I’m launching a YouTube channel and podcast, Hidden Track, which is all about the side interests of people otherwise famous for something else. Check out the first episode, an interview with former Meta CTO Mike Schroepfer, about cow burps, offshore datacenter pirate risk, and other ways in which green tech can be made cheap.

Please share with your friends and subscribe. Thank you so much for supporting me!

Just because you can doesn’t mean you should. I keep relearning this lesson.

There was a time when I took an engine and transmission out of a Lamborghini because I was too cheap to pay the dealership for a clutch change. On a Honda, you simply jack up the car, detach the engine and transmission from below, and insert your $300 clutch. For Lamborghinis, where apparently the cost of labor doesn’t matter, the clutch itself is $2,200 and the engine-out job entails an additional $9,000 of labor.

It took over a month of nights and weekends for me to remove this engine and transmission, with a ton of help from my wife logging the exact order of every wire and hose we detached (so as to make the process reversible), complete with taped labels on both ends of anything detached so we could later pair things back together. When we finally reattached everything and attempted to start the engine, it wouldn’t catch for 30 seconds but instead just cranked and cranked. During that time, I gave serious consideration to the very real possibility that I had ruined everything, creating a much more costly repair through my DIY “frugality.”

Turns out a reattached fuel hose takes a few dozen seconds to siphon enough gasoline back toward the pistons before an engine will restart. After an extended period of suspenseful dread, the engine finally started to life, and I could at last congratulate myself on saving $9,000 and Sticking It to The Man™.

I don’t think I’d ever do this again even if I still owned the car. The process itself was invaluable, in that I learned a boatload about car repair by tackling a comically ambitious task, but the fully amortized cost of my time, coupled with the possibility I could have made the world’s most expensive doorstop by committing one of many potential mistakes along the way, make the experience one I wouldn’t repeat.

Just because you can doesn’t mean you should.

False Victory Laps for Solopreneurship

With the rapidly increasing capabilities of AI, there’s been a gushing of enthusiasm around the potential for everyone to start their own companies. In fact, July 2026 saw the most companies created in any month in US history (577k).

There’s no question you can. Now’s a better time than any, arguably, for starting your own company. AI works tirelessly, and can also tackle things you might not be good at. It’s cheaper than ever to hang a shingle. The question is whether you should.

Based on having worked on Superphonic full-time the past 18 months, here are some drawbacks of running things yourself:

- It can be lonely. I had the good fortune of working with Guilherme Scotti for part of this time… but I felt lonely during the period where I didn’t have a coworker. Perhaps you’re the coding equivalent of Kelly Clarkson, never lonely when you’re alone. But most people appreciate having coworkers.

- Ideas are made better through partnership. In recent months, I’ve been brainstorming a bunch with the first PM in my career, from Microsoft Word back in 1997, Ashok Kuppusamy. Batting ideas around with Ashok has greatly improved both my happiness and the quality of my ideas. I’m hoping to announce more things from this partnership in coming months.

- The App Store is nearly impossible at this point. Although Superphonic is notably better than all other podcast players (e.g. it skips ads, it supports up to 5x playback speeds, it’s the only player that supports subscribing to topics, it auto-chapterizes so you can skip content quickly), discovery is impossible when the App Store has over 2M apps. Imagine my disappointment when I heard from an Apple employee — someone who created many of the apps you use every day — that even he finds the App Store to be a highly unlikely path to building a business these days.

- AI erodes moats. Relatedly, whatever you can build can copied just as easily by anyone else. Especially if they wait six months before copying you, at which point the AI is even better. The entire premise of The Hard Thing About Hard Things is the insight that durable businesses are built on hard things. Easy things, by definition, are easily mimicked. The fact that AI makes building The Next iFart™ effortless is an observation completely neutral to the prospects of creating successful businesses.

What Next?

Advances in AI have been both incredibly exciting, opening vistas heretofore seemingly impossible in what one person can build, as well as disillusioning, when I realize that whatever’s trivially easy for me to do is equally trivial for others. The main insight I’ve had, if anything, is that I miss working with more people.

Just because AI allows you to build everything yourself doesn’t mean you should.

This latest realization has made me consider a lot more seriously whether I should deliberately join a larger team, or formalize my work with Ashok and others into more structured collaborations instead of ad-hoc brainstorming.

No conclusions yet. LMK what you’ve been thinking.