Bittensor is a network that pays people and teams for producing useful AI — like intelligence, computation, and predictions — instead of paying people for burning electricity on arbitrary puzzles, the way Bitcoin does. It runs on its own asset, TAO. Vermilion pays you TAO, and this page explains what that actually is.
Picture a marketplace made of many smaller marketplaces. Bittensor calls each smaller marketplace a subnet. One subnet might specialize in AI text generation, another in image generation, another in financial predictions, another in raw computing power for training models. Each subnet has its own competition running inside it: participants submit work, and the network scores that work on quality.
The network itself does the scoring through a mechanism called the incentive mechanism. Do good work, get paid more TAO. Do low-quality or lazy work, get paid less. It is a market that prices intelligence and effort directly, the same way a real market prices anything else people want.
Bittensor launched in 2021 with a simple bet: the most valuable resource in the coming decade would not be raw computing cycles, it would be genuinely useful machine intelligence, and no single company should own the market that prices it. Rather than one corporation training and owning a model behind closed doors, Bittensor tries to build an open, competitive market where anyone can contribute a model, a dataset, or compute, and get paid according to how useful their contribution actually is.
That is the thesis in one sentence: make intelligence a commodity that markets can price, instead of a black box that one company owns.
Every subnet has two kinds of participants. Miners are the ones doing the work — running a model, doing the inference, generating the answer. Validators are the ones grading that work, checking it for quality, and reporting scores back to the network. Good validators are rewarded for grading honestly; miners are rewarded for the quality of what they produce, not for how much noise they make about it.
Think of it like a marketplace of freelancers (the miners) and a marketplace of reviewers (the validators) running at the same time, continuously, with the network itself paying out based on the reviews. No manager, no single company deciding who gets paid; the scoring mechanism does that job.
A subnet is a self-contained economy inside Bittensor, built around one specific task. Some subnets focus on generating text, some on images, some on predicting financial markets, some on raw compute for training other models. Each subnet sets its own rules for how work is measured, but all of them plug into the same underlying network and the same asset, TAO.
As of this writing, some of these subnets already generate real revenue from real customers, not just from speculation. That distinction matters: a subnet with paying customers is a product, not a promise.
The incentive mechanism is the part doing the actual work of a market: it takes in scores for quality and effort, and turns them into a payout schedule, automatically, on a regular cadence. There is no committee approving payouts by hand. The code does it, continuously, based on the numbers the validators report.
dTAO (dynamic TAO) is the mechanism that decides how much of the network's total emission (the amount of new TAO created and paid out) each subnet receives. Instead of a fixed allocation decided once, subnets compete for emission based on how much value the market assigns them — more demand for a subnet's token pulls more emission toward it. It is the layer that lets the network reallocate resources toward whatever is actually working, rather than whatever was decided to matter a year ago.
Vermilion does not run a subnet and is not a Bittensor product. It is a separate token on Solana that converts a transfer fee into TAO and pays it to holders. The connection is simple: holding Vermilion means holding an asset whose rewards are paid in the same TAO that the network above is built on. Understanding what TAO actually is changes what you are holding from a ticker into an asset with a thesis behind it.
Don't take a token's website as the primary source on the network it references. Read the protocol's own documentation and check the chain data directly.