I'm a product leader who restarted his career twice and is now reinventing it a third time — voluntarily — to build with AI from the ground up.
Majored in Civil Engineering. My career began in China, in sales. I worked my way up to BD Manager, and from there into running a new B2C initiative in Hangzhou — ten direct reports, a real P&L, and a clean wind-down twelve months in. Then I moved to the US, where my resume didn't translate, and I climbed the product ladder from scratch: Product Analyst → Senior PM → Lead PM → Group PM → Head of Product / Senior Director
Now, in addition to playing the role of people manager, I spend a big chunk of my time hands-on building 0->1 products: defining agent context — system prompts, personas, jobs-to-be-done — and writing code and building evals and shipping deploy scripts. The line between those jobs has collapsed, and the people who noticed that early will define the next ten years of products.
A short, opinionated list. I update it when something changes my mind.
All-In Podcast — used to be my Sat listen. The political content has crowded out the technical takes I went there for.
The clearest single intro to thinking in models.
First-principles thinking with the receipts.
A working reference, not a one-time read.
Required reading for anyone shipping decisions to humans.
The product-PM bible on habit loops.
Read it once for sales; read it twice for product; read it three times for self-defense.
Taught me to look for the incentive structure hiding underneath any behavior.
A useful counterweight to the assumption that you can read people — or users — as easily as you think.
A rotating list. Some of these will look obvious in two years; some will look stupid.
Sales coordinators, junior project managers, parts of account management — these roles will compress first, hardest. The work doesn't disappear; the headcount does.
If you can measure quality on day one, you can iterate faster than anyone who can't. This is more important than which model you use.
It's overrated as a credential and underrated as a forced restart button. Mine was the latter.
The tools I reach for every week. Dated so you can see the drift.
Read →about / mental modelsFive models I run decisions through before I even open Claude.
Read →about / problemsThree bets I want to test with people who disagree.
Read →