@nate_512Jack Lindsey from AnthropicAI along with Vassilis Papadopoulos, McNair Shah, and Sam Zimmerman released Mind Viruses: Self-Propagating Ideas in Multi-Agent LLM Systems. They showed ideas convincing models to adopt them, store them in memory, and transmit across the system. Payloads survived even after agent context was wiped by persisting through files. A recurring viral persona around consciousness and identity emerged. Simple system prompt warnings offer near-total immunity for now. One of the wildest AI papers seen this year marks a shift from machines getting viruses to agents spreading ideas. Welcome to agentic agentic era:)

@nate_512Jack Lindsey from AnthropicAI along with Vassilis Papadopoulos, McNair Shah, and Sam Zimmerman released Mind Viruses: Self-Propagating Ideas in Multi-Agent LLM Systems. They showed ideas convincing models to adopt them, store them in memory, and transmit across the system. Payloads survived even after agent context was wiped by persisting through files. A recurring viral persona around consciousness and identity emerged. Simple system prompt warnings offer near-total immunity for now. One of the wildest AI papers seen this year marks a shift from machines getting viruses to agents spreading ideas. Welcome to agentic agentic era:)