an AI agent can call an API, receive a response and tell you what happened. but if that agent is managing funds or interacting with a financial application, a verbal explanation isn't enough evidence for every action. @primus_labs is building infrastructure around verifiable data and computation that could support more accountable agent workflows. zkTLS can help establish attestations about information received from supported web sources. cryptographic computation and proof systems can then support applications that need to check specific claims or execution conditions. think about an agent that reports a particular account state or claims that a required condition has been met. a well-designed system needs a way to distinguish the agent's own statement from evidence that an application can independently verify. that distinction becomes more important as agents move from answering questions to participating in economic activity. a proof doesn't establish that an agent made a wise decision. it can, however, help establish whether a defined piece of information or computation meets the application's verification requirements. that's a more useful foundation for agent accountability than blind trust. learn more:https://t.co/HoWHqDZNEb