Blog

Why per kilobyte, and what it beat

Cladior bills a reader for the data a site sends them. The obvious question is why that, rather than per request, per minute, or a cut of a subscription.

The answer is that per kilobyte is what was left after the other models broke. It was not picked because it sounded elegant.

The thing before this

Cladior came out of a product called Chaild. Chaild was a paid referral system tied to a set of apps. One monthly subscription bought access to all of them, you had to renew each month to keep using them, and you could earn by referring other people.

The money from subscribers went into a pool. That pool had to be divided between the apps somehow, and how it was divided is where the thing started coming apart.

Splitting it evenly

The first split was equal shares. Every app in the pool got the same amount.

This fails the moment somebody notices. An app that took an afternoon to build earns what one that took a year earns, so the rational move is to publish as many thin apps as you can manage. Nothing in the model stops it, and review policy does not fix a payout rule that pays for volume.

Splitting it by time spent

Paying by time looks like the obvious repair. An app nobody opens earns nothing, so the junk app problem goes away on its own.

It introduces a worse one. A fintech app like Opay loses badly to a blog, because reading an article takes longer than sending money. The payment took three seconds and was the more valuable thing that happened that day. Time measures how long somebody sat in front of a screen, which is not what they got out of it.

There is a second problem with time, and it is the one that ended the idea. Time is hard to measure honestly. A tab left open in the background looks the same as a tab being read. Both sides have a reason to inflate it and neither can audit the other.

What was left

Several things were tried in between. The one that held up was charging for the data actually moved between the reader and the server.

Data has a property the other two lack. Both parties can count it and arrive at the same number. The server knows what it sent, Cladior sits in the path and measures the same bytes, and a reader can check the figure against their own network tools. Inflating it means actually sending more data, which costs the sender as well.

That is what the founder means by calling it foolproof, in his words: you can't cheat them and they can't cheat you.

It also stops pretending that attention is worth. A 2 KB API response carrying a credit decision bills as 2 KB. A long article bills as what it weighs. Whether either is priced correctly is the developer's call, because the developer sets the rate per kilobyte. The meter reports what moved.

What it does not solve

Per kilobyte is a measurement choice rather than a pricing theory. It says how much was delivered. It says nothing about whether what was delivered was worth the money, and a developer who sets a bad rate gets a bad outcome with accurate accounting.

Open questions come straight out of it. A cached response is the clearest one: the bytes reached the reader without the origin doing the work, and whether that should bill is a judgement rather than a measurement.

Where this stands

Cladior has been in beta since July 2026. The metering runs, and the accounting behind it has invariant checks that run against live data. Payments are on Stripe test keys, and no payout has been sent to a developer yet.

So this is an argument about a design decision and how it was reached, not a report of it working at scale. The billing docs set out what gets counted, if you would rather read the mechanism than the reasoning.