Translation

Intention-vectors compose and conflict

Multiple perspectives exerting intentions over the same data region produce a vector sum at that region. Aligned intentions produce coherent information. Conflicting intentions produce stress in the field. The model can in principle distinguish productive disagreement (vectors at angles, both contributing magnitude) from unresolved disagreement (vectors pointing opposite, summing to near-zero with high latent tension).

Multiple perspectives engaging the same data field with aligned good-faith intentions still produce different information shapes — different terms, different weights, different salience structures — because each information shape is grounded in a particular perspective-vector and there is no neutral ground onto which to project them.

The ops head's "feature" and the PM's "feature" and the tech lead's "feature" are not different names for the same information; they are different information, even when they point at substantially the same data.

This rules out a class of design ambitions knowledge systems regularly attempt and fail at: producing a canonical, neutral, role-independent view. There is no such view to produce. A system that tries to merge perspectives into one frame either silently privileges one perspective as canonical, or produces a flattened average none of the perspectives actually inhabits. Both are translation failures masquerading as resolutions.

What the system can do — and what attention-economy demands it do — is make perspective-differences visible without claiming to have resolved them. Rather than producing a neutral view, the system can show the same region's information shape from multiple perspectives, where the shapes overlap, and where they diverge. Translation becomes a first-class activity supported by the system, not an outcome it pretends to deliver.

A further depth in the irreducibility is worth naming. Each perspective's information shape is itself a contingent stabilization — the emission a perspective makes is the outer-boundary representation of an internal play of forces, not a transparent window into a univocal underlying intent. Translation is therefore not between two clear views of the same data; it is between two boundary-emissions, each of which is already a stabilization of internal flux on its side. The model is not just acknowledging that perspectives don't converge with each other; it's acknowledging that even within a single perspective there is no convergence to be reported.

Translation events, bridges, and translation-scars

When two perspectives bring their information shapes into dialogue and arrive at shared understanding — or fail to — that is an event with the same status as a decision or a research note. The trace of translation work over time accumulates into a bridging structure over the base topology, distinct from but related to it. Successful translations leave bridges; failed translations leave scars at the translation point rather than at the underlying data. Both belong in the substrate.

Knowledge is the experiential outcome when a perspective engages a projection of the represented state. It is a relationship, not a thing. The system stores conditions for knowledge production, not knowledge itself. The data / information / knowledge stratification is the right vocabulary for naming the layers of the phenomenon, but the system's representational scope ends at projection — knowledge is what happens when a perspective engages with a projection.

This is the closing commitment of the model, and the place where the model's discipline is most visible: it refuses to claim that any of what it represents is the thing that matters. What matters is what perspectives do with it.