

The world completely changed on November 24, 2025 with the release of Claude Opus 4.5, specifically when used in Claude Code. Not in some fun imagine-the-future way, but in an immediate, jaw-dropping way.
I didn't use Claude Code until late December 2025, but when I did it felt like magic. Developers became 10x faster and 10x more people could now be developers. The knowledge required to produce software is no longer programming and reading docs. It's been reduced to clearly articulating what you want in English.
My friend Slobo and I started running events called Shiphaus to show people the magic. As Claude Code is not trivial to get set up, and we wanted to help our friends level up. After onboarding them, we found our suspicions were correct: nearly everyone felt the same way.
There's a start-up adage that "ideas are cheap, execution is expensive." I fully subscribed to it until I tried Opus 4.5. I think it's been completely inverted. Execution is now cheap, and it's worth dedicating more time to ideas, specifically ones emerging from the second-order effects of AI.
There's another start-up adage about second-order effects: "It was easy to predict mass car ownership but hard to predict Walmart." Cars get mass produced, so everyone gets one. Everyone has one, so highways get built. Highways exist, so suburbs sprawl. Suburbs sprawl, so big-box retail emerges. The chain from Cars → Walmart is obvious in retrospect. But whoever connected the dots first had a great advantage.
So there is big ideas to discover in the analogy: The car is to Walmart as AI is to ???. Cars collapsed the constraint of travel time. AI collapses the constraint of everything done on a computer. What are the second-order effects of that lifted constraint?
One constraint I've been thinking about is computer applications. Apps were a solution built around a problem: humans are messy, machines are rigid. Apps normalized our inputs into machine-readable data by forcing it to be done one way. UX design was the obsession of tech companies, all in an effort to coach users toward a desired action.
But AI flips this. Now AI can translate English for the machine. There are a thousand ways to say "schedule a meeting with John next Tuesday" and AI understands all of them. Humans don't need to learn to click a specific sequence of buttons to perform an action. We can just speak or type, a much preferred form factor to clicking.
The world completely changed on November 24, 2025 with the release of Claude Opus 4.5, specifically when used in Claude Code. Not in some fun imagine-the-future way, but in an immediate, jaw-dropping way.
I didn't use Claude Code until late December 2025, but when I did it felt like magic. Developers became 10x faster and 10x more people could now be developers. The knowledge required to produce software is no longer programming and reading docs. It's been reduced to clearly articulating what you want in English.
My friend Slobo and I started running events called Shiphaus to show people the magic. As Claude Code is not trivial to get set up, and we wanted to help our friends level up. After onboarding them, we found our suspicions were correct: nearly everyone felt the same way.
There's a start-up adage that "ideas are cheap, execution is expensive." I fully subscribed to it until I tried Opus 4.5. I think it's been completely inverted. Execution is now cheap, and it's worth dedicating more time to ideas, specifically ones emerging from the second-order effects of AI.
There's another start-up adage about second-order effects: "It was easy to predict mass car ownership but hard to predict Walmart." Cars get mass produced, so everyone gets one. Everyone has one, so highways get built. Highways exist, so suburbs sprawl. Suburbs sprawl, so big-box retail emerges. The chain from Cars → Walmart is obvious in retrospect. But whoever connected the dots first had a great advantage.
So there is big ideas to discover in the analogy: The car is to Walmart as AI is to ???. Cars collapsed the constraint of travel time. AI collapses the constraint of everything done on a computer. What are the second-order effects of that lifted constraint?
One constraint I've been thinking about is computer applications. Apps were a solution built around a problem: humans are messy, machines are rigid. Apps normalized our inputs into machine-readable data by forcing it to be done one way. UX design was the obsession of tech companies, all in an effort to coach users toward a desired action.
But AI flips this. Now AI can translate English for the machine. There are a thousand ways to say "schedule a meeting with John next Tuesday" and AI understands all of them. Humans don't need to learn to click a specific sequence of buttons to perform an action. We can just speak or type, a much preferred form factor to clicking.
My guess: AI exists, so everyone gets an assistant. The assistant does your computer work. If the assistant is doing the work, it's the one interacting with apps. And agents don't want human UIs, they want APIs.
So the future isn't better human apps, it's APIs built for agents that do the interaction on your behalf while you just talk to your assistant. I'm building for that world at Network School because my prediction is 95% of people in the US have an agent assistant by 2027.
My guess: AI exists, so everyone gets an assistant. The assistant does your computer work. If the assistant is doing the work, it's the one interacting with apps. And agents don't want human UIs, they want APIs.
So the future isn't better human apps, it's APIs built for agents that do the interaction on your behalf while you just talk to your assistant. I'm building for that world at Network School because my prediction is 95% of people in the US have an agent assistant by 2027.
No comments yet