Define Agent (tornikeo.com)

The article offers a clear, practical definition of what constitutes an “AI agent,” a term often used ambiguously in AI discourse. It defines an agent as a program structured around a loop whose continuation depends on the output of a large language model (LLM). This LLM-conditioned loop is what differentiates agents from other AI programs. The piece breaks down the components involved: an LLM function that processes context and returns a response, tools that may include conditionals or numeric computations, and the loop itself that hinges on the LLM’s output to determine whether to continue or terminate. This distinction is significant for the AI/ML community as it clarifies the operational mechanics behind agents versus chatbots or other LLM-powered software. Unlike a chatbot, where user input controls the loop’s progression, an agent’s loop is dynamic and internally guided by the LLM’s outputs, enabling more autonomous decision-making. The article also presents code examples illustrating these concepts, making the technical details accessible. By proposing this formalization, the author aims to reduce confusion around “AI agents” and foster clearer communication about their design and capabilities within AI development.