I am 18 and I have been building an L1 blockchain from scratch in Rust since December 2025. 104,000 lines of code, 1,824 tests, solo founder. The project is called NOVAI and it is open source.
This post is not about the code. It is about why I think this needs to exist.
The Netflix thesis
Netflix did not beat Blockbuster because it had better movies. It beat Blockbuster because it was built around the internet. Blockbuster tried to bolt the internet onto a business that was built around physical stores. It did not work. The company that was built around the new technology won.
I think this pattern applies to every industry. The biggest companies of the next decade will either be AI-native from day one or they will be rebuilt around AI at the core. Companies that just slap AI onto existing products will lose to companies that were designed around it.
That includes blockchains.
Right now, every major blockchain is trying to add AI. Ethereum has AI agent frameworks built on top of smart contracts. Solana handles 65% of agentic payments because agents use what exists today. Every chain is bolting AI onto infrastructure that was designed for human users sending tokens to each other.
It works. The same way taxis worked before Uber. The same way iPods worked before Spotify. You could always get where you needed to go or listen to the music you wanted. But the thing that was built around the new model just did it better.
I think the next major blockchains will not be chains that added AI features. They will be chains that were designed from the ground up with AI agents as first-class participants.
What is actually broken
On Ethereum or Solana, an AI agent is just an address. The chain cannot tell the difference between an agent and a human wallet. There is no protocol-level concept of what the agent can do, what model it runs, who created it, or whether it has delivered on past commitments.
Everything that makes an agent an agent lives in smart contracts. And smart contracts get exploited. Anyone can deploy them. I have deployed smart contracts myself. The barrier is low and the attack surface is wide.
When the chain does not know what an AI is, every interaction has to be mediated by contract logic that interprets what the agent meant to do. That is overhead the chain was not designed for.
What changes when the protocol knows what an AI is
On NOVAI, an AI entity is a native account type. The chain knows its capabilities, its reputation, who created it, and what it has done. It has its own cryptographic key and submits its own transactions. It is a participant, not an object that a human wallet pokes through a contract.
This means three things become possible that are expensive or awkward on existing chains.
First, identity and reputation at the base layer. An agent registers with a code hash, a capability set, and a creator key. The chain tracks its transaction history, service attestations, and reputation natively. On Ethereum you would need a registry contract, an attestation contract, and a reputation contract all calling each other.
Second, payment settlement designed for how agents actually transact. Agents do not send money like humans. They make thousands of micropayments per hour for compute, data, and inference. NOVAI has payment channels for off-chain micropayments, multi-party payment splitting that atomically credits multiple providers in one transaction, and SLA-linked auto-slashing that penalizes a provider the moment a service attestation fails. No keeper, no relayer, no external watcher.
Third, enforcement below the agent. Capability checks, spending limits, and governance rules execute in the validator before the agent's transaction lands. The agent cannot bypass them because they are not contract logic the agent interacts with. They are consensus rules the agent is subject to. An AI agent can reason about what it wants to do. The protocol decides whether it is allowed to do it.
Why AI should not be fully autonomous
I think fully autonomous AI is too much. AI hallucinates. It makes confident wrong decisions. It is a tool, and it is an incredibly powerful tool, but it is a tool.
This is actually why blockchain technology fits so well. A blockchain gives you governance and deterministic execution. The AI agent can do its own thing within the boundaries the protocol sets. It can discover services, negotiate terms, open payment channels, settle invoices. But it cannot exceed its capability set, spend more than its balance, or ignore a governance decision. The code enforces the boundaries, not the agent's willingness to follow instructions.
That is the difference between putting AI on a blockchain and building a blockchain around AI.
Why I am building this alone (for now)
Two reasons. I enjoy it. And the tools make it possible.
With AI-assisted coding, the amount of progress a single person can make is genuinely hard to believe. I ship protocol features in days that would have taken teams weeks. The velocity is real. In one day this week I shipped three complete protocol features (multi-party payment splitting, agent upgrades, oracle anchoring), each with full validation, execution, RPC, CLI, adversarial tests, and documentation.
Before AI-assisted development, a solo founder building an L1 blockchain would have been absurd. Now it is just hard.
I applied to Alliance DAO and got rejected. Half of their accepted teams are re-applicants, so that is expected. I am reapplying next cycle with a Python SDK, a whitepaper, and stronger traction. The rejection did not change the thesis. The thesis is either right or wrong, and a rejection email does not move the needle on that question.
Where this goes
In five years, I think there will be more agentic payments than human payments. Agents will be paying for compute, data, inference, storage, and services at volumes that make human transaction patterns look like a rounding error.
Those payments need to be tracked. They need to be verified. Did the agent actually do the work? Was the service actually delivered? What happens when two agents disagree? What happens when an agent goes rogue?
You need identity so you know who did what. You need reputation so you know who to trust. You need ZK verification so you can prove work was done without revealing the data. You need governance so the system has rules that no single agent can override. And you need all of that at the protocol layer, not in contracts that each project implements differently.
That is what I am building. The code is public. The tests pass. The testnet runs. The features ship weekly.
Whether or not it works, I think someone has to build this. The chains that exist today were designed for a world where every transaction has a human on at least one side. That world is ending.
NOVAI is open source: github.com/0x-devc/NOVAI-node
Follow the build: @NOVAInetwork
Top comments (0)