How we grade evidence
Every rule this planner runs on carries one of five words that say how sure we are of it. The word sits beside the rule wherever it appears - on a plant page, a plant team, the planner itself - and it is a door to this page. Here is what each one commits to.
- Well established35 rules
- The rule names its mechanism, and the sources behind it have been read and checked.
- Promising62 rules
- The mechanism is named and the literature backs it, but it is not airtight yet.
- A good hunch13 rules
- Plausible by its mechanism, with field studies thin - a lean, not a law.
- Contested2 rules
- The evidence points both ways, so the planner does not enforce it.
- Refuted1 rule
- The claim did not survive its sources - kept so the folklore can be answered.
Two things a grade is not
It is not a guess dressed up. A rule cannot hold any grade at all unless it can say, in one sentence, how it works - a claim that only correlates is downgraded or dropped. And it is not a promise that a person has read the sources: that is a separate mark. 112 of the 113 rules are marked verified, meaning someone read a real source named on the rule and confirmed the claim; the rest are marked unverified and are never presented as fact. The model that helped write the corpus is not allowed to verify its own work.
Where the grades live
Each rule page shows its grade, its mechanism, its sources and its change history. The whole graded set is one file you can download and keep, and how this is built explains why the corpus, not the app, is the product. The things we are least sure of have their own page.