wtf is cold-review?

Cold Review is a prompt, not a piece of software, meant to be pasted into an AI chat tool. It gives project ideas a second opinion evaluation, examining evidence, checking competitors when the AI can browse the web, questioning assumptions that are not backed by data, and finishing with one of four decisions: Kill, Pivot, Test, or Build. To use it, you open the file called PROMPT.md, copy its full text, and paste it into an AI system such as ChatGPT or Claude. Then you add a clear description of your own project underneath the pasted prompt. If your AI tool supports web browsing, turning it on helps the review compare your idea against real competitors. Afterward, you are expected to independently check any important claims or sources the AI cites, since it cannot guarantee accuracy on its own. For a stronger review, the project recommends including a long list of details about your idea: what it is, what stage it is at, who the target user is, what problem it solves, how people currently work around that problem, why someone would switch to your solution, known competitors, your pricing or business model, how you plan to reach users, and any real evidence you have such as usage numbers, retention, feedback, or past test results. If you do not know something, the instructions say to write unknown or none rather than making something up. The project is upfront about its limits. It has not been tested formally across every AI model, and its output quality depends on which model you use, whether browsing is turned on, the quality of sources found, and how much detail you provide. Any similarity scores it produces are described as estimates, not facts. It also states clearly that it does not replace professional legal, financial, medical, safety, privacy, or security advice. The project is released under the MIT License, and its current version is 1.1.2 beta.