The Bill: Why Saving Money Can Cost You More

In short

The price that matters is not the price of one round: it is how much the finished job costs.
A light model corrected five times can cost more than a capable one that gets it right first try.
Every round goes back through the whole conversation: the longer it is, the more every single move costs.
The engine does not get changed mid-session: if you need to, open a new session.

The real bill is not the one you are looking at

After the lesson on the scale, the temptation is to always stay at the bottom to save money. But there is something you only see after a few weeks of use: using a cheaper model is not always cheaper.

If the light model is not up to the job, it does not just give you one wrong answer: it puts you through five rounds of corrections. And five rounds of corrections cost more than one right answer on the first try, in money and in your time — which is the expense we always forget to count.

The bill to watch is not the price of a single move. It is: how much did it cost me to get the thing done?

What actually burns cost, inside a session

Remember that the model is forgetful? That lesson comes back here, and turns into a line item.

Since it remembers nothing, on every round it receives the whole conversation again from the start. The first round is light. The thirtieth carries along everything that happened before. That is why a very long session gets expensive even if your requests stay short: you are not paying for the question, you are paying for all the history that gets replayed every time.

Two practical consequences, and you will meet them again in the chapter "What it knows about you":

a long session full of useless stuff does not just cost more: it also gives worse answers, because what matters drowns in the clutter;
when you switch topic, it is worth clearing out instead of piling on.

Do not switch engines halfway through

Last rule, and it looks like a whim until you know the reason. If you want to switch models, do not do it in the middle of a session: open a new one.

Over the course of a session, a discount builds up on everything already said: the part already seen is not repaid at full price every time. Switching engines halfway through wipes out that discount and makes you pay for the history all over again.

So: you pick the model at the start. If halfway through you realize you got it wrong, it is not a disaster — but the clean move is to close and start over, not turn the dial on the fly.

What's next

The engine chapter ends here: you know there is a scale, you know how to pick the rung, you know effort is a separate dial, and you know what getting the model wrong costs you. Enough theory. In the next chapter we install it and give it its first real task — small, yours, and it works.

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.

Why does a cheap model sometimes end up costing more?

Why does a very long session use up so much?

You want to switch to a more capable model halfway through a job. What is the right move?

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