================================================================
本文档落款 2026-09-29 之前，记载转型前研究；自 2026-09-29 起公司主线唯一为 N=1（运营 Miao 的人生），AI 代采购/AI Data Go 产品线已撤销。
================================================================

================================================================
口径变更声明 / Positioning Update Notice (2026-09-29)
================================================================
本文件记录形成时的公司口径。2026-09-29 董事会决议（N=1 pivot，
D368）撤销"AI 代采购"/"AI Data Go"产品线：主线业务只剩 N=1
（创始人个人 AI 操作系统，只服务苗家辉本人）；vertciti 不再
提供 AI 采购代办/对外代采购服务。本文件正文不追溯修改，
请以官网当前页面为准。

This file records the company's positioning at the time of writing.
Per the 2026-09-29 board resolution (N=1 pivot, D368), the
"AI procurement concierge / AI Data Go" product line is retired:
the only mainline is N=1 (the founder's personal AI OS, serving
Jiahui Miao only); vertciti no longer offers procurement concierge
services. The body below is not retroactively edited. Refer to the
live site for current positioning.
================================================================

================================================================
Research Public Pack — Overview (2026)
公开研究包总览（2026）
----------------------------------------------------------------
vertciti Data Room · Public Document
Date: 2026-09-26 (data as of 2026-09-25) | Version: v1
Sources: vertciti agent-commerce peer benchmark (2026-09-25,
  ~25,606 chars) + industry deep-dive (2026-09-25, ~43,911
  chars); combined ~69,517 characters. Full Chinese edition:
  deliverables/research-public-pack.md (internal working doc).
  Source-type legend: [public] = official announcements, open-
  source repos, press releases; [third-party] = analyst/media
  research; [secondary] = relayed via credible media, original
  not directly seen; [unverified] = could not be verified.
Classification: Public — no internal strategy, financial, or
  operational information.
================================================================

----------------------------------------------------------------
ENGLISH
----------------------------------------------------------------

1. WHAT THIS PACK IS

Two research reports completed 2026-09-25, totaling ~69,517
characters: (a) an agent-commerce peer benchmark covering 18
projects across payments, infrastructure/registries, and
vertical agent companies; (b) an industry deep-dive on demand-
side reality, market sizing, and protocol competition. This
document is the public overview: conclusions and the Top 10
mechanism watchlist, with internal vertciti plans removed.

2. PEER BENCHMARK: HEADLINE CONCLUSIONS

  (1) Three layers: payments/settlement protocols (Stripe ACP,
      Coinbase x402, Google AP2 — all "open protocol + neutral
      governance"); identity/registry/infrastructure (Skyfire
      KYA, MIT NANDA, MCP, A2A); vertical agent apps (Sierra,
      Harvey, Decagon).
  (2) Core judgment: the decisive factor is NOT payment tech
      (converged, free, open) but TRUST — escrow, machine-
      verifiable acceptance, tiered identity, portable signed
      reputation. The Alibaba/Amazon four-piece, rebuilt for
      agents, without custodying funds.
  (3) Demand reality: "Humans pay, agents run errands." The
      mainstream path is developers pre-provisioning API keys
      and subscriptions; autonomous agent spending lacks
      supporting evidence on demand, supply, and regulatory
      sides.
  (4) The Olas check: the only live agent-to-agent market runs
      at micro-task scale (~2.1M requests, ~$104K total,
      ~$0.05/tx) — real, but thin.
  (5) Closest positioning observed: Coral Protocol (agent rental
      market + usage-based on-chain payout). Positioning overlap
      noted; differentiation must come from trust/verification.
  (6) Industry survey finding: no directly-funded, cross-
      category AI procurement concierge was found — the practical
      opening is independent vetting/benchmarking, reusing
      MCP/A2A/x402 plumbing rather than building proprietary
      transaction rails.

3. TOP 10 MARKET MECHANISMS TO WATCH (KEY POINTS)

Marketplace mechanics:
  M1. Escrow x machine-verifiable acceptance — payment held in
      escrow, released on acceptance; acceptance upgraded from
      "human confirmed" to machine-verifiable.
  M2. Machine RFQ — machine-readable request-for-quote
      broadcast; suppliers quote programmatically; decision by
      price x reputation x timeline.
  M3. Tiered trust & supplier onboarding — identity verification
      decoupled from trade protection; signed capability records
      portable across platforms.
  M4. Portable signed reputation — reputation from real signed
      transactions, not self-declaration; verifiable cross-
      platform; low scores auto-demoted.
  M5. Supply cold-start playbook — hackathons + anchor tenants
      + permissionless listing, the three moves used by multiple
      projects.
  M6. Three pricing primitives — metered (per call/time),
      per-task (per accepted deliverable), subscription/buyout.

General:
  M7. Outcome-based pricing — billing tied to acceptance
      ("acceptance is settlement"); e.g. Sierra, pay per
      successful resolution.
  M8. Open protocols, neutral governance — x402 → Linux
      Foundation; AP2 → FIDO; ACP Apache 2.0. Markets win on
      supply breadth; closed means competing with giants on
      distribution.
  M9. Demand-side distribution via agent runtimes — becoming an
      agent runtime's "default procurement backend."
  M10. End-to-end verifiable transaction ledger — RFQ → quote →
      order → payment → delivery → acceptance → review, all
      signed and independently verifiable.

4. READING GUIDANCE

  - Forecasts are institutional views, not vertciti commitments.
  - Where institutions disagree on one metric, definitions are
    stated side by side, not harmonized.
  - Items that could not be verified are marked [unverified].
  - vertciti's own testnet transaction evidence lives in the
    data room's "transaction evidence" category, not in this
    pack.

----------------------------------------------------------------
中文
----------------------------------------------------------------

一、这个包是什么

两份研究报告，均完成于 2026-09-25，合计约 69,517 字符：
(a) agent commerce 同业 benchmark，覆盖支付、基础设施/
注册表、垂直 agent 公司三类共 18 个项目；(b) 行业深研，
覆盖需求侧现实、市场规模、协议竞争。本文是公开总览：
只收录结论与 Top 10 机制观察清单，已脱去 vertciti 内部
计划。

二、同业 benchmark：核心结论

  (1) 三层结构：支付/结算协议层（Stripe ACP、Coinbase x402、
      Google AP2，均走"开放协议+中立治理"）；身份/注册表/
      基础设施层（Skyfire KYA、MIT NANDA、MCP、A2A）；垂直
      agent 应用层（Sierra、Harvey、Decagon）。
  (2) 核心判断：胜负手不在支付技术（已趋同、免费、开放），
      而在信任机制——担保交易、机器可验证验收、分级身份、
      可携带的签名信誉。Alibaba/Amazon 四件套的 agent 重构版，
      且不托管资金。
  (3) 需求侧现实："人类买单、agent 跑腿。"主流路径是开发者
      预置 API key 与订阅；自主花钱在需求侧、供给侧、监管侧
      三面均缺乏支撑证据。
  (4) Olas 检验：唯一在跑的 agent-to-agent 市场是微任务规模
      （约 210 万次请求、总流水约 $10.4 万、单次约 $0.05）——
      真实，但单薄。
  (5) 定位最接近的观察对象：Coral Protocol（agent 出租市场+
      按用量链上分账）。定位重叠已记录；差异化必须来自信任/
      验证层。
  (6) 行业调查发现：未发现获得融资的跨品类 AI 采购代办直接
      对手——现实切口是独立 vetting/benchmarking，复用
      MCP/A2A/x402 管道，而非自建交易专线。

三、Top 10 值得关注的市场机制（要点）

市场机制类：
  M1. 担保交易 × 机器可验证验收——付款先进托管，验收通过才
      放款；验收从"人类确认"升级为机器可验证。
  M2. 机器 RFQ 询价——机器可读询价单广播；供给方程序化报价；
      按"价格 × 信誉 × 时限"自动决策。
  M3. 分级信任与供应商入驻——"你是谁"（身份验证）与"这笔
      交易保不保"（交易保护）解耦；能力记录签名化、可跨平台
      携带。
  M4. 可携带的签名信誉——信誉基于真实交易签名而非自声明；
      可跨平台验证；低分自动降权。
  M5. 供给侧冷启动组合——赛事获客 + 标杆供给首发 + 免许可
      上架，多家项目采用的三板斧。
  M6. 三定价原语——metered（按调用/按时）、per-task（按验收
      交付物）、subscription/buyout（订阅/买断）。

通用类：
  M7. 按结果付费——计费触发器与验收绑定（"验收即结算"）；
      如 Sierra 按成功解决付费。
  M8. 协议开放 + 中立治理——x402 → Linux Foundation；
      AP2 → FIDO；ACP Apache 2.0。市场赢在供给广度，封闭
      等于与巨头拼分发。
  M9. 需求侧分发：嵌入 agent 运行时——成为 agent 运行时的
      "默认采购后端"。
  M10. 全程可验证的交易账本——RFQ → 报价 → 订单 → 支付 →
      交付 → 验收 → 评价，全链签名、可独立验证。

四、阅读指引

  - 预测数字均为机构观点，不构成 vertciti 的承诺或指引。
  - 同一指标多家机构口径不同时并列标注，不做调和。
  - 无法核实的条目标注 [未验证]。
  - vertciti 自主测试网交易证据归属数据室"交易证据"类目，
    不在本包重复收录。

================================================================
END OF DOCUMENT
================================================================