I'd just like to interject for a moment. What you're refering to as LLM, is in fact, Harness/LLM, or as I've recently taken to calling it, Harness plus LLM. LLMs are not a useful tool unto themselves, but rather another component of a fully functioning agentic system made useful by the Harness's vector database, tool calling interface, and vital system components comprising a full agentic AI system.

Many vibe coders run a modified version of the Harness system every day, without realizing it. Through a peculiar turn of events, the version of Harness which is widely used today is often called LLM, and many of its users are not aware that it is basically the Harness system, developed by Anthropic, OpenCode, and so on.

There really are LLMs, and these people are using them, but they are just a part of the system they use. The LLM is the underlying language model: the weights and biases that actually represent the statistical information contained within an enormous corpus of language. The LLM is an essential part of an agentic system, but useless by itself; it can only function in the context of a complete harness. LLMs are normally used in combination with harnesses, the whole system is basically a harness with LLM added, or Harness/LLM. All the so-called LLM chatbots are really distributions of Harness/LLM!