What Claude Code really is
People keep asking me the difference between Claude Code, OpenClaw, Hermes, Claude Cowork and every other AI application out there, and the honest answer is that they are much easier to understand than they look. In this walkthrough I give you the one mental model that sorts all of them: the AI model is the brain, and the app around it is the body. Five minutes, nothing to install, and you will never confuse an LLM with an AI application again.
1. Three Claude apps, one detail that matters (0:00)
I start from a few Post-it notes. On the upper side sit the three main applications by Anthropic: Claude Chat, Claude Cowork and Claude Code. Notice that every single one of them has the word Claude in the name.
That might seem like a very small detail, but it is actually important. The word Claude refers to the brain of the application. All three apps are built around the same Claude intelligence, and that shared name is the clue.
2. The LLM is the brain (0:41)
The brain is the thing you always hear about: the LLM. And here is the part most people skip. The brain by itself cannot really do anything. The only thing an LLM can do is give you the next most probable word whenever you ask it a question.
That is it. No files, no browsing, no actions. Raw intelligence with no way to touch the world.
3. Apps are bodies for that brain (1:00)
So how do you make that brain useful? You build a body around it. That is exactly what the applications on the bottom row of my diagram are. Claude Chat, Claude Cowork and Claude Code have the same brain but correspond to different bodies. You can think of them as different Iron Man suits: one for Chat, a purple one for Cowork, an orange one for Claude Code.
The key thing to understand is that the brain in all three applications is literally the same. What changes are the capacities of that brain, because it sits inside a more capable body. Claude Code, Claude Cowork and Claude Chat are three different bodies for the same Claude brain.
4. Hermes and OpenClaw are bodies too, just open ones (1:57)
Where it gets interesting is when you start hearing about applications like Hermes and OpenClaw. What are they really? Very simple: they are also bodies. They are just not created by Anthropic, the company behind the Claude models. They are created by independent developers. Hermes belongs to a company called Nous Research, and OpenClaw belongs to Peter Steinberg.
The difference between the bodies created by Anthropic and these bodies is that these ones can host any brain out there. You can put a GPT brain inside, an Anthropic brain, maybe even a Gemini brain. You have a fixed body and you drop in whichever brain you want.
With Claude Chat, Claude Cowork and Claude Code it is different. These are private bodies, and you cannot put any brain you want into them. The one exception is Claude Code, and the only reason you can change Claude Code's brain is that the code got leaked.
5. What "harness" really means (3:29)
This flexible body has a name: a harness. It is a complicated word for something very simple, and it is your key learning today. The harness is everything that goes around the brain of your AI application.
So ChatGPT is a harness. Claude Code is a harness. Claude Cowork is a harness. And if you ever get advanced enough, you can even build your own harness, your own body designed for your own use cases.
This is why the same brain in different applications gives you extremely different performance. The suit decides what the intelligence can actually do.
6. Why the body matters more than the brain now (4:15)
Here is the practical takeaway. Given the state of the art LLMs we already have, what truly matters right now is the body, not the brain. The capacities you get are dictated by the harness: the tools it exposes, the context it manages, the actions it can take.
You have two routes. You can use the off the shelf bodies from Anthropic, or you can bring your own brain to a flexible harness like OpenClaw, Hermes, or whatever harness provider shows up next. Either way, when you evaluate an AI product from now on, stop asking only how smart the model is. Ask what the harness lets it do.
That is the whole model: same brain, different bodies, and the body is where the leverage is.
Where to go next
If you want to see this exact brain and body logic play out inside OpenAI's ecosystem, watch ChatGPT will die because of this. And if you are curious about the brains themselves and how open models are catching up, check Stop Ignoring Open Source AI.
