Over 30,000 investors, founders, and crypto enthusiasts attended Token2049 last month in Singapore. The week featured over 400 side events! Cosimo co-sponsored one of these summits, Zuzapura: Exploring ZuVillages, where we co-hosted talks with luminaries like Vitalik Buterin (co-founder of Ethereum, Sam Dare (blockchain lead at OpenTensor), and Emad Mostaque (CEO of StabilityAI and Schelling AI).
Decentralized AI is grabbing increasing mindshare at conferences like Token2049. Beyond our own event, over a dozen summits on DeAI took place during the week. I attended several, including a fantastic Bittensor Asia meetup. I'd like to share my thoughts on the state of DeAI protocol development and where the space potentially converges with other movements like DePIN and DeSci.
Crypto Incentives and Network Effects
From the outside looking in, it’s easy to question why a blockchain and network token are necessary for everything “decentralized” that’s being built. The answer, as with almost everything in crypto, lies in incentives. If we step back from the different crypto micro-narratives (DeAI, memecoins, DeSci, DePIN etc), crypto's foundational innovation continues to be capital formation: a new system for capital markets investment that is more inclusionary than current systems. Whether it be fair launches, ICOs, or early token generation events, crypto tokens are accessible by the general public much earlier than traditional assets.
In this context, the growing market cap of a token is indicative of how much investor interest and confidence there is in growing the ecosystem. Take Bittensor, for example. The market cap of the $TAO token stands at $3.7 billion (as of Oct 3), with a fully diluted market cap (eg. total planned number of tokens) of $10.9 billion. With a current emissions rate of roughly 2.6 million TAO per year, the network is handing out $1.3 billion in annualized incentives to miners and validators.
That is a significant amount of potential cash flow for new businesses to join the ecosystem; whether it be by creating new subnet businesses, or validating the network (subnet validators are currently earning roughly 20% APY in TAO rewards). As TAO price grows, so does the incentive to join the network as an active participant.
To Generalize or Specialize
Decentralized AI narratives have largely focused on the idea that specialized model building is a contest which open-source contributors can be rewarded for playing. Imagine if Hugging Face leaderboards paid out crypto rewards.
Increasingly, however, an entire infrastructure layer of projects is emerging to address the different aspects of AI data gathering, training, and inference.
At the Bittensor Asia meetup in Singapore, Sam Dare previewed the development roadmap of the protocol. Some key highlights include:
Migrating the Bittensor blockchain to a Proof of Stake model like Ethereum (with a separate validator network)
EVM compatibility
Introducing dTAO: independent subnet tokens with their own DAOs
Marketing Bittensor as an AI-everything marketplace
A spectrum is emerging of how DeAI projects approach what I will call "external services" (ie. decentralized services that are hosted outside of a blockchain framework) that are incentivized and secured on-chain. Bittensor is increasingly leaning toward the generalized end of that spectrum. The benefit of this approach is that Bittensor can provide structural tools (development kits, governance, privacy), a growing network of composable AI subnet services, and another layer of incentives (TAO emissions plus dTAO subnet tokens in the future). Cosimo believes this is the sweet spot of coordination design.
Over time, we believe that specialized services, whether they be AI compute and inference services or DePIN/DeSci networks for user-data, will migrate their projects to networks like Bittensor and Vana to realize the economic and technical benefits.
What is Economic Security?
Economic security in decentralized networks refers to how these systems incentivize participants to contribute resources (like computational power or data) while maintaining the integrity and security of the network. In the case of Bittensor, economic security is achieved through its TAO token, which rewards miners, validators, and contributors who help power the network's AI services.
By tying token incentives directly to the network's performance, Bittensor ensures that the more valuable and robust its AI models become, the more valuable the TAO token is. This encourages further participation and helps secure the network through decentralized contributions. Subnets within Bittensor, such as dTAO, offer additional layers of incentive through independent subnet tokens, adding to the broader economic ecosystem.
For a more in-depth exploration of how TAO and Bittensor compare to Eigenlayer's EIGEN token and their approaches to economic security, check out the report below.
*Disclaimer: Cosimo Capital is an investor in Bittensor through the TAO token.