Writing.
Long-form pieces about decisions, constraints and evidence, and what changed when the first answer did not hold.
All pieces
- The question that makes the bookI started making picture books for my son with AI, and the first ones missed.
- My writing room has a critic that isn't allowed to fix anythingThree weeks after the first episode of my show locked its script, I cut a line about my own background. It was wrong, and it had already been through every check I'd built.
- The generic resume kept winningI used to tell people to tailor their resume to every job they applied for. Then a man I spoke with about his job search showed me a test he'd been running on himself.
- AI didn't break the job searchAI didn't make the job search worse. It made something obvious that was always true.
- My second hire only gets one idea at a timeThe second role on my one-person team is the research bench, and what defines it is a limit. Each time it runs, it can take at most one idea off the scout's list. It has to source every claim in that idea, say who published each number, and then make the strongest case it can against it.
- The call that decided itA hire I made a few years ago was decided on a phone call with someone she used to work with, not in any of the interviews.
- Rethinking the roleA recruiter tells you the team is "rethinking the role" halfway through a search, and it's easy to assume you did something wrong. When I've been on the hiring side, it was often a different question: could part of this job be done with AI instead?
- How a voice agent hears youIf your first-round screen is run by a voice agent, it doesn't listen the way a person does, and the way most of us answer interview questions works against us. I've spent the last few months building an assistant that makes phone calls for me, so I've listened to a lot of recordings of software trying to understand people.
- The cover letters all sounded like one personSomewhere in the middle of a stack of applications, I realized I was reading the same cover letter over and over, written by different people.
- My first hire isn't allowed to decide anythingEvery Monday I'm going to take one role from my one-person team and show you how it works: what it does, how it's set up, something it actually produced, what broke, and a template you can copy. I'm starting where I'd tell anyone to start, at the bottom of the chart.
- Walk your front door as a botYou can find out in an afternoon which parts of your process only work because a person is on the other end. You don't need to build anything.
- What's actually happening with AI agents at the front doorAmazon blocked Meta's new shopping agent, Muse, from its store less than two weeks after it launched. Amazon said apps that buy on a customer's behalf should "respect service provider decisions about whether or not to participate." Meta, meanwhile, named Walmart, Sephora, Expedia, OpenTable and Shopify among its partners.
- The agent will read your policy literallyLate checkout cost extra. It said so on the website and on the card in every room, and when I was a manager at a hotel, I waived it all the time.
- The anniversary an agent won't mentionAn AI agent can book a table for two at eight and never mention it's their anniversary. When I was a manager at a hotel restaurant, catching things like that was the real job, and none of it was in the policy.
- The subscriptions were the cheap partAn AI subscription is only the start of what it costs to make something. My show's plans are $264 a month. Testing and usage cost another $674.81 in the 30 days to September 25, and no pricing page shows you that number.
- The third try was a different ideaWhen an AI tool misses, the instinct is to explain more. Often, that's exactly what makes the next version worse.
- The video series only you could makeIf you've spent years doing a job, there's a video series only you could make. Until recently, making it took a team.
- I run a show with a team that doesn't exist. Here's the org chart.At the end of 2019 I had two production quotes for a show I was trying to start. The smaller shop wanted $1,400 an episode. The other, with three cameras, a sound engineer and lights, came to $2,955 to $3,000.
- Nobody is absent until somebody says soNobody is absent until somebody says so.
- A guess with good credentialsAnything you write down is true the day you write it. After that it's a guess with good credentials.
- One bad practiceI ran one bad practice and wrote down what went wrong. What came back had things in it I didn't know.
- The first thing I made with AI wasn't an appI can't stop an AI from being confident and wrong about my work. I can stop it from being wrong about the facts, and that took a list, not a prompt.
- I don't know enough about AI yetI hear this constantly, and I said a version of it myself for most of last year.
- When self-service is the better answerI wouldn't want a customer paying me just because I'm the only one with the login.
- Inside the writing partnershipI'm not convinced a form can tell me enough to write under someone else's name.
- The cost of selling the workI'm building a service that uses AI, and I need to be careful not to price it as though my own time is free.
- Who is doing the extra work?I might be giving my own AI tool too much credit.
- Before someone else depends on itIf I build a treehouse for my son, I don't need to work out how to sell treehouses.
- A backup plan is only theory until someone uses itI can have a backup of every file and still not know which one I'm supposed to send the client.
- A marketplace is not a customer listIf you're building up clients on a marketplace so you can work with them directly later, I'd check that second part now.
- The customer does not see your handoffsIf I have to approve every little thing, I can't really expect someone else to take care of the customer.
- Saving time is not the same as finishing the jobWe weren't getting the results we wanted, so we'd change the process. Then we'd change it again.
- The business behind the one-person companyA one-person company can look wonderfully simple on an org chart. The customer may still be depending on a chain of other companies to get one answer.
- The line that made the idea smallerThe longer I work on an idea, the easier it gets to explain why it deserves another month. The harder sentence is the one that says when it does not.
- “We can add that” is a cheap promiseI have been in enough product meetings to know that “we can add that” gets treated as if somebody solved something.
- Knowing how to build it was the problemOne of the stranger problems I had building an AI product was that I was pretty good at finding the next thing to build.
- The market is not the buyerAt Subziwalla, I could describe the market in one breath: Indian and South Asian grocery delivery. It sounded specific. It still did not tell me what to build next.
- The hard part starts after the demo worksI had a working production system and chose not to use it for a week.
- The name changed. Here is what it is now and how it's actually going.This newsletter used to be called Product of One, and it was about career strategy. It is now called The Operator Economy, and it is about building businesses that one experienced person can own and run.
- The Feature Nobody UsedI shipped a feature nobody used because I never watched anyone use the thing it was supposed to fix.
- Fix the sound of the oceanA guest at the hotel where I ran customer operations in Mexico once asked me to fix the sound of the ocean.
- Thorough Onboarding, Empty CompletionI built my product's onboarding to be thorough. Then I watched people quit it halfway through.
- Nice vs. kindThere's a difference between being nice and being kind, and I think about it constantly when I'm building.
- AI Usage Excellence FrameworkThe AI Usage Excellence Framework provides a structured pathway for product managers to evolve their AI collaboration from basic task automation to sophisticated cognitive partnership. This framework emphasizes intentional, progressive integration of AI tools to enhance—not replace—human judgment and creativity.
- Decision-Driven DevelopmentDecision-Driven Development (DDD) is a framework that transforms how product teams approach their work. Instead of focusing on activities and deliverables, DDD ensures that every piece of work directly enables a critical product decision.
- Prompting Is Overrated. Context Is Everything.Everyone's talking about prompting. There are courses on "prompt engineering." LinkedIn is full of "10 prompts that will change your life." People are building entire businesses around teaching you to talk to AI.
- Sprint Metrics & HealthThe Sprint Metrics & Health Framework provides a systematic approach to tracking sprint performance and team wellbeing. It combines quantitative metrics with qualitative health indicators to create an early warning system that helps teams maintain sustainable pace and predictable delivery.