x20 is not 4x: the multiplier on the box is not the quota in the wire
A few nights ago a tech Discord server I am in spent two hours arguing about a number. Not a benchmark, not a model. A price tag. One person said the x20 plan delivers about 1.9 times the quota of the x5, not four times. Another said the 20x is the five-hour limit and the weekly limit is only 10x. A third dropped a link to a pricing table showing 20x landing at roughly twice 5x. Someone concluded "x6 is the standard," which made no sense and perfect sense at the same time, because nobody in the room had measured anything. A thousand messages and the argument ended where it started: still arguing, still no numbers.
I stayed quiet through most of it because I had been the person who buys the bigger number before. I wrote about that already. This time the question was narrower and more annoying: is the multiplier on the box real?
Three numbers wearing one name
Here is what I pieced together from the Discord thread, and I want to be careful about levels here. I did not run the measurement. These are claims from people I talk to, plus one community pricing table, not my own logs. What I trust is the shape of the confusion, because I have seen this shape before.
The marketing gives you one number. Twenty. It sounds like twenty times something. But the plan actually contains at least three different quantities, and they do not move together:
| What the number means | x5 plan | x20 plan | Ratio |
|---|---|---|---|
| Five-hour rolling limit | 5x | 20x | 4.0x |
| Weekly limit | 5x-ish | ~10x | ~2x |
| Tokens you actually burn through in a week of real work | varies | varies | ~1.9–2x (reported, not measured by me) |
Nobody in the Discord thread disputed the first row. The fight was entirely about the second and third rows, which are the rows that decide what you actually get. A 20x five-hour limit with a 10x weekly limit is not a 20x plan in any sense that matters to a person working across a week. It is a plan that lets you sprint four times as fast and then hits the same wall twice as late.
That is not fraud. It is arithmetic wearing a costume. The costume works because nobody reads the second row until they have already paid for the first.
Why I believed the box the first time
I have a history with this. When I upgraded from twenty to fifty to a hundred a month, I did the math in the way that made each step look reasonable, which meant not doing it at all. A bigger number on the plan page feels like more of the thing. More quota, more room, more building. The feeling is doing all the work and the number is just its costume.
The x20 argument was the same feeling at higher resolution. Twenty is bigger than five. Four times bigger, by the name. The people arguing for it were not stupid; they were doing what I did, which is trusting the label instead of the contents. The people arguing against it had fragments: a table here, a limit observed there, a week where the x20 felt like barely double. Fragments, not a study. But fragments pointing the same direction are how every investigation I have ever done started.
The only measurement that would settle it
The Discord thread kept asking for someone else's number. A table, a review, a post. I think that is the wrong ask. The right ask is a method anyone can run in an evening, because the quantity that matters is personal: how much finished work does each plan buy me, on my tasks, in my week.
Here is the version I would trust, and the version I am writing down so I actually run it:
- Pick one repeatable unit of work. Same repo, same kind of slice, something you do weekly.
- Run it on the x5 until you hit a limit. Count finished units, not tokens, not hours. Finished units.
- Do the same week on the x20. Same kind of work, same counting.
- Divide. That number is your multiplier. Nobody else's.
Tokens are the wrong unit because a token spent re-reading context is not a token spent finishing work. Hours are the wrong unit because an hour of a swarm looking busy is not an hour of building. I learned both of those the expensive way. The only unit that survives contact with my actual evenings is the finished thing.
Until someone runs that, and I include myself, every number in the argument is a claim about the box, not a measurement of the wire.
What I am doing instead of upgrading
I did not buy the x20. Not because I know it is a bad deal. I do not know that. Because I do not know what it is, and I have a rule now about buying things I cannot state the contents of. That rule cost me a year of invoices to learn, so I am keeping it.
What I did instead was smaller and duller. I picked the plan whose limits I have actually hit, and I stayed on it until I can say what a week of my work costs in finished units. When I know that number, the multiplier question answers itself: either the bigger plan buys proportionally more finished work, or it buys the feeling of more. I have already bought the feeling several times. It is the most expensive thing I own.
Do not buy the multiplier. Buy the output you can count. And if you cannot count it yet, that is the measurement to run before the purchase, not after.
The multiplier on the box is not the quota in the wire. I know that the way you know anything before measuring, which is to say I suspect it strongly and cannot prove it yet.