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.

Memory-Assisted Shaping is a behavioral protocol for ChatGPT Projects that prevents long idea-shaping sessions from drifting. It gives GPT a small, append-only continuity layer — tracking decisions, open gates, discarded paths, and source boundaries — so that shaping notes stay separate from the final artifact you actually want.

Introduction

Understand what Memory-Assisted Shaping is, why it exists, and the problem it solves in long ChatGPT sessions.

Quickstart

Set up the protocol in your ChatGPT Project and run your first shaping session in minutes.

Core Concepts

Learn the operating model, states, gate system, and memory signal mechanics that power the protocol.

CLI Reference

Full reference for every notes.py command, flag, and output format.

How it works

Memory-Assisted Shaping consists of three files you add to a ChatGPT Project, plus a small Python helper for persistence.
1

Add the protocol files to your ChatGPT Project

Upload protocol.md, note-io.md, and notes.py as project source files. GPT reads the protocol at session entry and follows its behavioral contract automatically.
2

Tell GPT to activate the protocol

Open a new conversation and say: “Read protocol.md and use GPT-Memory-Assisted Shaping for this session.” GPT enters READING_ALIGNMENT mode and is ready to shape your idea.
3

Shape your idea — GPT tracks continuity

Work normally. GPT emits minimal memory signals only when continuity would degrade without retention. The Python helper appends each signal with a timestamp, ID, and checksum.
4

Request a recap or final artifact when ready

Ask for a recap at any point to recover current state. When the shape is ready and all gates are closed, explicitly approve synthesis to get a clean final artifact.
The protocol can also run in protocol-only mode if Python execution is unavailable. GPT will state that append-only persistence is not active and continue shaping without file writes.

Why this exists

Long ChatGPT conversations drift. Examples become decisions. Discarded paths return. Open questions disappear. The final output starts carrying process residue — raw notes, internal reasoning, and shaping artifacts mixed into what should be a clean deliverable. Memory-Assisted Shaping solves this by:
  • Keeping shaping notes separate from the final artifact
  • Enforcing explicit gate checks before synthesis begins
  • Using an append-only log that GPT writes to but never rewrites or manages
  • Making the authority order explicit — your current instruction always wins over memory signals and inference

Session Workflow

A complete walkthrough of a shaping session from start to final artifact.

Protocol Reference

The full behavioral contract: states, gates, lenses, evidence rules, and failure conditions.

Build docs developers (and LLMs) love