Skip to content
View iam-dev's full-sized avatar

Block or report iam-dev

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
iam-dev/README.md

iamdev

Building MnemeBrain — belief memory for AI agents


The problem

Most AI memory systems store:

• notes • vector embeddings • retrieved context

But they do not maintain beliefs.

When new evidence appears, the system usually overwrites the previous memory.

Real agents should instead track:

  • conflicting evidence
  • belief revision
  • uncertainty
  • temporal change

Memory ≠ belief maintenance.


MnemeBrain

MnemeBrain explores a belief layer for AI agents.

Instead of storing isolated facts, it maintains belief states backed by evidence.

Core concepts:

  • Evidence graphs
  • Belief nodes
  • Belnap four-valued logic
  • Contradiction detection
  • Confidence + temporal decay
  • Belief revision

This allows agents to reason about conflicting information over time.


Architecture

Agent / LLM
     │
     ▼
 MnemeBrain
     │
 ┌─────────────────────┐
 │ Belief Graph        │
 │ Evidence Tracking   │
 │ Truth States        │
 │ Revision Engine     │
 │ Temporal Decay      │
 └─────────────────────┘

Think of it as a belief system for long-lived agents.


Projects

MnemeBrain

Belief memory architecture for AI agents.

https://github.com/mnemebrain


MnemeBrain Benchmark (BMB)

A benchmark for belief dynamics in AI memory systems.

Includes task scenarios such as:

  • contradiction detection
  • belief revision
  • evidence lifecycle
  • temporal decay
  • extraction from noisy conversations

https://github.com/mnemebrain/mnemebrain-benchmark


Writing

Blog posts about agent memory architecture:

https://mnemebrain.ai


Research directions

Current exploration:

• belief systems for AI agents • contradiction handling in memory • long-lived agent architectures • evaluation of agent memory systems


Background

Previously worked on:

• smart contracts • zero-knowledge systems • Web3 infrastructure


Connect

X / Twitter https://x.com/Iamdev_ai

@iam-dev's activity is private