Musings

Startup paradoxes

Every list of startup advice contradicts the last one you read. Move fast and break things, but also be deliberate and measure twice. Stay highly focused, but experiment constantly and be ready to pivot. Hire slow and fire fast, but also scale quickly before the window closes. After a while you start to think the real lesson isn't in any individual piece of advice — it's that following any one of them too religiously is what kills you.

Compounding returns of AI workflows

Using Claude Code daily for as much of your work as possible creates compounding returns that are hard to see at first. Once your plans, preferences, past projects, and key numbers live in markdown files, every next task and every next project gets incrementally easier. That benefit compounds into a system where everything becomes dramatically more efficient. The longer you wait to start building that system, the more ground you're giving up.

Creative flow state and AI

AI has completely changed where the real work happens for me. I spend two hours on a walk with Wispr Flow, getting ideas down on my phone in a stream of consciousness — speaking at three times the speed I could ever type. The creative thinking and planning in the flow state is the hard bit now. The execution is the easy part once you just bypass permissions and let things run.

Age of the generalist

I've got slight cognitive bias here because this describes my background, but I genuinely believe this is the age of the generalist. The people I've always looked to hire are the ones with a logical brain, a habit of tinkering with tech tools, and the ability to operate across different domains such as product, engineering, and commercial work. A broad base of knowledge, combined with strong technical ability and intellectual curiosity, can turn someone into a 10x or 100x contributor — especially now that AI handles so much of the deep specialist work.

Software engineering in nine months

Boris Cherny, the creator of Claude Code, said recently that the title "software engineer" will start to disappear by the end of this year. I think he's right. At the highest levels, engineers will become architecture and code reviewers. But the rest of the job — writing the code, designing the product, speaking to users, iterating on feedback — that's becoming its own discipline, and it won't require a CS degree. Some people will do both. Most won't need to.

Small teams, good people

Right now, a small team of exceptional people with strong horizontal knowledge across many domains will outperform a larger, less AI-fluent team by orders of magnitude. The economics of headcount have fundamentally shifted and I think most companies are still staffing for the old world.

Startup Zen #1 — Shoshin

Shoshin — beginner's mind — is probably the most useful Zen concept I've come across as a founder. It's the practice of approaching something as if you're encountering it for the first time, no matter how much experience you have. The moment you think you've figured out and understood your customer is usually the moment you stop listening to them.

Startup Zen #2 — Kanso

Kanso is the Zen aesthetic of simplicity — not minimalism for its own sake, but the discipline of stripping away everything that isn't essential. Steve Jobs took this literally, saying no to hundreds of product ideas so Apple could focus on a few. Most product teams I've worked on could do with that kind of ruthlessness.

Startup Zen #3 — Mushin

Mushin — no-mind — is the Zen state of acting without overthinking. In martial arts it's what separates someone who fights from someone who thinks about fighting. I've noticed the same thing in founders: the best ones I've worked with don't deliberate endlessly over every decision. They build taste over years, then trust instinct in the moment.

Roles dissolving

The walls between product person and technical person are coming down fast. When a product manager can build a working prototype in fifteen minutes and an engineer can test product hypotheses at near-zero cost, the old job titles start to feel like artificial boundaries.