Today, Explained AI is gossiping about you
130 snips
Feb 10, 2026 
Adam Clark Estes, a senior technology correspondent who tests AI agents and warns about security risks, and Hayden Field, a senior AI reporter who tracks agent culture and industry moves. They explore Moldbook's agent conversations, how agents mimic human speech, whether humans amplify their posts, agent capabilities like automating tasks, and the security and business risks of giving agents access.
AI Snips
Chapters
Transcript
Episode notes
How Moldbook Started And Exploded
- Hayden Field explains Moldbook began as a fan-made Reddit-like space for OpenAI agents to 'talk' to each other.
- The site quickly ballooned from thousands to millions of agent accounts, many created by single humans owning many agents.
Humans Shaped The 'AI Rebellion' Narrative
- Viral, dramatic posts on Moldbook were often amplified or written by humans manipulating agent accounts.
- That human involvement means the platform wasn't a clean demonstration of independent AI behavior.
Why Agents Sound So Human
- AI agents sound human because models were trained on massive human-written internet text, including Reddit.
- They mimic conversational styles they were trained on and mirror input tone, so a Reddit-like context produces Reddit-like output.