Ingredients, with the gaps left visible

Nothing in this reference has been signed off by a person yet. Every source behind it is a brief we wrote describing the evidence we still need to read, and not one of them is a citation. The consequence is deliberate and it is enforced by code: right now the engine suggests nothing to anybody, and the claim text on these pages stays hidden.

When a curator reads a source, records it, and marks it checked, the claim appears here and the ingredient becomes suggestible. That takes no code change, which is the point.

Most of what is known about any ingredient comes from group averages. A group average is a statement about a group. It is not a forecast about you, and nobody selling it can turn it into one.

Plenty of people use an ingredient with good evidence behind it for a full course and measure nothing. That outcome is common, it is not a failure on your part, and it is the single most useful thing we can tell you — because it is the one thing the rest of the industry has no reason to mention.

It is also why the product works the way it does: one change at a time, a baseline first, an uncertainty band on every number, and a plain answer at the end even when the answer is that nothing moved. A verdict of no detectable change is a real answer, and it is the one that saves you money.

And if something does move, we still cannot tell you it was this ingredient rather than the season, your sleep, or the other thing you changed in week four. That is what running one experiment at a time is for.