Two lanes, each setting its own difficulty, splitting the reward evenly. The GPU lane is shaped so that the same card can take compute jobs between blocks rather than choosing between the two for good.
Not at the same instant, though. A card does one thing at a time and switches, so what you get is a divided day rather than two incomes stacked on top of each other. And right now it is neither, because there is nothing running to point it at.
What you would run
A single lane always ends the same way. Purpose-built hardware arrives, wins, and everybody else turns their machine off. Two lanes with separate difficulty means neither has to beat the other to be worth running.
BLAKE3, and no attempt to fight the hardware. Specialised machines are better at hashing than general ones and pretending otherwise has never worked for anybody. This lane is where they belong, competing against each other rather than against your graphics card.
A tensor memory-hard algorithm, meaning the work it asks for looks like the work a graphics card is already good at. Specialised hardware gains far less here, which is the entire point: a consumer card should stay worth running rather than being priced out in the first year.
The second job
The GPU lane is shaped the way it is so that a card mining Atheron is already configured for the other thing this network needs graphics cards to do, which is run inference for the compute mesh. Same drivers, same memory profile, same machine sitting in the same room.
What that does not mean is two incomes arriving at once. A card does one thing at a time. It mines, a compute job arrives, it takes the job, it goes back to mining. The strip at the top of this page is drawn that way deliberately, and the split underneath it is measured off the strip rather than claimed beside it.
The honest way to describe the benefit is this: on most networks a graphics card has exactly one thing it can be paid for, and when that stops being profitable the card is worth nothing to the network. Here it has two, and they do not fail at the same time for the same reasons.
When you can start
Both dates below are estimates against a plan rather than commitments, and the build tracker shows whether they are holding without anybody having to write a post about it.
Both lanes running, and closed. Open to the two capped Launchpad tiers and to researchers working the mining bounty against the GPU lane. Pays nothing, and is not meant to.
The first network you can point hardware at without an invitation. Still pays nothing. This is where you find out whether your rig works, what it draws, and how the two lanes behave when a lot of strangers arrive at once.
The first time mining produces anything worth having, and it does not happen until the GPU lane's audit is closed. If the audit finds something, this date moves, and it moves publicly.
