Field notes

Most advice tells you how to feel about quitting. Write a threshold instead: a profit floor, a runway floor, and a date you check them. Research backs staying.

Kill an idea when a must-pass gate clearly fails, or when even the optimistic end of its range is bad. The five gates and two rules our audit engine actually runs.

'AI wrapper' isn't an insult — it's an unpriced risk. A 4-question test to tell whether the next model update makes your product better or redundant.

Vibe coding collapsed the cost of building — not the cost of building the wrong thing. Idea validation is now the whole game. How to do it in an afternoon.

How to price your product right with the Van Westendorp Price Sensitivity Meter — the four questions, running the survey, reading the graph, and its limits.

Sycophancy is structural, not a prompt bug. Why a chatbot can't tell you to kill your startup idea — and what real validation needs that a chat window lacks.

Mostly yes — but idea theft is the wrong worry. The real risk is the answer you get back. What happens to what you type, and what to do instead.

Saturation isn't a competitor count — it's demand vs supply, and demand is public. A 5-step method to read a market off Google, no tooling required.