---
mw_bundle: 1
id: ycb00hEg
title: "AI Memory Stack"
url: https://memory.wiki/b/ycb00hEg
document_count: 3
updated: 2026-05-24T17:48:04.238Z
analysis_generated_at: 2026-05-09T20:22:44.329Z
analysis_stale: true
source: "memory.wiki"
---
# AI Memory Stack

> Notes and references about AI memory systems including Mem0, Letta, OpenAI Memory, LLM Wiki, embeddings, and vector DBs. Background reading for the Memory.Wiki architecture.

> ⚠ _Analysis may be stale — one or more member docs were edited after the last analysis run. Re-run the canvas to refresh._

## Summary

These documents present mdfy.app as a comprehensive solution to AI memory fragmentation, offering a unified markdown hub that enables cross-platform content creation and sharing. The collection contrasts mdfy's open approach with proprietary AI memory systems like OpenAI's ChatGPT, while establishing technical specifications for interoperable AI context bundles. Together, they articulate a vision for user-controlled, portable AI memory that breaks vendor lock-in through open standards.

## Themes

- AI Memory Portability
- Cross-Platform Integration
- Open vs Closed Systems
- Markdown-Centric Workflows

## Cross-document insights

- mdfy positions memory ownership as the next battleground in AI vendor lock-in, suggesting that whoever controls the memory layer controls multi-AI workflows
- The Bundle specification represents a strategic attempt to create industry standards before proprietary formats dominate the AI context market
- The technical comparison between Letta and Mem0 reveals fundamental architectural disagreements about how AI memory should be structured and accessed
- mdfy's cross-platform approach suggests a hub-and-spoke model where content creation happens everywhere but memory persistence happens centrally

## Key takeaways

- mdfy.app positions itself as the solution to AI memory vendor lock-in by providing user-controlled, portable memory that works across all AI platforms
- The Bundle specification is mdfy's attempt to establish an open standard for AI context before proprietary formats dominate the market
- Cross-platform integration and data portability are core differentiators that enable users to maintain consistent workflows regardless of their preferred AI tool

## Open questions / gaps

- No discussion of enterprise security, compliance, or privacy considerations for organizational AI memory management
- Missing technical details about how mdfy's Bundle format actually integrates with different AI platforms in practice
- No business model or monetization strategy explanation for how mdfy sustains its open approach against well-funded proprietary competitors

## Notable connections

- **doc:ycd02zOQ** ↔ **doc:ycd00PnE** — The OpenAI critique sets up the problem that mdfy's comprehensive platform features are designed to solve
- **doc:ycd01N9A** ↔ **doc:ycd02zOQ** — The technical memory system comparison provides context for understanding why OpenAI's closed approach is problematic
- **doc:ycd04NxU** ↔ **doc:ycd00PnE** — The Bundle specification provides the technical foundation that enables mdfy's cross-platform sharing capabilities
- **doc:ycd04NxU** ↔ **doc:ycd02zOQ** — The open Bundle standard directly addresses the vendor lock-in problem identified in the OpenAI critique

## Concepts (this bundle)

- **Markdown Hub**
- **AI Memory Systems**
- **Vendor Lock-in**
- **Open Standards**
- **Context Management**

## Concept relations

- **Vendor Lock-in** ↔ **AI Memory Systems** — controls access
- **Open Standards** ↔ **Vendor Lock-in** — opposes restriction
- **Markdown Hub** ↔ **Open Standards** — implements philosophy

## Documents

### 1. [Letta vs Mem0](https://memory.wiki/ycd01N9A)
> Side-by-side comparison with Letta — what they share, where they diverge.
| | Mem0 | Letta | |---|---|---| | Storage | Vector + graph | Hierarchical (core / archival) | | Authorship | Auto-extract | Auto-extract + manual edit | | Hosted | Yes | Yes + self-host | | Format | JSON memories | Block-based | | OSS | Yes | Yes (formerly MemGPT) |
*sections:* Where they differ: Letta inherits MemGPT's "context windows are an OS resource" framing — | Where they're similar: Both treat memory as something the system extracts from the user's

### 2. [OpenAI Memory: Behind the Walled Garden](https://memory.wiki/ycd02zOQ)
> Why ChatGPT Memory is closed-by-design and what that costs users.
ChatGPT's "Memory" feature is closed. Memories live in OpenAI's database, shaped by their extraction model, only retrievable inside ChatGPT.
*sections:* What that means in practice: Switch to Claude → memories don't follow; Want to inspect what's stored → opaque; Want to deploy "what ChatGPT knows about my project" to Cursor → can't | The lock-in: Memory is the new vendor lock-in. Whoever owns your memory layer owns

### 3. [mdfy Bundle Spec v1.0 (Draft)](https://memory.wiki/ycd04NxU)
> Our own spec — the open standard piece.
A Bundle is a markdown document that wraps other markdown documents as deployable AI context.
*sections:* Frontmatter (YAML) | Body structure: H1 = bundle title; Optional blockquote = description; H2 sections per included document, in sort order; Each section: title, "Source: <url>", optional annotation, body | Hub subtype: type: hub means body contains an INDEX (TOC + recent + categories) | Goal: Open standard. Any tool can produce or consume Bundles. mdfy.app is


_Digest view — follow any link above to fetch that doc's full markdown. Add `?full=1` to this URL for the concatenated payload._