Short answers to the questions I get most, written once so I can answer with a link. Steal what works.
A project is validated on the day a stranger pays for it. Applause from family and friends rounds to zero, and so does a waitlist. I hold everything I build to this line because payment is the one signal that survives contact with the market. Until it happens, the honest word for the project is hobby, and hobbies are fine as long as they are labeled.
Every side project gets three numbers before it gets hours: a weekly hour cap, a kill date, and a revenue bar. If the bar is missed by the date, the project dies and the hours return to the main thing. The contract is written in advance, while I still like the idea, because future me is a biased judge.
A new idea goes straight to the wall. I hunt for funded competitors, price the switching cost, and give the idea one day to die. A fast no is money in the bank. Whatever survives earns the cheapest possible test that puts it in front of a stranger's wallet, and the test runs the same week.
Producing is free now. Anyone can generate ten drafts, ten features and ten landing pages before lunch. Taste is the scarce resource: knowing which nine to kill. I grade my weeks on what got deleted.
I skip wireframes and build the product itself, even when it costs more time. A sketch can only teach you about the sketch. The real thing teaches you how people behave when something is at stake, and that lesson is the entire point of building early.
Before I build, I want a named channel where the first twenty customers already gather: a specific room, list or feed. If that answer is fuzzy, the idea is fuzzy, and it goes back to the wall.
My shipping unit is a single HTML file on Cloudflare Pages. Everything inline, deployed by drag and drop, live in under a minute. The small scope keeps projects honest and keeps the delete key cheap. This page is one of those files, and so is every other page on this site.
I keep my life on a timeline with switchable lenses: work, people, places. Zoomed out, years group themselves into eras, and the next era gets a dotted line so the future stays a draft instead of a blank. The tool for this lives at /timeline, built as one file like everything else.
All of it calibrated on my own mistakes. Adjust the numbers, keep the knife.