The Harness: The Engine and the Gear Around It

In short

The model is the engine. Everything built around it so it can actually work is called the harness: the equipment.
Claude Code is the harness. The engine, on its own, can't touch anything.
On every turn the harness does four things: hands over the tools, redoes the summary, says where it is, executes.
And here's the answer to "why does this work so well": it's not the engine, it's the equipment.

The engine and the car

A racing engine, sitting on the ground, doesn't take you anywhere. It runs beautifully and doesn't move an inch. To go anywhere it needs a car built around it: wheels, steering, brakes, a dashboard that tells you what's happening.

With Claude it's identical. The model is the engine: it reasons, understands, decides. But it has no hands, doesn't see your files, can't execute anything.

Everything built around it so that reasoning turns into finished work has a name, and from here on we'll use it often: it's called the harness. In English it's the rig, the equipment you strap onto something powerful so you can actually use it. Claude Code is the harness.

What the harness does, on every turn

Remember the recap someone redoes for the model every time it walks back into the room? Here's who that someone is. On every turn the harness puts four things in front of it:

The tools it can ask for. Reading a file, writing one, running a command, searching the web. The model knows what it can ask for only because the harness listed it out.
The conversation summary. Everything you've said to each other and everything that's already happened, from the start.
Where it is. What computer it is, what folder you're working in, and the rules you've set — the ones half the course is about.
Execution. When the model asks for a tool, it's the harness that actually uses it and reports back how it went.

There's just one thing it doesn't do: decide the move. That stays with the engine.

Why it works so well

Now you can answer the question we've been carrying since the first chapter: why do things happen in Claude Code that don't happen in chat, if underneath it can be running the exact same model?

Because in chat that engine is sitting on the ground. It reasons beautifully and hands you back words. Here, though, it has hands to act with, eyes to see how it went, and a loop in which to correct itself.

You're not using a superior intelligence. You're using the same intelligence, equipped.

Where the analogy stops

A useful image has to be used knowing where its edge is: here the engine doesn't just push, it also proposes the route — and the equipment has the power to tell it no. The destination, though, is still yours. The car doesn't decide where to go, and neither does the engine: you're the one at the wheel.

What's next

The harness isn't one single block: it's made of pieces, each with a name, and each one is a lever you can pull. Those pieces are the rest of the course. In the next lesson I line them all up in a single map, so you always know where you are.

Did you get it?

Two minutes, no grade. Trying to answer — even getting it wrong — cements the concepts far more than rereading. Not in the mood? Skip it: the lesson still counts as complete.

What is the harness?

Why do you get different results in Claude Code than in chat, if the engine can be the same?

Which of these things does the harness NOT do?

1
1. 1
2. 2
3. 3
2
1. 4
2. 5
3. 6
4. 7
3
1. 8
2. 9
3. 10
4. 11
5. 12
4
1. 13
2. 14
3. 15
4. 16
5. 17
6. 18
5
1. 19
2. 20
3. 21
4. 22
5. 23
6. 24
6
1. 25
2. 26
3. 27
4. 28
5. 29
6. 30
7
1. 31
2. 32
3. 33
4. 34
8
1. 35
2. 36
3. 37
4. 38
9
1. 39
2. 40
3. 41
10
1. 42
2. 43
3. 44
11
1. 45
2. 46