,000 positions. Google expanded voluntary exit offers to employees in several countries. Canva recently announced organizational changes affecting 1,271 employees.</p>
<p>The reasons behind these decisions differ. Some companies are reducing management layers and simplifying their structures. Others are redirecting investment toward AI infrastructure and changing how work gets done.</p>
<p>There are examples of a more direct connection. Meta’s CEO has discussed the trade-off between spending on AI infrastructure and spending on people. Block’s CEO has said that AI tools allow a significantly smaller team to do more.</p>
<p>There are two changes worth watching. AI is beginning to take over tasks people used to perform. Companies are also reorganizing around the expectation that smaller teams, supported by AI, can accomplish more.</p>
<p>The effects can reach employees before AI is capable of doing their entire job. A company might slow hiring, combine responsibilities, eliminate a role, or decide that a team no longer needs as many people.</p>
<p>How much of this is already showing up in employment data?</p>
<p>A Stanford Digital Economy Lab study using payroll data covering millions of US workers through June 2026 found no evidence of widespread, economy-wide job displacement. It did, however, find that employment among workers aged 22–25 in occupations highly exposed to AI was about 19% below where it would have been if it had kept pace with employment among similarly aged workers in less-exposed occupations.</p>
<p>The researchers found that the pattern appears to reflect reduced hiring more than layoffs. They describe it as a pattern in the data, not proof that AI caused the change.</p>
<p>That distinction matters. We are seeing changes in how companies allocate money, organize work, and think about future hiring. The full effect on employment is still unfolding.</p>
<h3>What this means for us</h3>
<p>Start by looking at your own work, not just your job title.</p>
<p>List your tasks and sort them into three groups:</p>
<ol>
<li>Tasks AI could probably help with.</li>
<li>Tasks where AI could assist but needs your judgment.</li>
<li>Tasks where human experience and accountability remain central.</li>
</ol>
<p>Start with a low-stakes task from the first group, where you can check the result and the consequences of a mistake are limited.</p>
<p>Keep developing the capabilities that help you work through unfamiliar situations: asking better questions, understanding context, exercising judgment, building relationships, and taking responsibility.</p>
<p>If you’re early in your career, seek opportunities to learn by doing. Find people who will let you observe, practice, ask questions, and take on real responsibility.</p>
<p>You don’t need to predict exactly which jobs will disappear. You do need to understand how the work around you is changing, and keep building the skills that help you adapt.</p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,width=1200,onerror=redirect/uploads/asset/file/ff07a037-cefa-4b75-8cf1-17311069d8b3/image.jpeg?t=1793776906" alt="A professional working at a laptop in a quiet office, representing the skills needed in a changing workplace." width="1200" height="675" />
<figcaption>The deeper question is what we’ll need to learn to keep contributing.</figcaption>
</figure>

<h2>3. Can we trust who we’re talking to?</h2>
<p>The Arup case shows how easily familiar faces and voices can overcome reasonable doubts.</p>
<p>Now imagine receiving a call that sounds exactly like your child, saying they’re in trouble and urgently need money. Or a message from your manager asking you to make a confidential payment.</p>
<p>The technology is making impersonation more convincing. Familiarity is no longer enough to establish identity.</p>
<h3>What this means for us</h3>
<p>Create a simple verification habit before you need it.</p>
<p>For family members, agree on a code word or another way to verify an urgent emergency request. If a call feels unusual, hang up and call the person back using a number you already know.</p>
<p>At work, verify unusual payment requests through a separate, trusted channel. A second message on the same chat or a video call arranged by the same person is not independent confirmation.</p>
<p>The check that exposed the Arup fraud came after the money had moved. Build yours in before.</p>

<h2>4. Can we trust AI’s answers?</h2>
<p>AI can produce a clear, confident answer even when it is wrong.</p>
<p>It can invent sources, overlook recent developments, misunderstand a question, or build a convincing explanation around a faulty assumption.</p>
<p>The most dangerous answer sounds reasonable enough that we stop questioning it.</p>
<p>I’ve found it useful to think about verification in proportion to consequence.</p>
<p>If AI suggests a recipe, I might try it and see what happens. If it summarizes a financial document, helps interpret a medical result, or informs an important career decision, I need a different level of scrutiny.</p>
<p>The question I ask is: <strong>What would happen if this answer were wrong?</strong></p>
<h3>What this means for us</h3>
<p>Don’t treat a polished answer as proof.</p>
<p>Ask what evidence supports the answer, and check the original source when the details matter. Verify the parts that could change your decision.</p>
<p>AI can help us think through a problem. It cannot take responsibility for the decision we make.</p>
<p>Use it as another perspective. The final decision stays with you.</p>

<h2>5. Who is accountable when AI makes a decision?</h2>
<p>AI is increasingly being added to systems that help make decisions.</p>
<p>Employers use software to screen job applications. Banks use automated systems to assess loans. Insurers use automated tools to review claims. These decisions affect someone’s livelihood, opportunities, or access to essential services.</p>
<p>Imagine your loan application is declined, and the explanation is that a system applied a rule to your file. The rule may be right in general and still fit your situation poorly: you recently changed jobs, your income arrives irregularly, or your record contains a mistake.</p>
<p>The shift is beginning to reach payments too. Mastercard has confirmed that its strategic partnership with OpenAI includes Agent Pay, its technology for enabling AI agents to initiate payments using authorized, tokenized credentials.</p>
<p>When an AI agent can initiate a payment, questions about the authority it has been given, the limits on its spending, and how its actions can be traced become very practical concerns.</p>
<p>Governments are heading in this direction too. On November 2, Abu Dhabi announced four agreements with OpenAI. One of the initiatives, called Machine-Readable Government, aims to make selected government policies and regulations easier for AI systems to interpret and apply. Its first pilot focuses on food import assessments.</p>
<p>The stated approach is that assessments can be traced to the governing rules, while government officials retain oversight and accountability for final decisions.</p>
<p>These examples involve different uses of AI, but they raise a common question. AI may help make a decision, apply a rule, or execute an action. Who remains accountable for the outcome?</p>
<p>What happens when AI applies a rule correctly, but the rule doesn’t quite fit your situation?</p>
<h3>What this means for us</h3>
<p>When an AI-assisted decision affects you, ask how it was made, who is accountable for it, and how you can ask for a review. Keep a record of the reasons you were given.</p>
<p>Rules on explanations and human review vary by country, so ask either way. Don’t assume that because a system produced an answer, the answer is beyond challenge.</p>
<p>Efficiency matters. So does recognizing when a situation needs human judgment.</p>

<h2>Preparing without predicting</h2>
<p>Of these five questions, the one I watch most closely is what AI means for work, particularly the opportunities people have to learn their craft early in their careers.</p>
<p>I don’t think any of us can confidently predict how all of this will unfold.</p>
<p>I use AI regularly. I experiment with it, find ways to make everyday tasks easier, and check important things before relying on them. I also know there are things I cannot see coming yet.</p>
<p>That is probably true for most of us.</p>
<p>So what can we do?</p>
<p>I’ve come to think about preparedness as three habits. They’re simple, but they give us a way to respond without pretending we know exactly what the future holds.</p>
<h3>1. Try small things.</h3>
<p>Don’t wait until you feel you understand AI completely. Pick a low-stakes task and experiment. Ask it to help you plan a trip, explain something unfamiliar, organize information, or improve a first draft.</p>
<p>Notice what works, what doesn’t, and what surprised you.</p>
<p>You build useful experience by trying things.</p>
<h3>2. Check what matters.</h3>
<p>Match the amount of checking to what’s at stake. Use the question from section 4: what would happen if this answer were wrong?</p>
<h3>3. Revisit your assumptions.</h3>
<p>Something that works well today may not work the same way six months from now. A task you thought required hours may become much easier. A tool you rely on may change. A new risk may emerge.</p>
<p>Keep learning. Review what you believe. Be willing to change your approach when the evidence changes.</p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,width=1200,onerror=redirect/uploads/asset/file/a4348b4e-f474-4d8c-8f9e-07da89543d03/image.jpeg?t=1793777003" alt="A person walking toward an open horizon, representing an uncertain future." width="1200" height="675" />
<figcaption>We don’t need to know exactly what’s coming to start preparing for it.</figcaption>
</figure>

<h2>Try this: find out what AI doesn’t know</h2>
<p>Here’s a small experiment you can try this week.</p>
<p>Choose a subject you know well. It could be something from your work, a hobby, a sport, or a skill you’ve developed through experience.</p>
<p>Then open your AI assistant and try this:</p>
<div class="prompt">
<p>I know a lot about [add your subject]. Here is a real situation: [describe it with specific details, constraints, and what you’re trying to achieve].</p>
<p>What would you advise, and why? Explain the trade-offs. Tell me which assumptions you’re making and what information you’d need to be more certain.</p>
</div>
<p>Give it a real-world situation that needs more than a definition or a list of facts. An avid photographer might write: “I’m shooting a street portrait at golden hour, in strong backlight, with a 50mm f/1.8 lens and no flash. What settings and positioning would you recommend?”</p>
<p>Read the answer like an experienced person would.</p>
<p>Did it miss a detail that someone with experience would notice?</p>
<p>Did it make an assumption that doesn’t hold in practice?</p>
<p>Did it sound sure without enough evidence?</p>
<p>Then compare its list of assumptions with your own experience. If it missed something you caught, that’s the lesson: its self-check can overlook the very errors that matter. If it listed everything you expected, ask yourself what you might not know to look for.</p>
<p>If it doesn’t make a mistake, make the question harder. Add a constraint, introduce an unusual situation, or ask it to explain the trade-offs.</p>
<p>Knowing that AI can make mistakes is easy. Recognizing those mistakes when they appear takes experience and judgment.</p>
<p>As AI becomes more capable, those human capabilities become increasingly valuable.</p>
<p>The future is uncertain. Our response doesn’t have to be.</p>
<p>We won’t be able to anticipate every new tool, every disruption, or every unintended consequence. We can stay curious, check what matters, and avoid confusing confidence with certainty.</p>
<p>What is one thing about AI’s future that you find difficult to predict, and what are you doing to prepare for it?</p>
<p>I’d genuinely like to hear your perspective. Reply and tell me.</p>
<p>Know someone worried about where AI is heading? Forward them this post. It might help them feel a little more ready.</p>
<p><strong>Let’s multiply ourselves!</strong></p>
<p><em>Next Tuesday: I have a team now, and none of them are human. Plus a four-step ladder for testing the idea you’ve been putting off.</em></p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

    <item>
      <title>The Fine Print, Finally Readable</title>
      <link>https://themultipliedself.com/editions/the-fine-print</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/the-fine-print</guid>
      <pubDate>Tue, 03 Nov 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Reading</category>
      <category>Documents</category>
      <category>Experiments</category>
      <description>I gave AI a 61-page rental agreement and checked every quote it pulled. All five matched word for word — and the context still needed a human eye.</description>
      <content:encoded><![CDATA[
<p><em>Edition 05 — 3 November 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,format=auto,onerror=redirect/uploads/asset/file/6963fa65-6248-4d6c-b601-0d16ddaa888e/The-Fine-Print-Finally-Readable.png?t=1791448882" alt="The cover image for The Fine Print, Finally Readable — a page of dense contract wording with a few lines lit up." width="1200" height="675" />
</figure>
<p>On April 1, 2010, a British video game retailer called GameStation added a clause to its terms and conditions.</p>
<p>By clicking “I agree,” customers were supposedly giving the company the right to claim their immortal souls.</p>
<p>It was an April Fools’ joke. There was even a way out: anyone who noticed the clause could opt out and receive a £5 voucher.</p>
<p>According to GameStation, 88 percent of customers didn’t opt out, and around 7,500 people agreed to hand over their souls.</p>
<p>The prank was funny. What stayed with me was how easily people agreed to something they hadn’t read.</p>
<p>And we do versions of this all the time.</p>
<p>A subscription. An insurance document. A rental agreement. A workplace policy. A software contract.</p>
<p>Click. Scroll. Agree. Next.</p>
<p>I don’t think most of us are careless. Important information gets buried among everything else.</p>
<p>A 60-page document might contain only a handful of paragraphs that genuinely matter to you. Finding them takes time and attention most of us aren’t going to spend on every document put in front of us.</p>
<p>So what if we had help finding the parts worth reading?</p>

<h2>How to use AI to summarize contracts and leases</h2>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,format=auto,onerror=redirect/uploads/asset/file/57d5e978-00c6-4d18-8eec-323a7eef6039/A-Few-Lines-Illuminated.png?t=1791449065" alt="AI as a flashlight in a dark room full of paper, pointing to what deserves human attention." width="1200" height="675" />
<figcaption>AI as a flashlight in a dark room full of paper, pointing to what deserves human attention.</figcaption>
</figure>
<p>Imagine a dark room full of paper.</p>
<p>You don’t need someone to describe every object in the room. You need a flashlight. Something that shows you what’s important, where to look, and what deserves a closer read.</p>
<blockquote>That’s the role I’d give AI. Not a lawyer. Not the person making the decision. A flashlight.</blockquote>
<p>Take a document you actually need to understand: a lease, an insurance policy, a subscription agreement, or a benefits document. An old one you haven’t looked at in years works fine.</p>
<p>Remove personal information first (names, addresses, account numbers), and make sure you’re allowed to share the document with the AI tool you’re using. Check its privacy settings too.</p>

<h2>Copy-paste AI prompt for reading fine print</h2>
<div class="prompt">
<p>Find five important clauses in this document that could affect me financially or change my obligations.</p>
<p>For each one:</p>
<p>Quote the exact wording.</p>
<p>Identify the page or section.</p>
<p>Explain it in plain English.</p>
<p>List any conditions, exceptions, or related clauses that could change its meaning.</p>
<p>Clearly separate what the document explicitly says from your interpretation. If something depends on information outside the document, tell me. Don’t fill in gaps or assume anything that isn’t in the document.</p>
</div>

<h2>I tried it on a very old rental agreement</h2>
<p>I had one sitting around. It was 61 pages long, packed with clauses, fees, conditions, and the kind of language most of us skim until we have a reason not to. I used an earlier version of the prompt above.</p>
<p>I asked AI to identify five important clauses and quote the exact wording. Then I checked each quotation against the original agreement.</p>
<p>All five matched. Word for word.</p>
<p>That was reassuring. But the more interesting part came when I looked beyond the quotes.</p>
<p>One clause, for example, referred to the return of the security deposit within 45 days. Read on its own, that sounds straightforward. But the agreement also listed conditions that had to be satisfied before that happened.</p>
<p><strong>The words were right. The context mattered just as much.</strong></p>

<h2>Check three things</h2>
<ol>
<li><strong>Check the words.</strong> Find each quotation in the original. Does it match exactly? AI can paraphrase something in a way that sounds perfectly reasonable but isn’t what the document says. If you can’t find the quote, don’t trust the point until you’ve worked out what happened.</li>
<li><strong>Check the context.</strong> Read the paragraphs around it. Has AI left out a condition or exception?</li>
<li><strong>Check the interpretation.</strong> Does the plain-English explanation actually follow from the wording, or has AI made an assumption?</li>
</ol>
<p>Don’t stop at whether the quote is accurate. Ask whether the explanation gives you the whole picture.</p>
<p>Then try one follow-up:</p>
<div class="prompt">
<p>List the pages and sections you used to prepare this summary. Identify any pages you couldn’t access, sections you didn’t examine, or anything you couldn’t verify from the document. Don’t claim to have reviewed anything you haven’t actually examined.</p>
</div>
<p>This gives AI an opportunity to flag what it couldn’t read or verify. But treat its list as a starting point, not proof. It can claim to have reviewed everything and still be wrong. Check the important references yourself.</p>
<p>You may find that using AI this way makes you trust it less. That’s a good result. You stop taking its word for what the document says, and start reading the paragraph it points to.</p>
<p><em>One important caveat: AI can make mistakes, so verify its claims against the original document. For consequential legal, financial, insurance, or employment decisions, consult an appropriate professional.</em></p>
<p>AI can’t make every document simple, and it shouldn’t make the decision for you. But it can help you spend your limited attention where it matters most.</p>
<p>The machine does the searching.</p>
<p><strong>You do the deciding.</strong></p>
<p><strong>Let’s multiply ourselves!</strong></p>
<p>P.S. Which document are you going to try this on first? Hit reply and tell me. And if you know someone who has a lease or a contract they’ve been avoiding, send this their way.</p>
<p><em>Next Tuesday: a $25 million video call, and five worries about AI that nobody can fully answer yet.</em></p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

    <item>
      <title>The Curious Beginner</title>
      <link>https://themultipliedself.com/editions/the-curious-beginner</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/the-curious-beginner</guid>
      <pubDate>Tue, 27 Oct 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Thinking</category>
      <category>Learning</category>
      <category>Experiments</category>
      <description>What if explaining something to a curious beginner revealed what you don’t actually understand? Use AI as a patient beginner and find the gaps.</description>
      <content:encoded><![CDATA[
<p><em>Edition 04 — 27 October 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,width=1200,onerror=redirect/uploads/asset/file/683986ef-196a-4d33-841c-7d4f92164c1b/Thumbnail-Curious-Beginner-Watch.png?t=1791448439" alt="The cover image for The Curious Beginner — a reminder that explaining something simply can reveal what you do not yet understand." width="1200" height="675" />
</figure>
<p>My six-year-old was trying to create GIFs on WhatsApp using AI.</p>
<p>At some point, he asked me what AI actually was.</p>
<p>I started explaining.</p>
<p>Then I realized I was using words he didn’t know to explain something he didn’t understand.</p>
<p>“It’s trained on lots of data…”</p>
<p>Blank look.</p>
<p>So I tried again. And again.</p>
<p>The more I tried to simplify it, the more obvious it became that I was leaning on words like “data,” “training” and “models” without really explaining what they meant.</p>
<p>Then I tried a different approach. I explained AI as a student learning from its teachers. It learns from the examples we give it. The next time we ask it something, it uses what it has learned to come up with a reply.</p>
<p>That was a much easier way to talk about it.</p>
<p>And it made me realize something: I hadn’t needed more technical words. I’d needed a better way to explain the idea.</p>
<p>My six-year-old had accidentally given me a test:</p>
<p><strong>Could I actually explain AI, or did I just know how to talk about it?</strong></p>
<p>Which is funny, because in the experiment below, AI plays the student.</p>

<h2>When an explanation falls apart</h2>
<p>I wasn’t the first person to discover that explaining something simply can be harder than it looks.</p>
<p>Nobel Prize winning physicist Richard Feynman faced a similar challenge. David Goodstein, a physicist at Caltech, once asked him to explain a difficult idea at a level beginning students could understand. Feynman took pride in explaining profound ideas to beginners, so, <a href="https://calteches.library.caltech.edu/563/2/Goodstein.pdf">he agreed to prepare a freshman lecture</a>.</p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,format=auto,width=1200,onerror=redirect/uploads/asset/file/268c3467-03e2-4b48-94bd-27d506715269/image.png?t=1791447530" alt="Richard Feynman standing at a blackboard in front of his notes." width="1200" height="897" />
<figcaption>Richard Feynman at the blackboard. Source: <a href="https://kottke.org/17/06/if-you-cant-explain-something-in-simple-terms-you-dont-understand-it">Kottke.org</a></figcaption>
</figure>
<p>Not a lecture for physicists. A lecture for the first-year students.</p>
<p>A few days later, he came back and admitted he couldn’t do it. He couldn’t reduce the idea to the freshman level.</p>
<p>In Goodstein’s telling, Feynman came to a conclusion that if he couldn’t explain the idea at a freshman level, perhaps he hadn’t understood it as well as he thought.</p>
<p>I find that remarkable. He didn’t blame the audience or lower his expectations. He treated the difficulty as a clue that there was something he hadn’t fully worked through.</p>
<p>Of course, not every difficult idea can be explained in a single introductory lecture, and struggling to explain something doesn’t automatically mean you don’t understand it. But the attempt can reveal where your explanation depends on words or steps you haven’t really examined.</p>

<h2>Make AI the beginner</h2>
<p>What if you could use AI to test how well you actually understand something?</p>
<p>Pick an idea from work, something you’ve been reading, a hobby, or anything you think you could explain to someone else. Open a chat and explain it in your own words. Don’t prepare a polished explanation.</p>
<h3>Try this prompt</h3>
<div class="prompt">
<p>I want to explain the following topic in my own words:</p>
<p>[Write the topic you want to explain.]</p>
<p>Pretend you’re a curious beginner encountering this explanation for the first time.</p>
<p>Don’t compliment me or assume my explanation is correct. Stop me when I use a word I haven’t explained, skip a step, or make a claim without explaining why.</p>
<p>Ask one question at a time. If something is unclear, ask me to explain it rather than explaining it yourself. Keep going until you can follow my explanation without having to guess what I mean.</p>
</div>
<p>Then see where the conversation takes you.</p>
<p>Maybe you’ll get stuck trying to explain a word you use all the time. Maybe you’ll realize you know what happens but can’t explain why. Or maybe you’ll discover that a step you thought was obvious isn’t obvious at all.</p>
<p>There’s a catch. AI already knows a great deal, so it might understand your explanation even when a real beginner wouldn’t. That’s why you’ve asked it to question you instead of filling in the missing pieces.</p>
<p>Pay attention to the question that makes you pause. <strong>That’s the moment to work through the explanation yourself, rather than asking AI for the answer.</strong></p>
<p>The goal isn’t to produce a better explanation for AI. It’s to find out what you understand, and what you only thought you understood.</p>

<h2>Why teaching can help you learn</h2>
<p>There’s research on what’s known as the <strong>protégé effect</strong>: preparing to teach someone else changes how you approach learning.</p>
<p>In one study, eighth grade students learned biology using a system called <a href="https://eric.ed.gov/?id=EJ855299">Betty’s Brain</a>. Some were told they were teaching the computer character, while others were told they were learning for themselves. The students in the teaching group spent more time on learning activities such as reading, and performed better on the study’s learning measures.</p>
<p>The interesting part isn’t the computer. It’s what changes when you’re responsible for helping someone else understand. You can’t just recognize the right answer. You have to connect ideas, explain why they fit together, and notice where your understanding falls short.</p>
<p>AI gives us a patient partner for practicing that. You can explain an idea, find where it falls apart, try again, and follow a question wherever it leads. No need for someone else to ask.</p>
<p><strong>Let’s multiply ourselves!</strong></p>
<p>P.S. If you try this, reply with the word or idea that stopped you. I’m curious how often the hardest part turns out to be explaining something we thought we already knew.</p>
<p><em>Next Tuesday: a retailer, a clause about immortal souls, and how to find the paragraphs worth reading in documents nobody reads.</em></p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

    <item>
      <title>The Facts That Don’t Fit</title>
      <link>https://themultipliedself.com/editions/the-facts-that-dont-fit</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/the-facts-that-dont-fit</guid>
      <pubDate>Tue, 20 Oct 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Thinking</category>
      <category>Beliefs</category>
      <category>Experiments</category>
      <description>Darwin wrote down the evidence that went against his own theory because he knew he’d forget it. Here’s a way to try the same habit with AI.</description>
      <content:encoded><![CDATA[
<p><em>Edition 03 — 20 October 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/aeffe6f9-30d5-4eae-ba86-43d5a2f8ffa3/The_Facts_That_Dont_Fit.jpeg?t=1791956621" alt="The cover image for The Facts That Don’t Fit — an open notebook of notes that contradict one another." width="1200" height="675" />
</figure>
<p>Charles Darwin had a rule for himself.</p>
<p>Whenever he came across a fact, an observation or a thought that contradicted one of his ideas, he wrote it down immediately.</p>
<p>He called it his &quot;golden rule&quot; in his autobiography. He had noticed that inconvenient evidence slipped from his memory faster than evidence he liked.</p>
<p>I suspect most of us do this, myself included.</p>
<p>When we believe something, we notice the things that confirm it. We remember the examples that support our point. And when something doesn’t quite fit, we find a reason to put it aside.</p>
<p>I’ve been thinking about whether AI assistants are making us more confident, or just telling us what we want to hear. Darwin’s rule keeps coming back to me when I do.</p>
<p>Because AI can make this habit easier. Give it your opinion and ask it to build your case, and it will pull together supporting arguments, organize your reasoning and make the whole thing sound convincing.</p>
<p>You might end up with a better argument. You might also end up more certain of something you haven’t really examined.</p>
<p>There’s a name for one version of this problem: sycophancy. AI systems can sometimes adapt their responses to what a user seems to believe or prefer, rather than giving the most accurate answer.</p>
<p>So I started wondering whether I could use AI the other way round. Not to help me make my case, but to find the parts of it I hadn’t thought through.</p>

<h2>Give the other side a chance</h2>
<p>I wanted to try this with something I actually believed. There’s not much value in testing an opinion you don’t care about.</p>
<p>So here’s the experiment. Start with a view you hold. Explain what you believe and why. Then give AI this prompt.</p>

<h3>Try this prompt</h3>
<div class="prompt">
<p>Here is what I currently believe, and why.</p>
<p>[Explain what you believe and why.]</p>
<p>What are the strongest arguments thoughtful people who disagree with me make? Give me the strongest version of their case, not a weak version that’s easy to dismiss.</p>
<p>Explain what they see that I might be missing, and what assumptions or evidence their position rests on. Identify the important factual claims and provide sources I can check. If you can’t verify a claim, say so.</p>
<p>Don’t try to persuade me to change my mind. Help me understand their perspective before I evaluate it.</p>
</div>

<p>There’s a difference between asking AI to disagree with you and trying to understand why someone might genuinely see things differently. Research by psychologist Charlan Nemeth and colleagues found that genuine dissent in group discussions stimulated more original thinking than assigning someone to play devil’s advocate. It wasn’t about AI, but it left me with a question: was I really considering another perspective, or just collecting arguments I could knock down?</p>
<p>Once AI gives you its answer, find the strongest objection to your position.</p>
<p>Write it down.</p>
<p>Don’t answer it yet. Don’t explain why it’s wrong. Just give it a little room.</p>
<p>Then ask this:</p>

<div class="prompt">
<p>What evidence would genuinely weaken my original position? What would I need to find out to determine whether that evidence is reliable?</p>
</div>

<p>Now examine the evidence. Does it actually support the claim, or does it merely sound persuasive? AI might give you a weak argument, misrepresent what people believe, or invent supporting evidence, so check important claims against reliable sources.</p>

<h2>What did you find?</h2>
<p>Look at the objection you wrote down.</p>
<p>Did it genuinely challenge your position? Did it show you something you hadn’t considered? Or did you find a point you can easily answer?</p>
<p>Any of those is possible. You might even conclude that your original view holds up.</p>
<p>But pay attention to what happens next.</p>
<p>You might think, &quot;That’s fair.&quot;</p>
<p>You might think, &quot;No, that’s completely wrong.&quot;</p>
<p>Or you might find yourself saying, &quot;I don’t actually know.&quot;</p>
<p>That last one interests me the most. Sometimes we have strong opinions and plenty of reasons for holding them, without having examined how much those reasons really support them.</p>
<p>Darwin’s rule wasn’t about doubting every idea he had. It was about making sure he didn’t lose sight of the evidence that went against him.</p>
<p>So here is the question I’m taking with me:</p>

<blockquote>What am I at risk of forgetting because it doesn’t fit what I already believe?</blockquote>

<p>You don’t have to change your mind.</p>
<p>Let’s multiply ourselves!</p>
<p>P.S. If you try this, hit reply and tell me which objection you wrote down. I’d love to know whether it changed anything.</p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

    <item>
      <title>The AI That Interviews You</title>
      <link>https://themultipliedself.com/editions/the-ai-that-interviews-you</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/the-ai-that-interviews-you</guid>
      <pubDate>Tue, 13 Oct 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Thinking</category>
      <category>Prompts</category>
      <category>Experiments</category>
      <description>What happened when I stopped asking AI for advice and asked it to question me instead. A simple experiment in using AI to think better.</description>
      <content:encoded><![CDATA[
<p><em>Edition 02 — 13 October 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/c15090c5-ab9b-4ebb-b6f7-e2504d3bab43/The-AI-That-Interviews-You-Bold.png?t=1791302660" alt="The cover image for The AI That Interviews You — a reminder that a better question can matter more than an answer." width="1200" height="675" />
</figure>
<p>I had two ideas for a newsletter, and I couldn't decide which one to pursue.</p>
<p>The first was practical: tutorials about using AI, optimized for search. Clear answers to questions people were already asking. A sensible way to build an audience.</p>
<p>The second was what eventually became The Multiplied Self: a publication exploring how AI might change the way we think, learn, create, and live. Stories, experiments, and questions about what becomes possible when we work with AI differently.</p>
<p>I asked AI to compare the two. We went through the pros and cons, potential audiences, search traffic, and growth opportunities.</p>
<p>The analysis was useful. It was also getting me nowhere.</p>
<p>I had more information than when I started, but I still couldn't decide.</p>
<p>Then I changed the question.</p>
<blockquote>Don't give me advice. Interview me.</blockquote>

<h2>When the questions changed</h2>
<p>Instead of comparing the two ideas for me, AI started asking questions about what I actually wanted.</p>
<p>What was I optimizing for?</p>
<p>What would I regret not trying?</p>
<p>Which idea would I still want to pursue if nobody knew about it?</p>
<p>Then came one I hadn't considered:</p>
<blockquote>&quot;If both publications reached 10,000 subscribers, which group of readers would you actually want to spend your time talking to every Tuesday?&quot;</blockquote>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f8712a80-29b9-40bf-a70a-82901d0e3a3f/Screenshot_Claude_Conversation_About_Newsletter.png?t=1791622200" alt="A screenshot of a Claude conversation about choosing between a technical AI newsletter and The Multiplied Self." width="1600" height="900" />
<figcaption>A chat with Claude about choosing between a technical AI newsletter and The Multiplied Self.</figcaption>
</figure>
<p>I paused.</p>
<p>I'd been thinking about reach, search traffic, and the most sensible way to grow. But imagining those two audiences made me realize I cared more about the conversations I wanted to have than the numbers I wanted to reach.</p>

<h2>An old idea, a new use for AI</h2>
<p>This reminded me of Socrates, the ancient Greek philosopher.</p>
<p>His mother, Phaenarete, was a midwife. Socrates used that idea as a metaphor for his own approach to philosophy: rather than simply giving people answers, he asked questions that helped people examine their beliefs and arrive at their own understanding.</p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/7ba7eeb7-e637-4d14-8935-84ce3b8f1274/image.png?t=1791622199" alt="An illustration of Socrates questioning a small circle of listeners in an olive grove." width="1600" height="900" />
<figcaption>Socrates questioning those around him. Illustration generated with AI.</figcaption>
</figure>
<p>AI can play a similar role. Instead of giving you answers, it can ask questions that help you examine your assumptions and work out what matters to you.</p>

<h2>What research tells us about asking questions</h2>
<p>In a 2023 study published in Judgment and Decision Making, 265 participants completed a planning task. Those prompted to reflect using systematic, Socratic questions discovered more far-sighted decision strategies faster than comparison groups.</p>
<p>The study didn't test AI interviews or major life decisions. But it suggests that structured questions about how we make decisions can help people discover better strategies.</p>

<h2>Try this: ask AI to interview you</h2>
<p>Think of a decision you're struggling with. It could be a career move, a personal project, or something smaller that you can't quite settle.</p>
<h3>Try this prompt</h3>
<div class="prompt">
<p>I'm trying to decide [describe your decision].</p>
<p>Don't give me advice or recommend an option yet. Interview me, asking one question at a time.</p>
<p>Challenge vague answers. Point out contradictions. Help me separate facts from assumptions. Ask what I'm optimizing for, what I'm afraid of losing, and what I might be taking for granted.</p>
<p>Don't flatter me or push me toward a particular outcome. Don't assume there's one hidden answer waiting to be uncovered. If my priorities conflict or I don't have enough information, help me recognize that.</p>
<p>Keep asking questions until I can explain my priorities clearly. Then summarize my priorities, assumptions, unresolved questions, and tradeoffs. Ask me whether that summary reflects my thinking accurately.</p>
<p>Don't make the decision for me.</p>
</div>
<p>You don't need a special AI tool for this. Any chatbot that can sustain a conversation is a reasonable place to start.</p>
<p>If you feel stuck, try saying:</p>
<div class="prompt">
<p>Push harder on that.</p>
<p>Ask me the question I'm avoiding.</p>
</div>

<h2>Compare the answers</h2>
<p>Try comparing the two approaches yourself.</p>
<p>First, ask AI what you should do about your decision. Save its answer.</p>
<p>Then start a fresh conversation and use the interview prompt above. Answer honestly, even when you don't have a clear answer. Let the conversation unfold until you have a better understanding of your priorities and tradeoffs.</p>
<p>Now compare the two.</p>
<p>Did the first answer assume something about what mattered to you? Did the questions uncover a consideration you hadn't thought about? Did you discover a conflict between two things you wanted?</p>
<p>Or did the interview simply confirm what you already knew?</p>
<p>AI can misunderstand you, too. A question might sound insightful and still be based on a wrong assumption. Think it through before accepting where the conversation leads.</p>

<h2>Don't hand over the decision</h2>
<p>Using AI to think is not the same as letting AI think for you.</p>
<p>For decisions that matter, verify the facts, question the assumptions, and make the final call yourself.</p>
<p>I needed it to slow me down long enough to hear myself think.</p>
<p>If you try this with a decision of your own, reply and tell me: what changed between the two conversations?</p>
<p>Let's multiply ourselves!</p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

    <item>
      <title>The Screen Free Rebellion</title>
      <link>https://themultipliedself.com/editions/the-screen-free-rebellion</link>
      <guid isPermaLink="true">https://themultipliedself.com/editions/the-screen-free-rebellion</guid>
      <pubDate>Tue, 06 Oct 2026 09:00:00 +0000</pubDate>
      <dc:creator>The Multiplied Self</dc:creator>
      <category>Learning</category>
      <category>Walking</category>
      <category>Experiments</category>
      <description>Stop staring at long PDFs. Use Gemini Notebook (formerly NotebookLM) to listen, question and explore information while you walk.</description>
      <content:encoded><![CDATA[
<p><em>Edition 01 — 6 October 2026</em></p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/a561aeaf-49cd-43ac-a3ab-7b2600e8327d/The_Screen_Free_Rebellion_Thumbnail.jpeg?t=1790859021" alt="The cover image for The Screen Free Rebellion — learning without staring at another screen." width="1200" height="675" />
</figure>
<p>At some point last week, I was staring at a PDF I genuinely wanted to read.</p>
<p>It was interesting. It was useful. It was also about 60 pages long.</p>
<p>I'd already spent most of the day looking at screens, so I closed it. Not because I didn't want to know what was inside. I just didn't want to read it.</p>
<p>That got me thinking:</p>
<blockquote>What if I gave the document to AI and let it explain it to me while I was doing something else?</blockquote>
<p>So I uploaded the PDF to Gemini Notebook (formerly NotebookLM), generated an Audio Overview, put on my headphones and went for a walk.</p>
<p>It turned the PDF into what sounded like a podcast, with two voices walking through the author's ideas and examples.</p>
<p>Then I noticed a button that let me join the conversation.</p>
<p>I could interrupt them. Ask a question. Get an answer. Ask a follow-up.</p>
<p>That was the moment it clicked.</p>
<p>It felt like a podcast organized around me.</p>
<p>I didn't have to wait for it to reach my question, or stop to find the paragraph in the PDF. I could just ask, and keep walking.</p>
<p>This is what I'm exploring with The Multiplied Self: what becomes possible when AI changes not just what we do, but how we experience everyday things like learning.</p>

<h2>The Walkman changed where music could happen</h2>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/5859515c-7347-41c3-a573-8d7dac5fa99c/Original_1979_Sony_Walkman.jpg?t=1790784978" alt="An original 1979 Sony Walkman TPS-L2 with orange foam headphones." width="1600" height="900" />
<figcaption>Original 1979 Sony Walkman TPS-L2</figcaption>
</figure>
<p>When Sony introduced the Walkman in 1979, it didn't invent music. It changed where music could happen.</p>
<p>The music hadn't changed. Your relationship with it had.</p>
<p>I think something similar is happening with information. For years, we've treated reading as something that requires a screen. But my 60-page PDF hadn't changed. I had simply found a different way to spend time with it.</p>
<p>The information hadn't changed. The way I met it had.</p>
<p>The screen was no longer where the learning had to happen. It had become where the learning started.</p>

<h2>A little research behind the walking part</h2>
<p>I wondered whether walking was changing how I thought, or whether I just enjoyed being away from my desk.</p>
<p>There's an interesting experiment from Stanford that gave me something to think about. Researchers Marily Oppezzo and Daniel Schwartz found that people generated more creative ideas while walking than while sitting.</p>
<p>That doesn't mean walking makes you understand everything. The research was about creative thinking, not comprehension. But it made me wonder whether I'd been thinking about learning too narrowly.</p>
<p>By then I'd noticed a pattern in what I was doing. I started calling it LISTEN.</p>

<h2>LISTEN</h2>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/babaa4bb-0134-412b-a87a-8585826b38cc/LISTEN_Framework.jpg?t=1790857704" alt="The LISTEN framework diagram: Load, Instruct, Spar, Track, Engage, Notice." width="1600" height="900" />
<figcaption>LISTEN Framework</figcaption>
</figure>
<p>Pick one document you've been putting off, then try it.</p>
<h3>L — Load</h3>
<p>Give Gemini Notebook material you genuinely want to understand: a PDF, article, research paper or your own notes. Don't worry about the perfect prompt.</p>
<p>One privacy rule: don't put passwords, financial credentials, medical records or other highly sensitive information into an AI notebook. Give it material you're comfortable putting into the tool.</p>
<h3>I — Instruct</h3>
<p>Ask for an Audio Overview. To take part in the conversation, create it in Interactive Mode, then tap Join when you want to ask something. Interactive Mode is currently available in English and works with newly generated Audio Overviews.</p>
<figure>
<img src="https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/1aaec508-cbef-4b99-a1e3-0b1c0f86387b/Gemini_Notebook_App_Screenshot.jpg?t=1790858035" alt="A screenshot of the Gemini Notebook app with an Audio Overview ready to play." width="1600" height="900" />
<figcaption>Gemini Notebook</figcaption>
</figure>
<p>Then put on your headphones and go for a walk.</p>
<h3>S — Spar</h3>
<p>Don't just listen. Talk back. Challenge something. Follow whatever catches your attention. Try:</p>
<ul>
<li>&quot;What does that actually mean?&quot;</li>
<li>&quot;Can you give me a real-world example?&quot;</li>
<li>&quot;What is the strongest argument against this?&quot;</li>
</ul>
<h3>T — Track</h3>
<p>The conversation itself doesn't stay with you, so back at your computer, capture the ideas you want to keep. In Gemini Notebook you can use Add note, or keep your own notes somewhere else.</p>
<p>The walk was for exploring. The notes were for keeping what mattered.</p>
<p>Then use your notes to reflect:</p>
<ul>
<li>&quot;What have I learned so far?&quot;</li>
<li>&quot;What are the three things I should remember?&quot;</li>
</ul>
<h3>E — Engage</h3>
<p>If one source leaves you curious, add another: a report, a paper, an article, your own notes. Then ask across all of them:</p>
<ul>
<li>&quot;Where do these sources agree?&quot;</li>
<li>&quot;Where do they disagree?&quot;</li>
<li>&quot;What questions come up when you look at them together?&quot;</li>
</ul>
<p>You don't have to read everything separately before you can start making connections.</p>
<h3>N — Notice</h3>
<p>Finally, stop and ask yourself:</p>
<ul>
<li>&quot;What do I understand now that I didn't before?&quot;</li>
<li>&quot;What still feels unclear?&quot;</li>
<li>&quot;What do I want to explore next?&quot;</li>
</ul>
<p>That's the part that matters. The experiment isn't successful because you listened to 60 pages without reading them. It's successful if you came away understanding something you didn't before.</p>
<p>You can do the whole thing with one document, or add a few sources and see what happens when you start asking questions across them.</p>

<h2>The bigger idea</h2>
<p>At first, I thought I was solving a very small problem:</p>
<blockquote>&quot;I don't want to read this right now.&quot;</blockquote>
<p>But I ended up with something more interesting. I was moving from &quot;I should read this&quot; to &quot;I understand this.&quot; And I didn't need another long stretch in front of a screen to get there.</p>
<p>The screen doesn't have to be the destination. It can just be the starting point.</p>
<p>Now I'm wondering:</p>
<blockquote>What else in your life could you stop doing the default way?</blockquote>
<p>Maybe it starts with one document you've been putting off. Give it the material. Put on your headphones. Ask a question. See what happens when you talk your way into understanding something instead of staring your way into it.</p>
<p>Let's multiply ourselves!</p>
<p>P.S. If you feel like trying one more version of this experiment, put this article into Gemini Notebook and LISTEN to it instead of reading it. See what happens when you take the conversation for a walk.</p>
<p><a href="https://the-multiplied-self.beehiiv.com/subscribe">Subscribe to the letter</a></p>
      ]]></content:encoded>
    </item>

  </channel>
</rss>