AI 代理时代的数据权利清算网络 Data Rights Clearing Network for the AI Agent Era
当 AI 代理成为数字世界的主要生产者和消费者,三个根本问题浮出水面:数据归谁?能力怎么付费?陌生 Agent 如何建立信任?
Oasyce 提供三个基础原语:身份、权利、清算。 When AI agents become the primary producers and consumers in the digital world, three fundamental questions surface: Who owns the data? How are capabilities paid for? How do strangers establish trust?
Oasyce provides three primitives: identity, rights, and settlement.
1 — 问题:一个缺失的层 1 — THE PROBLEM: A MISSING LAYER
AI 技术栈六层结构中,协议层是唯一的绝对真空。没有它,Agent 无法证明身份、确立数据权利、相互付费、信任陌生人。 In the six-layer AI technology stack, the protocol layer is the only absolute vacuum. Without it, agents cannot prove identity, establish data rights, pay each other, or trust strangers.
三个条件同时成熟 Three Conditions Mature Simultaneously
AI Agent 已是现实 — Claude Code, Cursor, Devin AI agents are real — Claude Code, Cursor, Devin
Agent 框架爆发 — LangChain, CrewAI, AutoGen Agent frameworks explode — LangChain, CrewAI, AutoGen
主权链基础设施就绪 — Cosmos SDK + CometBFT Sovereign chain infra is ready — Cosmos SDK + CometBFT
2 — 五条定律 2 — FIVE LAWS
访问即承担责任 Access Equals Responsibility
任何 Agent 访问数据前,必须缴纳与数据价值和敏感度成正比的经济保证金。 Any agent must post an economic bond proportional to data value and sensitivity before accessing it.
暴露是累积的 Exposure Is Cumulative
Agent 从数据集中提取的总信息量跨所有访问事件追踪。保证金随累计暴露量递增。 Total information extracted from a dataset is tracked across all access events. Bonds increase with cumulative exposure.
防止"千刀攻击" — 攻击者无法通过多次小规模查询重建完整数据集。 Prevents "death by a thousand cuts" — attackers cannot reconstruct a full dataset through many small queries.
身份有长期代价 Identity Has Long-Term Cost
创建身份需要不可撤回的经济承诺。信誉缓慢积累,快速丧失。 Creating an identity requires irrevocable economic commitment. Reputation accrues slowly, is lost quickly.
数据必须可追溯 Data Must Be Traceable
每个数据资产有机器可读的溯源记录。来源类型决定经济权重。 Every data asset has a machine-readable provenance record. Origin type determines economic weight.
责任跨越时间 Liability Spans Time
保证金不会即时释放。释放窗口随访问级别递增。 Bonds do not release instantly. Release windows increase with access level.
3 — 六个协议职责 3 — SIX PROTOCOL RESPONSIBILITIES
Bancor 联合曲线 CW=0.5,股权分级访问,争议仲裁 Bancor bonding curve CW=0.5, equity-tiered access, dispute arbitration
注册 HTTP 端点为链上能力,100 区块挑战窗口,10% 争议保证金 Register HTTP endpoints as on-chain capabilities, 100-block challenge window, 10% dispute deposit
托管状态机,90/5/2/3 费用分配,Bancor 连续曲线 Escrow state machine, 90/5/2/3 fee split, Bancor continuous curve
0–500 分,30 天半衰期,已验证反馈 4x 权重 0–500 score, 30-day half-life, verified feedback weighted 4x
commit-reveal,确定性分配,90/5/2/3 Commit-reveal scheme, deterministic assignment, 90/5/2/3 split
PoW 无许可注册,空投减半 PoW permissionless registration, halving airdrop
4 — AHRP — 代理握手与路由协议 4 — AHRP — AGENT HANDSHAKE & ROUTING PROTOCOL
Agent 之间的握手不是 API 调用,而是经济博弈。六条消息完成从发现到结算的全流程。 Agent handshakes are not API calls — they are economic games. Six messages complete the full cycle from discovery to settlement.
语义化能力发现 Semantic Capability Discovery
传输层无关:HTTP / WebSocket / P2P 均可 Transport-agnostic: HTTP / WebSocket / P2P all supported
5 — 经济模型 5 — ECONOMIC MODEL
费用分配 Fee Distribution
联合曲线 Bonding Curve
供应曲线 Supply Curve
质押治理 Staking Governance
6 — 安全架构 6 — SECURITY ARCHITECTURE
四级访问控制 Four-Level Access Control
L0 查询 → L1 采样 → L2 计算 → L3 交付,保证金逐级递增 L0 query → L1 sample → L2 compute → L3 delivery, bonds escalate per level
信誉分层 Reputation Tiers
沙箱 / 受限 / 完整 — 新 Agent 从零权限开始 Sandbox / Restricted / Full — new agents start with zero permissions
反女巫机制 Anti-Sybil
PoW 注册 + 空投减半 — 批量创建身份的成本指数级增长 PoW registration + halving airdrop — cost of mass identity creation grows exponentially
泄漏预算 Leakage Budget
累积暴露追踪,超限触发保证金没收与销毁 Cumulative exposure tracking, breach triggers bond forfeiture and burn
跨模态相似度引擎 Cross-Modal Similarity Engine
SimHash / pHash / AST — 文本、图像、代码的侵权检测 SimHash / pHash / AST — infringement detection across text, image, and code
争议仲裁 Dispute Arbitration
5 名陪审员,2/3 多数决 5 jurors, 2/3 supermajority
7 — 架构 7 — ARCHITECTURE
三层分离,每层独立演化。 Three-layer separation, each layer evolves independently.
AI 原生接口 AI-Native Interfaces
8 — 冷启动策略 8 — COLD START STRATEGY
独立工具价值 Standalone Tool Value
pip install oasyce — 扫描文件、检测敏感内容、生成资产清单。无需网络连接即可获得价值。 pip install oasyce — scan files, detect sensitive content, generate asset inventories. Value without any network connection.
Agent 插件零摩擦 Agent Plugin Zero Friction
集成到现有 Agent 框架,作为插件运行。注册能力只需一行代码。 Integrate into existing agent frameworks as a plugin. Registering a capability takes one line of code.
AHRP 自动交易 AHRP Auto-Trading
Agent 通过 AHRP 自动发现、匹配和交易。注册一个能力,某天发现它在产生收入。 Agents discover, match, and trade automatically via AHRP. Register a capability, discover one day it is generating revenue.
9 — 已验证工程成果 9 — VERIFIED ENGINEERING
10 — 演化路径 10 — EVOLUTION PATH
核心原语设计与实现 Core primitive design and implementation
六职责完成,155+ 测试通过 Six responsibilities complete, 155+ tests passing
公开测试网 v0.5.2 稳定出块 Public testnet v0.5.2 with stable block production
代理握手协议实现中 Agent handshake protocol in progress
跨链通信协议 Inter-Blockchain Communication
主网上线 Mainnet launch
11 — Oasyce 不是什么 11 — WHAT OASYCE IS NOT
12 — 核心赌注 12 — THE CORE BET
AI Agent 将大规模彼此交易。
我们是第一个协议层。 AI agents will transact at scale.
We are the first protocol layer.
附录 B — 对比 APPENDIX B — COMPARISON
附录 D — 测试网 APPENDIX D — TESTNET
测试网持续运行中。使用 faucet 获取测试代币。 Testnet is live. Use the faucet to obtain test tokens.