The Promise of AI (Part III) - Defanging friction and just doing things

Friction shapes the world a lot more than policy

— Tobi Lutke (This Week in Startups podcast March 2021)

There are many more great business ideas in the world than there are businesses that turn those ideas into tools and products. "That could have been me" is what you might hear some people say. That happens at least partly because it can be too overwhelming to get started. The friction of getting started is what stops many people in their tracks.

I bought the domain jeremycowcher.com five years ago and had done nothing with it until a few weeks ago. I can give plenty of plausible reasons why I never acted and yet I suspect that I was actually overwhelmed and put it in the "too hard" basket. The friction was too great for me. Since you're reading this post on a website with that domain, you can see that five years of inaction was remedied. It was actually remedied by AI. I write these posts in a TextEdit file on my Mac and Claude Code does the rest.

In Part I of this series, I wrote about how AI is giving the best founders, engineers and designers more leverage and ability to create greater value. In Part II, I outlined why I expect to see a wave of amazing companies led by designer founders, unlocked by AI.

For this Part III, I'm focusing on how everyone can do things. As Canva gave everyone the ability to design, AI will eventually give everyone the ability to build things. Most of these things will never become products, maybe not even tools. That's the point. By democratizing the ability to build things, AI will mean there will be more things that are built that are not valuable, and potentially don't even spread beyond the initial builder. There will also be more things that turn out to be incredibly valuable.

Doing things is infectious. I got my start building things with AI after succumbing to the rave reviews on X for Claude Code and listening to Chris Hutchins on the "All the Hacks" podcast talk about using Claude to build different projects for him. Having done some Python courses, I felt comfortable enough in VS Code to get started with Claude Code. The truth is that the courses weren't necessary — anyone can get started with Claude Code. In fact, Claude Code has gotten me started on two other projects (Elementary Timer and IdeaKache), which I'll cover in Part IV.

Think of creators like Bill Simmons, who has achieved incredible success, and generated jobs for many people, because of the power of the internet. As the internet turned Simmons' writing into a global media business, AI can turn anyone's idea into something real.

As one small example of how AI will reduce the friction of getting started, take the beginning of DoorDash. DoorDash has been in the news because of viral AI prediction articles, but it started with a simple landing page and personal phone number for restaurant delivery. While the DoorDash founding team had incredible technical talent, the landing page was simple. Before AI, if you didn't have any technical talent, the friction of building a simple landing page would stop many people. Thanks to AI, everyone can now build a simple landing page. As Ben Thompson noted, the hard part of building the three-sided market for DoorDash would remain. That first step, however, is available to everyone now.

AI is not perfect. I worked with Claude to build a financial model for a public company last week and was underwhelmed. I was so underwhelmed I ultimately landed on not trusting it because of data issues. I suspect I'll feel differently in the near future. So while AI is not perfect, it's definitely good enough to defang the friction of getting started.

And all the really successful people I know have a very strong action bias. They just do things. The easiest way to figure out if something is viable or not is by doing it. At least do the first step, and the second step, and the third, and then decide

— Naval Ravikant (source)