The Advantage #08: The Machine Learns Your Voice, the iPhone Moment Repeats, and the Newsletter Writes Itself

Thanks for tuning into the eighth episode of The Advantage. A short, weekly note where I share what I am working on, something worth watching, a lesson from history, and one practical edge you can try right away.

If you know someone who may enjoy this, please feel free to forward it.

If you were forwarded this email and are enjoying it, you can subscribe here.

I just finished writing a deep dive on why the biggest career risk in AI is not the machine itself. It is the person in your industry who figured out how to manage it before you did. This is not theory. It is backed by field studies from UC Berkeley, Harvard, BCG, and MIT, and it draws on a pattern I have watched repeat across every major disruption I have invested in. Here is an excerpt from the full read:

AI is not a tool you pick up when you need it. It is an employee. A very smart, PhD-level, ambitious, eager to please employee who never gets tired and almost never tells you "I don't know." That sounds like a superpower, and it is. But it is also the trap.

AI output is often wrong in quiet ways. It sounds right. It uses the right tone and structure. But it can be missing key constraints, built on invented facts, or structurally flawed in ways that only someone with real expertise would catch. When that slips through, it is not an AI failure. It is a management failure.

I have experienced this firsthand. Like most people, I started using AI in basic ways. My thinking shifted when I stopped treating it like a search engine and started treating it like part of my org chart. I now use AI as an analyst, a model builder, and an editor. It builds my Excel models, drafts my memos, and pressure tests my thinking. But the 80% it gets you on the first pass is not where the value lives. The value is in the last 20%, which requires serious back and forth.

Quick intro:

This week's Worth Watching is a 26-minute YouTube video called "How To Use Claude Co-Work Better Than 99% of People" by Michele Torti. If you have been hearing about AI tools but are not sure where to start, or if you have been dabbling but feel like you are barely scratching the surface, this is the video. Torti does not just walk you through features. He builds a complete automation system from scratch, on camera, in real time. It is one of the clearest demonstrations I have seen of what AI-assisted work actually looks like when done right.

What I loved about it:

The central argument hit me hard: 99% of people use Claude like a chatbot. They type a question, get an answer, and move on. The top 1% treat it like an autonomous coworker. The difference is not intelligence or technical ability. It is architecture. The people getting 10x results are the ones who build systems around the tool instead of just talking to it.

Torti introduces a concept he calls the README as the "brain." Instead of prompting Claude fresh every time, you create a structured document that tells Claude who you are, what you are building, and how you want things done. Claude reads it at the start of every session and works from that foundation. It is the difference between hiring a new temp every day and onboarding a full-time employee who learns your preferences over time.

What makes this click is the live demo. He builds what he calls "Repurpose-OS," a system that takes a single YouTube video and automatically generates a blog post, social media content, and newsletter copy, all formatted and ready to publish. Watching him do it in real time made the abstract idea of "AI productivity" feel concrete and immediate.

Here is my 20-second recap if you do not have the full 26 minutes:

Stop prompting, start programming. Most people treat AI like a search bar. The leverage comes from building repeatable systems: structured folders, clear instructions, and a README file that acts as the AI's operating manual. One-off prompts give you one-off results. Systems give you compounding returns.

The README is the boss. A well-written README file becomes the AI's memory and decision-making framework. It stores your voice, your preferences, your project structure, and your quality standards. Every session starts from a foundation instead of a blank slate. This is where the real productivity multiplier lives.

Separation of concerns matters. Just like good software architecture, good AI workflows separate different types of work into distinct modules. Research goes in one place. Writing goes in another. Formatting goes in a third. This makes each piece auditable, editable, and improvable without breaking the whole system.

The bar is still low. The vast majority of people are not doing any of this. They are copy-pasting into a chat window and calling it "using AI." If you invest even a few hours into building a proper system, you are already ahead of 99% of users. The gap between casual use and systematic use is enormous, and it is still wide open.

Pro Move: You do not need to watch it. Put it on like a podcast and listen while you walk or drive.

What Happened:

On January 9, 2007, Steve Jobs walked onto the Macworld stage and introduced three products: a widescreen iPod with touch controls, a revolutionary mobile phone, and a breakthrough internet communications device. Then he paused. "Are you getting it? These are not three separate devices. This is one device." The crowd erupted. The iPhone was not the first smartphone. IBM's Simon existed in 1994. BlackBerry dominated enterprise. Nokia owned global market share. But none of them had built a product that non-technical people actually wanted to use. Within 74 days, Apple sold one million iPhones. By the end of 2008, the App Store had launched and third-party developers were building an ecosystem Apple never could have created alone. Nokia's market share, once north of 40%, collapsed to under 3% within six years. The technology existed before Jobs. The paradigm shift did not.

Insight behind it:

The iPhone did not win because it was the most powerful device. It won because it removed the barrier between capability and usability. Every prior smartphone required technical fluency. The iPhone made the power accessible to everyone. That is when adoption goes from linear to exponential. The breakthrough is never the technology itself. It is the moment when the technology becomes invisible and the experience becomes intuitive. When that happens, the market does not grow incrementally. It restructures entirely.

Modern application:

Claude Co-Work is following this exact pattern. AI tools have existed for years. ChatGPT, coding assistants, automation platforms. But most of them still require technical fluency to use well. You need to know how to prompt, how to structure workflows, how to debug outputs. Co-Work changes the interface. It gives non-technical users the ability to direct AI the way a manager directs a team: with natural language, clear objectives, and iterative feedback. That is the iPhone moment for AI productivity. The technology was already here. The usability just caught up. If you are waiting for AI to "get good enough" before you start, you are making the Nokia mistake. The shift is not coming. It already arrived. The question is whether you are building on top of it or watching from the sideline.

Why it works:

Full transparency: you are reading a newsletter that was assembled by Claude Co-Work. Not just drafted. Assembled. The research, the writing, the formatting, the section-by-section construction, all of it was produced by Claude working autonomously from a set of instructions I provided.

I am telling you this for a reason. This is not a gimmick. It is a proof of concept for the single most important productivity shift I have encountered since I started building companies. AI does not just help you think faster. When set up correctly, it executes entire workflows end to end, at a level of quality that would have taken me an entire weekend to produce manually.

Here is why this matters: most people are still using AI to write an email or summarize a document. That is sampling. Real adoption is when AI plugs into your actual workflow, with structure, review steps, and quality standards that keep the output reliable. Co-Work is the first tool I have used that makes that kind of adoption accessible without writing a single line of code.

How I use it:

Here is exactly what happened. I sat down and gave Claude a detailed prompt: the theme for this issue, the YouTube video to review, the blog post to reference, and the structure I wanted each section to follow. Claude then opened my browser, navigated to Gemini for the video transcript, accessed my custom GPTs in ChatGPT, downloaded and analyzed all seven of my past newsletters to learn my voice, and built a comprehensive style guide from scratch. Then it wrote every section, assembled them into a single document, and entered the final draft into Beehiiv. (This is the Newsletter CRM I use)

My total active involvement was roughly 15 minutes of upfront direction, plus a final review pass. The rest was autonomous. (It was actually close to 1 hour - 2 hours of total time, plus I had to do minor formatting at the end. However, this was the first setup.)

This is not magic. It is management. I treated Claude the same way I described in my blog post: not as a tool I pick up when I need it, but as an employee I onboard, direct, and hold to a standard. The 80% first pass is not where the value lives. The value is in the system that makes the first pass possible and the review process that sharpens the last 20%.

The challenge:

Try Claude Co-Work yourself this week. Pick one workflow you do repeatedly, something that takes 30 to 60 minutes. Write out the steps as if you were training a new hire. Then hand it to Claude and see what happens. You do not need to be technical. You need to be clear. Start with something low-stakes. A weekly report. A content outline. A research summary. The first time will feel clunky. That is normal. By the third time, you will wonder how you ever did it manually.

The biggest career risk in the AI era is not the machine. It is the human who learned to run the machine before you did.

Over the last several months, I've built multiple custom GPTs, trained Gemini, ChatGPT, and Claude across a range of tasks, and taught them to understand my voice and objectives. I've figured out which models do what best, integrated multiple systems across my AI workflows, and developed a clear point of view on my newsletter: its structure, tone, and quality bar. I've done every job in the process, which means I can explain each one in detail, and I know what great looks like.

The prompt used multiple LLMs, custom GPTs, sub-prompts that are very detailed, and many different tools to verify quality and tone. I intentionally didn't edit this newsletter. I wanted to demonstrate what Claude Cowork could do on its own. Without that constraint, I would have made several edits and structural changes that I believe would have delivered a meaningfully better product.

That said, my ability to land at this level of output is the result of spending 20 to 30 hours a week on AI for the last six months. The learnings, the model training, the understanding of capabilities: it's all starting to compound. Like any tool, becoming proficient requires real effort. There are no shortcuts.

I'm not saying any of this to brag. In fact, I feel massively behind. It's an exciting time, but I've mentally prepared myself: the next few years are going to be the grind of all grinds. As the old saying goes, hard life today, easy life tomorrow. Easy life today, hard life tomorrow.

Thanks for reading,

Join The Advantage, click here to subscribe to the newsletter.