Lesson 2 of 6
The LLM as an operating system
6 min read
What if the chatbot isn't the product, but the processor? A growing idea treats the model as the center of a whole system. What plugs into it?
The model as the coordinator
Researcher Andrej Karpathy calls it the "LLM OS": picture the model as the CPU of a new kind of computer. On its own it just predicts text. But wire it to tools, a memory, other models, and the live web, and it becomes the thing that coordinates them — deciding what to call, reading the result, and folding it back into the answer.
The model isn't the whole system — it's the part that coordinates the rest.
Why the picture matters
This reframes what 'using AI' means. The model's own knowledge is fixed and imperfect; the power comes from what it can reach — a calculator for exact math, a memory for what you told it last week, a specialist model for images. Judge a system less by the raw model and more by how well it's wired to the parts around it.
A model's usefulness comes as much from what it's wired to as from the model itself.
It's an analogy, not a finished blueprint — but it's how many new AI products are being built: a model at the center, tools and memory around it.
The gist
- —One vision: the model is the CPU of a larger system.
- —It coordinates tools, memory, other models, and the web.
- —The wiring around a model matters as much as the model.
In the 'LLM OS' picture, what's the model's job?
Continue in the app
Take the whole The Future of AI course — tracked
Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.