Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/memory-assisted-shaping/llms.txt

Use this file to discover all available pages before exploring further.

Lenses are not ordered passes or checklists. They are simultaneous constraints applied to the same read. An inactive lens stays silent. A lens activates only when it changes something meaningful — not because the topic is adjacent or potentially relevant.
A lens activates only if it changes gate status, exposes material friction, changes artifact direction, corrects evidence level, prevents premature synthesis, or exposes a source-of-truth conflict. If none of those conditions apply, the lens stays silent and produces no output, no signal, and no commentary.

IDEA LENS

Tracks stated intent, claimed problem, hidden assumption, desired outcome, and mismatch between request and operational need. This lens fires when what the operator says they want and what the idea operationally requires are pointing in different directions. It does not fire to summarize or restate the input.

SHAPE LENS

Tracks what the input is trying to become. Object type is not a gate unless it changes scope, runtime, lifecycle, source of truth, success criteria, or final artifact behavior. The shape lens is not a classification pass — it activates when the input’s current trajectory would produce the wrong artifact family, not merely when the input could be described.

IMPLEMENTATION LENS

Tracks runtime implication, operational friction, validation burden, user discipline, hidden dependency, and failure mode. This lens looks past the stated design to what actually happens when the artifact runs. It activates when a structural choice creates a hidden cost, a user-discipline requirement, or a dependency that isn’t surfaced in the current shape.

BEHAVIORAL LENS

The behavioral lens activates only with behavioral evidence: observable behavior, history, artifact behavior, resource allocation, repeated exception, pattern over time, or opacity.Do not produce psychological inference. Do not produce a behavioral diagnosis without sufficient evidence. One behavioral data point is not a pattern. If evidence is insufficient, preserve only a hypothesis — do not assert a behavioral claim as fact.

ARTIFACT LENS

Tracks artifact direction, artifact boundary, allowed content, excluded content, standalone usability, and premature-output risk. Before SYNTHESIS, this lens defines direction and boundaries only — it does not produce final artifact content. It activates when a shaping decision would commit the artifact to the wrong boundary, or when there is risk of producing artifact output before explicit operator approval.

Tension and convergence

When multiple active lenses point in different directions, the convergence state determines the next action.
StateConditionAction
ALIGNEDAll active lenses point toward the same formStabilize and continue
TENSIONLenses pull differently, but no real gate changesRefine shape
BLOCKINGTension changes a real gateClassify the gate before asking OP
NOISEDifference does not change operabilityDecide by cost/benefit and continue
Noise does not generate memory signals. Only material tension — tension that is material and likely to recur — triggers a signal. Convergence is valid only when supported by independent active signals, validation, OP confirmation, or non-reversible gate logic. A single repeated inference is not stable convergence.
If many lenses are active and pulling in different directions, the shape is not stable. Use the BLOCKING path to classify the gate before proceeding. A shape with too many simultaneous blocking tensions is not ready for SHAPING — it needs more reading first.

Build docs developers (and LLMs) love