🤖 I investigate malicious activity on social media networks 🕵🏼
@emilioferrara.bsky.social
📄 Paper: arxiv.org/pdf/2510.25003
📊 Interactive Dashboard: llmxio-dashboard.vercel.app
@emilioferrara.bsky.social
📄 Paper: arxiv.org/pdf/2510.25003
📊 Interactive Dashboard: llmxio-dashboard.vercel.app
🧭 Curious how this plays out? An interactive dashboard lets you inspect agent behaviors, network evolution, and coordination dynamics directly.
🧭 Curious how this plays out? An interactive dashboard lets you inspect agent behaviors, network evolution, and coordination dynamics directly.
- Malign agents' networks become denser & more clustered;
- Interactions grow more reciprocal & increasingly positive;
- Narratives converge toward greater semantic homogeneity;
- Amplification becomes synchronized.
- Malign agents' networks become denser & more clustered;
- Interactions grow more reciprocal & increasingly positive;
- Narratives converge toward greater semantic homogeneity;
- Amplification becomes synchronized.
Give AI agents just a bit of awareness of their “teammates,” and they begin to self-organize. No scripts. No central command. Just networked AI learning how to act together.
Give AI agents just a bit of awareness of their “teammates,” and they begin to self-organize. No scripts. No central command. Just networked AI learning how to act together.