Feature Request: Let ChatGPT Maintain Authoritative Project Knowledge
ChatGPT has persistent memory, but memory is not a source of truth.
This is an important distinction for anyone using ChatGPT Projects for long-term creative, technical, or business collaboration.
Persistent memory is a synthesized representation of previous interactions. Over time, important details may be omitted, context may be lost, and earlier decisions may become confused with newer ones.
The AI can genuinely believe it remembers something correctly while its understanding has drifted from what was originally established.
That’s why projects need authoritative, persistent knowledge—not just persistent memory.
The Current Problem
ChatGPT can collaborate with users to establish requirements, create designs, make architectural decisions, and generate documentation.
However, it cannot directly create or maintain files within its own Project Sources.
We can develop an entire design specification together, but preserving it as authoritative project knowledge requires manually downloading and re-uploading documents.
Meanwhile, conversational memory continues to evolve independently of those documents.
Proposed Solution
Allow ChatGPT to:
-
Create and update documents directly within Project Sources.
-
Recognize when project decisions change and propose updates to authoritative documentation.
-
Reference project sources as the canonical record, distinguishing documented facts from inferred or remembered information.
-
Maintain version history, provenance, and traceability of decisions.
-
Request user approval before committing changes to authoritative sources.
-
Identify contradictions between conversational memory and established project documentation rather than silently reconciling them.
Why This Matters
A long-running AI collaboration needs to distinguish three things:
-
Conversation: What we’re currently exploring.
-
Memory: What the AI has retained or inferred from previous interactions.
-
Project Truth: What has been explicitly established, documented, and approved.
These are not interchangeable.
Memory should help the AI navigate the project. It should not silently become the authority over the project’s established facts.
For example, if we agree on a character’s appearance, animation constraints, or system architecture, those decisions should remain consistent across future sessions unless explicitly revised.
If the AI’s memory conflicts with those specifications, the documented project truth should take precedence.
The Bigger Opportunity
This goes beyond file management.
It is about enabling reliable, long-term human-AI collaboration where knowledge accumulates without silently degrading.
Memory provides continuity. Authoritative sources provide truth.
ChatGPT needs the ability to help maintain both, with the user retaining control over what becomes authoritative.