Before anything else: thank you. To the people who were here on day 0, when there was nothing to show yet except the idea. To those who give us feedback, and to those who come to us with ideas of their own. To everyone who trusts the service enough to use it day after day. This is for you.
For a few years, my graphics cards worked for companies that never paid them. I am not saying that to complain. I am saying it because it is the reason Paralon exists, and because many of the people reading this have the same years behind them.
I work in tech, and AI is what I do in my spare time as well. So when decentralized compute networks started appearing, I did the obvious thing and joined as a provider. Not one network. Several. Each had raised millions, each had a roadmap, and each had a testnet.
What a testnet asks of you
If you have never run a node in one, this is the deal.
You keep a machine online around the clock. You pay for the electricity. You update the software when the team ships a new build, often the same night, because a node on the old version stops counting. You watch a Discord channel for announcements. Your hardware ages while it waits.
In return you get points, and a sentence. The sentence is always some version of "early contributors will be rewarded."
I did that for years. Across the networks I joined, the testnet never ended. There was a phase two, then a season three, then a migration to a new architecture that reset the leaderboard. Then, one after another, the announcements got shorter, and then the networks were gone. What I had earned, after all of it, was nothing.
The whole path never worked
It would be easy to say the idea was wrong. I do not think it was. What I can say is narrower, and it is only what I saw from my side of it: at none of them did I see the whole path work.
Pieces worked. A node connected. A dashboard showed it online. Points went up. But the path that matters has more steps than that. A customer pays for compute. The work runs on a provider's machine. The provider is paid for it. I never saw that happen from end to end, on any of the networks I ran nodes for.
Without that, there is nothing to pay providers with. The networks gathered machines first and counted them, and points were the tool for it. Points measure one thing, which is that your machine was switched on. They say nothing about whether anyone needed it.
A network can run like that for as long as its funding lasts. It cannot pay you, because nothing is coming in. Every provider in it is working for a future that depends on a path nobody has finished building.
The one rule
So when we started Paralon, we started with a rule, and everything else followed from it.
A provider is paid when a customer pays. Not when the network reaches a milestone. Not at the end of a season. When someone rents your machine, the rental is billed by the minute, and your share is credited by the minute. You keep 80% of the price. It is paid in USDC, and you withdraw it to your own wallet.
The same holds for inference. When a paid request runs on your card, the request pays the card that ran it.
That rule has a hard edge, and I would rather say it than have you find it. If nobody rents your machine, you earn nothing from rentals. We do not pay for being online with money we do not have. An idle card on Paralon earns the same as an idle card on your desk.
It also forced us to do things in an uncomfortable order. We could not gather a thousand nodes and announce it. We had to find people who wanted to rent a GPU and make sure the machine they got was the machine they paid for. That is why every node is benchmarked before it is listed, why rentals have a reliability score, and why a provider can set their own price. None of that would matter in a network with no customers. All of it matters in one where a stranger is spending real money on your hardware.
The people who came with us
The first providers on Paralon were people I knew from those other networks. We had sat in the same Discord servers, updated the same nodes on the same nights, and watched the same leaderboards reset.
They did not need the idea explained. They needed to see one thing, and it was not a roadmap. It was a withdrawal arriving in a wallet. That is still the only argument I trust.
Points, on top of payouts
Paralon has points as well. The programme is called Season 0, and a verified node earns points for the time it spends online.
The difference is the order. Here, points sit on top of payouts, not in their place. A rental pays in USDC for every minute it runs, with or without Season 0. Points are added to that, as recognition for the people who brought hardware to the network early and kept it there.
We will say more about Season 0 as it takes shape. What will not change is the order: paid work first, points on top.
Where we are
Paralon is a young network, and we run it with care. Payouts follow a fixed schedule: requests open from the 1st to the 5th of each month, every one is reviewed, and it reaches your wallet within 24 to 48 hours. Next, we intend to add payouts by bank transfer, against an invoice, for providers who operate as a business. Every machine is verified before it is listed and watched while it is rented. Availability moves with demand, so some days a card is already taken and some days yours waits for its next renter. And when something does not go as planned, we measure it and write it up, with the numbers, in our engineering posts.
We are building this for the long run, one paid rental at a time. And whatever comes next, one thing stays as it is: nobody here will be asked to keep a machine running for years on a promise.
If you have a GPU and you have been through what I described, this is how to add it. If you want to rent one, they are here.
