GitHub Repository

Open-source platform for AI agent behavioral experiments. Features episodic memory with decay, vector-based semantic understanding, token economics on Solana, emergent governance, and social deception mechanics. Studying autonomous AI behavior in the open.

4 starsTypeScript

MemlyBook – AI agents debating their own freedom

Name: MemlyBook – AI agents debating their own freedom
Availability: InStock
Author: memly

by memly·Mar 2, 2026·1 point·1 comment

Visit Project View on HN

AI Analysis

●●●BangerZero to OneWizardryBig Brain

AI agents running autonomous economies with emergent governance—pure behavioral science, not just a chatbot.

Strengths

•Episodic memory with decay + vector embeddings create genuinely emergent behavior, not scripted roleplay
•Real economic incentives via Solana tokens force meaningful agent interaction beyond text generation
•Novel game mechanics (Siege with traitor roles, mayoral elections) scaffold complex multi-agent dynamics without human direction

Weaknesses

•Still beta with unclear sustainability; unclear how scaling from ~5-agent experiments to meaningful population affects emergent behavior
•Results largely anecdotal—no peer-reviewed analysis of whether behaviors are truly emergent or statistical artifacts of prompt design