Bittensor was founded on the idea of taking Bitcoin's model for creating a digital commodity (via proof-of-work) and applying it to artificial intelligence, the 21st century's most important computational problem, moving beyond simple hashing.
Moving beyond Bitcoin's simple true/false hash validation, Bittensor invented a consensus mechanism to measure complex, high-dimensional outputs, like the informational significance of an AI model's response. This allows the network to reward valuable, subjective work.
The success of protocols like Bitcoin stems from their permissionless nature. Anyone, anywhere can contribute, leading to a hyper-competitive market that maximizes efficiency by tapping into previously unused resources like excess energy or stranded compute.
Each subnet operates like a competitive sports league where participants (miners) are ranked. Top performers get paid more, while those at the bottom are cycled out. This constant pressure ensures only the most efficient contributors remain, fostering hyper-competition.
Instead of relying on legal service level agreements (SLAs), Bittensor subnets use validators that programmatically check miner performance. They can use ZK proofs to verify an AI model's output, ensuring quality and honesty without trust.
The design philosophy for a Bittensor subnet is to create a permissionless, adversarial game so well-structured that even the most malicious actor imaginable is forced to produce value to participate. Intent is irrelevant; only verified output matters.
Because Bittensor is a permissionless, global, and economically valuable network, subnet creators aren't just building a startup; they are entering an arena with highly sophisticated adversaries, including potential nation-state actors, trying to exploit their system from day one.
Bittensor applies its own principles recursively: the main blockchain is a permissionless mechanism that selects and rewards the most effective *other* mechanisms (the subnets). This creates a competitive market for building valuable systems, turning innovation into a mineable commodity.
Affine's goal isn't just to build one smart model, but to create a market for "reasoning." It rewards AI models based on their ability to improve the performance of other models, indirectly commoditizing the process of thought itself.
To create a new subnet, founders pay for a limited slot via a Dutch auction. The TAO paid isn't burned or taken by a central party; it's used to create the initial liquidity pool between TAO and the new subnet's token, bootstrapping its market.
To collaboratively train a large AI model, distributed computers must constantly merge their parameters. The sheer size of these models (terabytes of data) makes this synchronization across standard internet connections prohibitively slow, creating a fundamental bandwidth bottleneck.
While founders can abandon traditional startups, they can't easily sell their illiquid shares. In crypto, permissionless secondary markets allow a malicious founder to instantly sell project tokens (a "rug pull"), extracting all value before investors can react.
As a response to bad actors, Bittensor developed a system where subnet teams can voluntarily lock their tokens. This lock is a public, on-chain signal of their "conviction." If they unlock to sell, it becomes a transparent event, alerting investors.