Start here
When should I not use Discover Agents?
Why it works this way
Discover Agents is for ideas that would cost you real time or money to get wrong, so if being wrong would cost you a weekend, build it instead.
Reading the web takes minutes. Checking an idea takes days or weeks of conversations with real people. That time is worth spending when a wrong build would cost you real time or money. When it would cost a weekend, the build is the test, and running tests first is the slower way to find out.
What we gave up: the pull to say yes to everyone. Some people will read this page and close the product, and that is a fine result.
What it is
Discover Agents is for ideas that would cost you real time or money to get wrong. It is not for every idea. This guide lists the cases where the honest answer is to skip it, so you can leave early and keep your time.
How to use it
Read the five questions below before you start a board. A yes to any of them means Discover Agents may be the wrong tool for this idea.
- Would being wrong cost you a weekend or less? Build it and see. A rough build is the cheaper test, and the checking here takes days or weeks.
- Is there nobody you can talk to who has this problem? Most tests run on real people, and what simulated people, who are AI, say never counts toward a test. The few tests that need no customer can rule an idea out and can never confirm one. The product lists where your people gather and writes the message to send, but you do the asking.
- Do you want a yes more than an answer? Discover Agents starts from the assumption that someone is already working on your idea, shows the honest case against building, and calls no result proof. See What if the answer is no?.
- Do you want the agent built for you? It writes the tests. The building is yours. See Why does Discover Agents stop at the test?.
- Are you weighing something that is not an AI agent? The Blueprint, its five questions and the shapes are all written for AI agents.
If you can answer no to all five, a board is worth opening. Start with What can I type into the first box?.
What it will never do
- Build your agent for you. If you want it built, the building is yours.
- Tell you an idea is proven. A result is a signal from one test.
- Count a simulated person as evidence. With nobody to talk to yet, simulated people let you practise your questions and nothing more.
If something looks off
- You already started a board and now think you did not need it. Stop whenever you like. A board saves as you work, and Remove from list keeps it. See Does it save, and how do I come back?.
- You cannot tell whether building is cheaper. If you can put a rough version in front of people this weekend, do that. If you cannot say what being wrong would cost, that is a reason to look first.
- You have no one to talk to yet. You can still read the research and plan the tests. See How do I find people to talk to?.
- Your idea is not an agent. The Blueprint, its five questions and the shapes are about agents, and they will not fit.