A free idea validator is a lead magnet. Its job is to get you to sign up, so it says yes. A yes from something that can't say no carries no information.

We build Cruxproof, a research engine that checks a startup idea before you build it. We wanted one that could come back with "don't build this version" and be right about it. Here is what that changed in how it works.

1. It attacks the premise first

Almost every idea rests on one factual claim: "nobody affordable does this", "incumbents only sell to enterprises". The engine treats that claim as the main research target. If it's false, most of the idea falls with it, so the report leads with that instead of burying it.

2. Competitors alone never mean "don't build"

Most real businesses have competitors. They are the baseline you have to match and then beat, not a verdict. The question is whether there is a documented gap to beat them with. So there are three outcomes:

  • Build: the premise holds and there is at least one specific, documented weakness a newcomer can use.
  • Reshape: the idea as stated has a problem, but a narrower or differently priced version looks viable. This is the common honest answer.
  • Don't build: kept for fatal problems no reshaping fixes, such as a hard legal or platform blocker, or economics that can't work at any price.

3. Kind verdicts, never kind facts

When two verdicts are both supported, it may choose the more favourable one, and it has to put the uncertainty in the confidence section instead of rounding down to no. But leniency stops at the verdict. It must not invent a wedge, soften a real blocker, or say "build" when it found no documented weakness.

4. Prices come from the vendor's own page

Roundup articles publish prices that are wrong by large factors, and many are written by the vendors they rank. When a price decides the analysis, the engine looks for the vendor's pricing page. If it only finds a price secondhand, it says so and names the source.

5. It won't make up a search volume

Web search can't measure volume, so the engine never states one unless it found a published figure and can name where. Autocomplete shows a query exists and is common; it is not a volume. Evidence like competitors paying for the obvious ad terms, or whether the problem is discussed unprompted, is reported as exactly what it is: indirect.

6. Every material number is tagged

Each one is marked as verified at source, secondhand, or untraceable. If research fails, the report says the research failed.

7. Roadblocks are checked every time

Payment processors, ad policies, app store rules and licensing kill more products than competitors do, and founders rarely check them first. The engine works through that list even when you didn't ask, and reports only what genuinely applies.

The honest limit

It can be wrong. It researches live and cites as it goes, but you should check the two or three findings your decision turns on against the linked sources.

We launched on Product Hunt today. Launch week offer: $10 free credit, enough for two validations, no card, at cruxproof.com. It is our own product, so weigh that accordingly.