Research · AI × crypto · June 2026
Why AI and blockchain need each other.
In brief
- AI is consolidating into a few companies that own the models, the compute, and the data. Blockchain's role is to build the open counterweight — letting users own, pay for, and govern intelligence directly.
- NEAR is our primary example: co-founded by an author of the Transformer paper, it is explicitly positioning as the chain for user-owned AI and autonomous agents.
- Bittensor (TAO) builds a market that pays for useful machine intelligence; Venice (VVV) token-gates private inference — two different shots at the same target.
- The theme is early and noisy. We treat it as a frontier conviction, sized as such, and watch usage over narrative.
The two defining technologies of the decade are colliding. Artificial intelligence is the most powerful tool ever built — and the most centralizing, concentrating capability inside a handful of firms that own the models, the GPUs, and the training data. Blockchain is the opposite instinct: open networks, user ownership, and incentives without a gatekeeper. The interesting question is not "AI or crypto," but what gets built where the two meet.
The convergence thesis
Strip away the hype and the overlap is concrete. AI needs things crypto is unusually good at providing: payments and metering for compute and inference, provenance for data and model outputs, coordination of distributed contributors, and ownership that isn't a corporate account that can be revoked. As software agents begin to transact on their own, they will need wallets, not credit cards — money that is programmable, global, and permissionless. That is the structural bet: an open economic layer for machine intelligence.
NEAR — from Transformers to user-owned AI
Our primary example is NEAR, and the reason starts with its origins. NEAR was co-founded by Illia Polosukhin, one of the authors of "Attention Is All You Need" — the 2017 paper that introduced the Transformer, the architecture underneath essentially every modern large language model. The team came out of machine learning before it built a blockchain, and it has now turned back toward AI with a specific thesis: user-owned AI.
The argument is that intelligence is becoming the interface to everything, and if that intelligence is owned entirely by a few platforms, users lose control of their data, their identity, and their economic agency all at once. NEAR's response leans on infrastructure it already built — human-readable accounts, low fees, and chain abstraction (one account acting across many chains) — as the rails for autonomous agents that can hold assets, pay for services, and act on a user's behalf without surrendering control to a platform. It is a credible attempt to make "your AI works for you, and you own it" more than a slogan.
Bittensor (TAO) — a market for intelligence
Where NEAR focuses on ownership and agents, Bittensor attacks the production side: it builds an open market that rewards useful machine intelligence. The network is split into subnets, each a continuous competition for a task — language, images, prediction, data. Miners produce outputs; validators score them; the protocol (via its Yuma consensus) turns those scores into TAO rewards. The token's monetary policy deliberately echoes Bitcoin — a capped supply with halvings — and the later dTAO upgrade gives each subnet its own market-priced token, so capital flows toward the intelligence the market values most. We cover the mechanism in full in the Bittensor deep-dive.
Venice (VVV) — paying for private inference
The third angle is access. Venice is a privacy-first AI application — it runs open-source models and does not retain user conversations — and it uses a token, VVV, to decentralize the economics of inference. Rather than a monthly subscription to a company, holders can stake VVV to earn a continuous, pro-rata claim on the network's daily inference and API capacity. It is a clean illustration of a crypto-native idea: turn access to a compute resource into an ownable, productive asset instead of a rented service — with privacy as the headline feature rather than an afterthought.
How we frame it
These three are not the same trade. NEAR is an infrastructure-and-ownership bet, Bittensor an incentive-design bet on producing intelligence, Venice an access-and-privacy bet on consuming it. Held together, they sketch the shape of an open AI stack: produce intelligence (Bittensor), own and direct it (NEAR), access it privately (Venice). For Corvoza this sits across two standing themes — the AI × crypto frontier and sovereign money & privacy — and it is exactly the kind of structural, early conviction the book is built to take. Plan, then act: theses are sized to risk, with a defined way to be wrong.
Risks
- Narrative risk — "AI + crypto" attracts capital faster than it produces durable usage; tokens can run far ahead of fundamentals.
- Centralized competition — incumbents have enormous compute and distribution advantages.
- Measuring value — incentivizing genuinely useful intelligence is unsolved and can be gamed.
- Execution & token mechanics — emissions, unlocks, and subnet dynamics can pressure price independent of progress.
Key terms
- Transformer — the neural-network architecture behind modern LLMs (NEAR co-founder co-authored its founding paper).
- Autonomous agent — software that acts and transacts on a user's behalf, increasingly with its own wallet.
- Subnet (Bittensor) — a market/competition for one kind of AI task.
- Inference — running a trained model to produce an output; the resource Venice meters with VVV.
- Chain abstraction — using one account seamlessly across many chains.
Related — the NEAR deep-dive →
Corvoza Research is for informational purposes only — not financial, legal, or tax advice, and not a recommendation to buy or sell any asset. Digital assets are volatile and may result in total loss of capital. Corvoza is operated by Centrent, part of the Trancent world.