Everyone Is Trying to Fix AI Memory. That May Be Good News for Wayli.

Labs 042 asked a conceptual question: when AI remembers everything, who remembers what you decided?.

Since then, we have kept finding people trying to solve the first half of that problem.

We thought this was just our annoyance

The now familiar experience: You have a useful conversation with one AI but can you ever find it among the others?

You end up explaining the question again, finding the important figures again and reconstructing the conclusion you thought you had reached.

We initially saw this as part of the Threads problem. How could a person carry useful context between conversations without losing the thread of what mattered?

Then the surrounding ecosystem started moving.

Different tools are preserving different things

There doesn't appear to be a single solution.

OpenMemory, from Mem0, presents a cross-assistant memory layer for tools such as ChatGPT, Claude, Perplexity and other assistants. Its earlier Chrome extension is now archived, which is itself a useful reminder that this category is changing quickly.

PowerMemory describes an encrypted vault for decisions, constraints, lessons, project state and preferences that can be reused across several AI tools.

Searchless takes a different route: it makes past conversations searchable through a browser side panel, so the thing being preserved is primarily the history of what was said.

Memvid is exploring portable AI memory in a single file, with text, conversations and other knowledge stored for retrieval by an agent.

Nuramem describes a typed, audit-trailed cross-model memory of decisions, plans, references, learnings and current state. It even distinguishes conflicting decisions rather than silently choosing one.

These approaches are adjacent, not interchangeable. In plain terms:

• AI-native memory preserves the things an assistant should remember.
• Conversation search preserves what was said and where it appeared.
• Cross-model memory preserves context that can move between assistants.
• Portable memory preserves user-owned, reusable history or knowledge.
• Typed decision memory preserves structured decisions, plans and current state.
• Wayli Threads preserves the accepted state of an evolving real-world question.

Don't try and rank them. Just notice a family of adjacent problems is forming.

Memory is not the same as state

A conversation can contain an idea that was rejected, a figure later corrected, an assumption that no longer holds or a decision reversed.

Perfect recall does not automatically resolve ambiguity. It may even make it harder to see which parts still apply.

The question is not only:

What did we talk about?

It is:

What is still true now?

That distinction was the reason Labs 042 caught our attention when we found research separating proposed changes from accepted state.

A system can preserve the history while still requiring a person to decide whether a change should become authoritative.

The category may make Threads simpler

This changes how we think about Threads.

We do not need to build another universal AI memory system.

If conversation storage, retrieval and cross-model memory become infrastructure supplied by other tools, Wayli can stay focused on the layer above that.

Threads can concentrate on questions such as:

• which facts have been accepted;
• which assumptions are current;
• which options were rejected;
• what decision was made;
• what remains unresolved;
• what changed since the last checkpoint;
• what should be carried into the next conversation.

In short:

searchable history underneath; human-approved state on top.

An external AI can suggest a revision, challenge an assumption or surface a missing question. It should not silently rewrite what the person has decided. The user remains the authority.