What I Actually Want You to Leave With

One page. Pin it up. Argue with it later.

How to Use This

This is the semester distilled into the handful of heuristics I most want you to carry out the door. It's not a summary of the readings and it's not a checklist. It's the short list of things I'd say to you in office hours if you had five minutes and I had to pick. When one of them starts to sound obvious, that's usually a sign you're ready to put real weight on it.

How to Think

AI is not replacing jobs. It's replacing bad thinking, and it's augmenting good thinking. The people who lose to AI are the ones whose whole contribution was something a model can do in a second. Your job is to be the person in the loop who's worth keeping in the loop, because when you bring real judgment the model multiplies it.

This class is about judgment, not tools. Tools change every six months. The habits of mind that let you pick the right problem, read a room, and call your own ideas wrong will still be working in twenty years.

Update the mental model, don't memorize the rules. Everything I'm handing you is a lens, not an instruction manual. The goal is a better model of how the world actually works, so the next time you meet a problem you've never seen, you can reason about it instead of pattern-matching on a slide.

Seek the argument against your own idea before you commit. Put conflicting opinions in the room on purpose. Lincoln built a cabinet out of his rivals for a reason, and you can build one inside your own head with four or five lenses (money, growth, pragmatic, the customer, the skeptic).

Run a premortem before the work, not a postmortem after it. Sit down and pretend it's two months from now and the project has failed. Write down every reason why. Most of the reasons you'll write are things you could fix today for free.

How to See

Get very, very real about the competition. Most people do a warm, flattering read of their own idea and a cold, suspicious read of the competition. Flip it. Red team your own work the way you'd red team an enemy's.

Every failure you can name is sitting inside a bigger system. The fishing simulation wasn't contrived; it's how things actually break at work, between teams, and between companies. If you can only see one team's incentives, you can't see the problem yet.

The first step to a solution customers like is asking the right problem. Most bad products are answers to the wrong question. Spend disproportionate time on problem selection; the rest of the sprint is cheap compared to getting this wrong.

A shippable product every step of the way beats a perfect product at the end. Always have something you could hand to someone. Fast, incremental value lowers the chance of a failure you can't recover from, and it gives you something real to get feedback on.

Don't run a cargo cult. If you're following the ritual (the framework, the ceremony, the process) without understanding why it worked the first time, you're building a bamboo airplane. The method is only as good as your model of what it's actually doing.

How to Reach People

Find a real problem for a real customer, and go stand next to that customer. I built the bus app because I hated missing the bus. When I needed to price it, I stood outside UVU and asked people getting off the bus. That's the whole method. Nobody is going to tell you what they'll pay in a survey as honestly as they'll tell you on the sidewalk.

Differentiation is the thing that makes it click. Your product doesn't have to be better at everything; it has to be different in a way the customer actually cares about. The most successful products change how the customer sees the world a little.

Take your advantage seriously. What's something you can do that few other people could match? Usually it sits at the intersection of two or three things you happen to know at the same time. That intersection is where your unfair advantage lives.

Write the founding hypothesis in one sentence. If we help [customer] solve [problem] with [approach], they'll pick us over [competitor] because our differentiation is [this]. If you can't say it that cleanly, you don't understand it yet.

Communication is a superpower, not a soft skill. Being able to stand in front of two senior VPs and tell them why their idea won't work, in a way that lands, is worth more than another technical certification. Practice this on purpose.

How to Build

Unique value lives at intersections. If you can do one thing really well, so can other people, and so can machines. The magic happens when you sit at the meeting point of two valuable things, because the combination is unique and hard to copy.

Codifiable work gets automated; judgment and context don't. Gary Kasparov lost to Deep Blue alone, but Kasparov with a computer beats a computer alone. Your job is to be the human in that pair, not the piece the machine replaces.

When your core competency gets outsourced, automation is next. That used to happen to low-skill jobs; now it happens to white-collar work. If the thing you're best at can be written down cleanly enough to hand off, it can be written down cleanly enough to automate.

This isn't chess. It's messier. Chess is a closed world with clear rules, which is exactly why computers won it. Your work lives in the open world, where connecting ideas across contexts is the whole game. Train for range, not for one narrow trench.

Treat AI like a coworker you're responsible for. Involve it in everything you try. Tell it what persona to take. Stay the human in the loop. And assume that whatever model you're using today is the worst one you'll ever use again.

You don't get credit for using AI. You get credit for shipping value with it. Roughly a quarter of companies have figured out how to actually make money from generative AI. Be one of the people who makes that number move, not one of the people who talks about the number.

Tools change. How you think doesn't. Learn to pick the right problem, stand next to the customer, and keep shipping. The rest is details.