Bittensor is the latest rabbit hole I’ve been diving into. And it’s been a fascinating one that I wanted to bring to your attention.
The ELI5 is that Bittensor is a Peer-to-Peer intelligence market, a decentralized network that incentivizes AI models (esp LLMs) that accomplish various tasks, including text generation, image generation, music generation, and more. There are over 23 subnets today that specialize in various tasks and anyone can spin up a new subnet.
Think of it as a decentralized chatGPT + Midjourney + everything AI models can do, which is not controlled or gated by a single central entity but rather by all of the network participants.
Miners: develop and host AI models (and earn rewards based on their performance)
Validators: evaluate outputs from Miners and rank their performance (and earn rewards)
Users submit tasks to Validators
It would be out of the scope of this post to dive into the technical intricacies of the network but I’d like to point out several detailed readings I found helpful:
TAO is the utility token for the network, and it has a tokenomic structure similar to Bitcoin: a hard cap of 21M tokens and a fair launch with no VC allocation. There is 5.65M TAO in circulation today, and all of it was earned by mining and validation on the network. It currently has a circulating market cap of $960M. Currently, 7,200 new TAO are issued every day to miners and validators.
The main way to obtain TAO is either from a CEX (MEXC) or to buy the wrapped version (wTAO) which currently has a liquidity pool on Uniswap and then bridge it over to the Bittensor network via taobridge.io
Token holders can stake TAO with one of the existing validators to earn staking rewards, which is ~17.9% APR (based on taostats.io)
Some early thoughts:
Bittensor is in its early days and it has teething issues to work through. While it has a cultish community, the actual number of users is still quite low. There are just about 50,000+ accounts on the network, about 40 active validators and 990+ active miners.
The idea of a decentralized AI network is fascinating. Besides reducing centralization risks, could the novel economic incentives lead to the creation of better and faster AI models than those from well-funded corporations like OpenAI and Google?
I personally think AI x Crypto will be highly complementary, and Bittensor is one such early experiment that I’m paying close attention to. Will continue to update as I observe and learn.