Yannic Explains: LLMs vs Agents vs Assistants vs Workflows

In this video, Yannic Kilcher, PhD, co-founder and CTO at DeepJudge, unpacks some of the most overused buzzwords in AI. A large language model (LLM) is simply a statistical model of text: you put text in, it predicts text out—nothing more. Terms like agent and assistant are mostly generalized labels, with “agent” often implying more steps, planning, or tool use, but in practice the line is blurry. Finally, a workflow is about structure: chaining steps together, sometimes with decision points, to complete a task. 

‍