Honest by design

A map is only worth leaning on if you can trust every line of it. Jensen’s core rule is that it never guesses.

The rule

Everything Jensen shows you is one of two things:

1. Observed. Something it read directly in your source.
2. Declared. Something your team asserted on purpose.

Nothing is inferred, and nothing is filled in to look complete. The two kinds stay distinguishable, each carrying the evidence behind it, so you can always ask where a fact came from. What Jensen does not know is marked unknown until someone tells it.

Why it matters

That single guarantee is what makes the map safe for a human to rely on and safe for an AI to build on. When the map says two services are connected, they are. An assistant reading a map that never guesses cannot inherit a hallucinated architecture from it, which is the failure that makes AI assistants expensive on large codebases.

What this rules out

• Jensen will not draw an edge because two services have similar names.
• It will not name an owner it cannot evidence.
• It will not describe a route it did not find.
• It will not present an assistant’s inference as fact. When an assistant proposes an architecture, Jensen stores it as inferred and shows it to you for confirmation. It never enters the map until you confirm it, and an assistant cannot confirm its own inference on your behalf.

Where declaring fits

Some structure cannot be read out of any single repository: which services exist across your estate, how they connect, over what protocol. That is what declaration is for. You state it, and Jensen folds it into the same picture, still labelled as declared rather than observed.

See Declaring topology.