Software, Bicycling, Brooklyn.
Fifteen years: Junior engineer to VP of Engineering. I still ship code every day. Engineering Leadership, AI, Product, Security. I studied poetry because it's elegant and packed with meaning. Then I designed software architecture for the same reasons. Now, I make the second by writing the first.
Thoughts
Where I stand on AI and softwareSoftware is fun again.
In the 1990s and early 2000s there weren’t many of us. The people writing software were the ones who were born to.
Then software became a big industry. It grew faster than it could stay fun. It became org charts, grooming meetings, and story points. We spent more time describing the work than doing it.
AI is unwinding that. I can make something good in an evening, and something great in a week. It feels the way it did back then, and I love it.
The factory is the wrong metaphorFactories make the same thing a million times. That’s not what we’re doing.
At a conference recently, I heard the phrase “software factory” at least thirty times. It’s the wrong metaphor. A factory makes identical products, plans everything before the line starts, and can’t change course once it’s running. None of that describes this work. But do we even need a metaphor? Aren’t we just building software?
When programmers fed punch cards into rented mainframes, it was called building software. When we dragged buttons onto forms in Visual Basic, it was called building software. When we write the rules, tools, and feedback loops for agents, it’s still called building software.
Software always suckedIt was built for the median user, and the median user is nobody.
Every app you’ve ever used was a compromise between what you needed and what a few million other people needed. We called it good software when the compromise was tolerable.
The home screen on my phone will soon be almost entirely custom icons: apps I had AI build for me, because each one is at least a little better than the one it replaced. Those custom apps are also saving me at least $50 a month in subscriptions I no longer pay.
Prepackaged software is fine for the things I don’t care much about. But for the things I do care about, like messaging people, my health, finding information, my money, taking notes, and a dozen others, almost all of the apps stink. I don’t want them anymore, and I don’t want their fees.
For the next few years, most people will still want an app that has already been decided for them. They don’t have the time or interest to customize one. That only lasts until apps get good at an automated feedback loop: using in-app analytics to reshape themselves in real time for the person holding them.
Those preferences compound, and they’re yours. Carry them from app to app, and each one already knows you the first time you open it. When I buy a new phone in five years and turn it on, I don’t want to see a single stock icon on that screen.
When everyone is a meat proxyA follow-up to Don't be a meat proxy, for whole organizations.
Niklas Gruhn's Don't be a meat proxy told us to stop pasting the model's answer into Slack. Fair.
Now look at a whole organization. Every white-collar person in it uses their own expertise to steer a model toward research, ideas, discoveries, and recommended decisions, then hands the result to the next person, who does the same. Most of what moves between us is an AI artifact with a human stamp on it.
So what are we? What should an organization look like now? I don't have an answer. I just carry the feeling around every day: that I'm stamping work I didn't quite do.
Projects
- Heyyy HealthAn LLM-powered health coach. Designed, built, launched, still maintained.
- Daily ScoopA personalized daily newsletter built from your RSS feeds, your newsletter subscriptions, and AI research on your interests.
- NYC Alternate Side ParkingAn app that tells you when to move the car.
- My Career CopilotCareer exploration for teens and veterans, through day-in-the-life videos.
Learnings
- BendHave a six-month forced-upgrade policy on your mobile app. Otherwise you always have to be backwards compatible with an infinite history of app releases, which makes engineering and product way harder.
- Forerunner10x engineers are real. I’ve worked with a handful over the years. If they are in meetings and code reviews all day instead of building, the team is composed wrong. I’d fix that first.
- Bowery FarmingThe age of robots has yet to arrive.
- Care/ofCross-org success starts with leaders, but only lasts if individuals are given the agency to solve problems together without those leaders around.
- NBCUniversalThirty microservices, each owned by a small pod, worked because the company was big. The overhead (project setup, CI, API versioning, etc.) added a lot of complexity that would be unbearable for a small or medium team.
- AdvomaticClients know what they want, but they rarely know what they need. Finding it for them feels like being a magician.
- Institute for Integrative NutritionCubicles are underrated. Focus++
- Poets & WritersIncorrect data leads to incorrect decisions. On the other hand, no data can at least leave room for reasonably good ones.