Roadmap
These are plans, not commitments. The phases and gates below come from the architecture paper, Neo-Blockchains: The Fold Architecture, and describe what the project intends to test before building further. Each gate can change the plan. What is built today is in What exists today.
Build the use before the network
The paper’s plan is to sell the trust layer before building the network: start with verifiable receipts and Agent Charters on existing chains, let usage grow, and build the neo-blockchain itself only once volume justifies it. The order in which it intends to offer the pieces:
| Order | Product | First users | Why in this order |
|---|---|---|---|
| 1 | Receipts: verifiable AI-output and content receipts with trust formulas | AI providers; enterprises deploying AI | Regulatory pull and fraud losses; needs no new chain |
| 2 | Charters: identity, limits and bonded accountability for agents | Companies letting agents spend money | A bounded worst case; plugs into agent payment protocols |
| 3 | Compliance-by-proof settlement | Fintechs and payment providers | Private stablecoin settlement with provable checks |
| 4 | Verifiable Clearing markets | GPU-compute and energy marketplaces | Certified-optimal allocation |
| 5 | The full neo-blockchain | Everyone above | Moves the proven pieces onto a network built for them |
Phases and gates
The plan has five phases separated by four gates. Each gate has criteria that must be met, or the plan changes:
| Gate | Proposed criteria | If missed |
|---|---|---|
| Gate 1: end of Phase 0 (month 6) | 3 signed design partners; the Assurance Algebra specified and reviewed by two outside cryptographers | License the algebra and SDK to existing chains instead of building a network |
| Gate 2: end of Phase 1 (month 18) | 100 million receipts a month; 10 paying customers; under 0.0001 USD per receipt | Stay an application-layer protocol on existing chains |
| Gate 3: end of Phase 2 (month 36) | A testnet sustains 100,000 operations per second with a World Proof at least every 30 seconds; two independent audits with no open critical findings | Delay the main network; launch with periodic proofs and lower caps |
| Gate 4: end of Phase 3 (month 48) | The first cleared market reaches steady volume with a published optimality gap under 0.1% | Keep Clearing to exact-certificate markets only |
The network build starts only after Gate 2.
What this repository has not built yet
The reference implementation is a devnet. From the implementation status, the main pieces of the architecture that are not built, or only simulated:
- Succinct proofs. World Proofs are replay proofs. A succinct,
hash-based proof stack would plug into the existing
ProofStackinterface. - A proving market, and payments to provers.
- Networking between machines, beyond what is simplified today:
sion nodeprocesses talk QUIC, tested on one machine, with peers from files; a validator that stops cannot rejoin. - An uncertified DAG (Mysticeti class) for Contention Consensus; the current protocol is a certified DAG.
- Sampled consensus committees and a value-in-flight cap at slashable stake.
- Freezing a censoring shard until it meets its obligation; today its committee is replaced.
- Personhood, confidential amounts and compliance by zero-knowledge proof.
- Adapters for ERC-8004 and C2PA (x402, AP2 mandates and ACP are built; ACP’s card payments run against a simulated acquirer and 3-D Secure issuer).
- Archival of published data, and incentives to keep it.
- A release of
sion, signed with the release key, and the agent run against the real model providers and messaging services: it has been tested with a scripted model and stand-ins for their APIs. - The mobile app. The desktop app is built; its Android and iOS builds are configured, not attempted.
Open research problems
The paper lists problems that are open for everyone, not only for this project, among them aggregating post-quantum signatures (lattice and hash-based) efficiently inside proofs. Progress on them moves the targets above; none of them is solved here.