Compare · three kinds of system
What Apodicta is, by what it is not
Apodicta is compared with a kind of system, never with a named product. Each page takes the question a buyer actually asks, "is this not just retrieval, a guardrail, a rules engine?", and answers it by what each one binds: what the model reads, what a reviewer sees, or what a line may say.
Retrieval-augmented generation
Retrieval changes what the model reads. It does not bind what the model writes.
The most common answer to hallucination, and the reason it is not enough: a model that has read the right case can still cite it for the wrong thing, or cite one it never read.
AI guardrails
A guardrail scores a probability. A gate decides whether a line may print.
A guardrail flags an output after it is written and leaves the decision to a person. A gate checks each line against the record before it exists, and holds what it cannot prove.
Legal rules engines
A rules engine looks up. This derives the structure, then reasons inside it.
A rules engine is deterministic because a person wrote every branch, which is what it costs when the law moves. Apodicta derives what a charge must prove from the authority, then reasons inside that structure.
The mechanism all three are compared against is on how it works; the terms are in the glossary.