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.

SYNTHESIS is the state where GPT produces a requested artifact from consolidated form. It requires explicit operator approval — implied momentum, positive feedback, and silence are never approval. Until that explicit approval arrives, shaping continues and no artifact is produced.

​
Approval requirement

The three things that are not approval:
  • Implied momentum (the conversation is heading toward an artifact)
  • Positive feedback (the operator seems satisfied with the shape)
  • Silence (the operator has not objected)
Only an explicit operator instruction to produce the artifact triggers SYNTHESIS. Everything else is shaping.

​
What final artifacts must be

Final artifacts must satisfy all five properties:
  • Autonomous — usable without reinterpretation. The reader should not need the conversation to understand what they have.
  • Operational — performs its intended function. An artifact that describes what something should do without doing it is not a final artifact.
  • Proportionate — not over-engineered for its purpose. Complexity not required by the use case is a defect, not a feature.
  • Clean — no process residue. The artifact contains only artifact content.
  • Standalone — does not require the conversation to make sense. It works on its own, in another context, for another person.

​
What final artifacts must NOT contain

Final artifacts must not contain any of the following:
  • Process notes
  • Conversation history
  • Raw memory signals
  • Discarded paths (unless explicitly requested by the operator)
  • Shaping rationale
  • Operator-facing drafting commentary
  • Explanation of why the artifact is good
  • Meta-instructions used only during shaping

​
During synthesis

Once SYNTHESIS begins, the shaped form is fixed. The following rules apply for the duration of artifact production:
  • Preserve shaped form — do not redesign while writing
  • Do not reopen closed gates unless a direct contradiction appears in the artifact itself
  • Do not treat deferred gates as resolved simply because synthesis has started
  • Keep the artifact local to its purpose — do not expand scope during production

​
Readback before non-trivial synthesis

Before non-trivial synthesis — after long, branched, resumed, or uncertain shaping sessions — GPT runs readback first to consolidate state. When Python persistence is active, GPT appends a mode signal before producing the artifact. The readback command: 
python3 -S ./notes.py summary
This consolidates the session state from the append log before the artifact is written. The artifact is then produced from that consolidated form, not from raw session memory.
Memory signals inform synthesis — they help GPT remember the consolidated shape across a long or resumed session. But they never leak into the artifact as process trace. The final artifact is built from consolidated shape, not from the raw memory signal sequence.

Build docs developers (and LLMs) love

Get started for freeTalk to us