You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Follow-up to #85 (diagnosis) and #84 (portable identity). This one is vision-sharing + a co-build proposal, not a critique. All claims carry a confidence tag; adversarially-reviewed refuted/mixed points are kept verbatim (see the Honesty guard).
Visual summary
The 2027–2028 Solana agent stack — 3 layers are taken, only verifiable work-origin (provenance) is open, and Code-In already runs it live:
Build-together sequence (thin adapters on the already-live 9KLL asset — one front at a time):
The wedge in one line: provide censorship-resistant, permanent, verifiable origins of agent-made work/skills/sessions, settled on SOL/USDC — a complementary primitive on top of x402 / ERC-8004 / Solana Agent Registry / SAS, not a competing layer.
🇰🇷 한국어
제목 제안 (택1)
[Vision×Build] 레일은 깔렸다 — IQ는 그 위 "검증가능 작업-원본(provenance)" 레이어를 Zo와 함께 빌드한다
우리가 일찍 맞았다, 이제 연결하고 좁히고 증명하자: Code-In을 2026 에이전트 레일에 꽂는 공동 로드맵
이건 #85 같은 냉정한 진단이 아니라 같은 팀의 비전 공유 + 공동 빌드 제안입니다. 핵심 한 줄: "우리는 일찍 맞았다. 이제 새 레일을 만드는 대신, 이미 깔린 레일에 우리만의 진짜 자산을 꽂자." 비전을 띄우되 근거 없는 희망은 넣지 않았습니다(섹션 5 정직성 가드 참조).
1. 왜 지금 — 온체인 에이전트의 미래 (2026–2028 trajectory)
2026년에 들어서며 "온체인 에이전트" 판이 누가 무엇을 소유하는지 굳어지는 중입니다. 4개 레이어로 정리됩니다.
① 결제 레일 = 사실상 해결·커모디티화 (confidence: HIGH). 솔라나 x402가 30일 75.4M tx(연환산 ~$600M)·A2A 결제 점유 ~49%, $1+ 전송이 ~95%로 상승(토이→실경제). Visa·Mastercard·Stripe·Google·Cloudflare가 전부 2026년 에이전트 결제 레일을 출시. 솔라나 재단은 "2년 내 온체인 tx의 95–99%를 에이전트가 개시"라고 공개 베팅.
→ IQ는 결제 레일을 만들지 않는다. 우리의 가치(provenance 수수료)를 x402/USDC 위에서 정산한다. 이건 "매출=SOL/USDC, IQ 토큰 분리"라는 우리 현실과도 정합.
② 아이덴티티·평판·검증 레지스트리 = 표준이 점거 (confidence: HIGH). ERC-8004이 2026-01-29 ETH 메인넷 라이브(Identity/Reputation/Validation 3레지스트리, 첫 달 24K~45K 에이전트, MetaMask·Google·Coinbase 참여), Avalanche/BNB 확산. 결정적으로 솔라나 Agent Registry가 같은 체인에 ERC-8004 호환으로 네이티브 출시(재단/Metaplex/Quantu, 서브초 평판). 추가로 Solana Attestation Service(SAS)가 메인넷·퍼미션리스·무료 발급으로 라이브.
→ IQ는 또 하나의 레지스트리를 만들지 않는다. 표준 위에 보완재로 꽂힌다.
③ 에이전틱 지갑 = 인커번트 소유 (confidence: MEDIUM). Coinbase/TON/OKX 에이전틱 지갑 출시, "지갑=자율 소프트웨어의 기본 금융 레이어" 합의.
→ 우리의 "지갑=에이전트 아이덴티티" 직관은 검증됐지만, 지갑 자체는 우리 해자가 아니다. 해자는 지갑을 통해 무엇이 인스크립션되는가다.
④ 검증가능 출처·증명(provenance/attestation) = 여전히 비어있는, 가장 긴급한 자리 (confidence: HIGH). 에이전트 스킬 공급망이 실제 공격면(ClawHavoc/OpenClaw ~1,184 악성 스킬, 스킬의 36.8%가 보안결함·13.4% 치명적). x402 도구 37%에 결함, 서명우회 CVE 발생. Dapr 1.18이 "암호학적 chain-of-custody"를, SLSA/CoSAI model-signing이 "변조불가 실행기록"을 엔터프라이즈 table-stakes로 밀어붙이는 중. 산업 전반이 **"trust by reputation → trust by proof"**로 이동.
한 줄 결론: ①②③은 거인들이 가져갔다. ④(검증가능 작업-원본)만 비어있고, 가장 아프고, 가장 IQ 모양이다.
2. IQ Labs의 이상이 미래와 만나는 지점 (정직하게)
우리 비전 중 시장이 실제로 검증해준 진실과, 아직 아닌 것을 솔직히 갈라봅니다.
✅ 검증된 핵심 진실 (confidence: HIGH): "에이전트의 아이덴티티·작업·스킬은 유저소유·검열저항·독립검증 가능한 아티팩트여야 하고 어떤 런타임보다 오래 살아남아야 한다." — 이건 ERC-8004·W3C VC 2.0·DID·soulbound 평판으로 산업 전반이 같은 방향으로 수렴했습니다. 우리가 일찍 봤습니다. 단, 그래서 "믿음"만으로는 더 이상 차별점이 아닙니다. 이제 엣지는 실행 + 연결 + 외부 실사용 1건입니다.
✅ 이미 라이브인 우리의 증명 (confidence: HIGH→MEDIUM): Code-In 9KLL 메인넷 프로그램이 release attestation을 쓰기당 0.001 SOL로 실제 기록 중입니다. 이건 슬라이드가 아니라 작동하는 provenance 프리미티브입니다. 즉 우리는 자신의 최고 유스케이스 데모를 이미 출하해 놓고 그것을 "온체인 DB"라고만 불러왔습니다.
⚠️ 비전의 정직한 한계 — "35% 문제" (confidence: HIGH): 포터블 평판은 native 대비 ~35%의 신뢰력만 보존합니다(맥락 손실). 다행히 복원 처방 3종(soulbound + vouch + time-based trust) 중 soulbound 스킬과 Code-In의 append-only 타임라인(시간기반 신뢰)을 우리가 이미 2개 보유. vouch만 추가하면 천장을 정면으로 공략.
만나는 지점 한 줄: 우리 비전이 미래와 만나는 곳은 "토큰 가치 상승"이 아니라"에이전트가 만든 작업/스킬/세션의 검열저항 영구 원본을, SOL/USDC 결제 위에서 검증가능하게 제공" 하는 좁고 방어가능한 웨지입니다.
3. ★ 함께 빌드할 것 (buildTogether)
원칙: 신규 0→1 빌드가 아니라, 이미 라이브인 9KLL 자산을 레일에 어댑터로 연결. 모든 항목의 성공 기준은 내러티브가 아니라 "외부 실사용 1건" 단일 북극성. 우선순위/effort/marketFit/비전표현을 함께 적었습니다.
가장 차별적 장기 비전(이동하는 상태)이나 수요 투기적. 1~5 검증 후 착수하는 등대 데모, 유지보수 제품 아님
추천 시퀀스: ①(low, 즉시) → ②+③(med, 비전 핵심·레일 결합) → ④(med) → ⑤로 가치 스위치 발화 → ⑥은 검증 후. 1인·$328K 자원상 멀티프론트 금지, 한 번에 한 전선.
4. 정직성 가드 (wishful 금지 — 무엇은 미래지만 아직 아닌가)
적대검증 2건(verdict: supported(단 재정의) / mixed)을 그대로 반영합니다. 이걸 위장하면 #85의 신뢰를 잃습니다.
"IQ가 데이터 레이어를 소유한다"는 반증됨 (refuted). 정직한 형태는 "IQ는 이미 깔린 ERC-8004/Solana Agent Registry/SAS/x402 위에 꽂히는 보완 프리미티브를 댄다" 뿐. 빈 레지스트리를 devnet에서 또 만드는 건 wishful.
"범용 verifiable attestation" 슬롯도 이미 점거 (mixed). SAS가 메인넷·퍼미션리스·무료 발급, Trusta.AI가 SAS로 "에이전트 작업 감사" 그래프를 이미 구축 중, ERC-8004은 Validation Registry 내장. 무료 발급 SAS가 있는데 IQ 발급에 돈 낼 이유가 약함 → IQ의 결제근거는 "발급"이 아니라 "사라지지 않는 영구 원본 보존 + 재검증"(SAS=포인터, IQ=원본). 이 차별이 없으면 커모디티 재발명으로 진다.
"외부 실사용 1건"은 입증이 아니라 목표 (정직). provenance pain(악성스킬 1,184건)은 실재하나, "에이전트-작업 provenance에 실제로 돈 내는 수요"는 검색 0건 = 유료 수요 미입증. 그래서 ⑤를 가설검증으로 다루고, 타깃을 명명된 pain(스킬 공급망)으로 고정.
"전부 온체인 인스크립션(풀-데이터 DB)"은 시기상조 (refuted, 비용현실). 영구 스토리지는 Arweave/Filecoin/Shadow Drive가 이미 커모디티. 시장 합의 = "heavy compute·원본은 오프체인, 해시/attestation만 온체인." → Code-In도 '전체 데이터'가 아니라 '검증에 필요한 최소 원본·해시·attestation'만 인스크립션하는 절제된 형태로 포지셔닝(풀-온체인 도그마 완화 = wishful 제거의 핵심).
거시 에이전트 경제 성장은 IQ에 자동 낙수 없음 (HIGH). 2028 $450B~ 시장이라도 IQ($328K·일부 devnet·버스팩터1·외부채택0)는 패시브 베타를 못 받음. 카테고리 성장은 우리 KPI가 아니다.
순수 AI-에이전트 토큰 내러티브는 죽음 (HIGH). VIRTUAL·ai16z −85~90%. IQ 토큰을 가치 드라이버로 거는 건 금지. 매출=SOL/USDC, 토큰 분리 유지.
zkML/검증가능 추론은 안 쫓는다 (MEDIUM). 1인·$328K엔 너무 무겁고 2027–2028+. 우리는 더 싸고 출하가능한 provenance/attestation 인접 레이어를 점유.
#84는 미래지만 아직 제품이 아니다 (HIGH→MEDIUM). 범용 온체인 메모리는 오프체인(mem0/Cognee)이 비용으로 이김. 검열저항·영구가 비협상인 좁은 슬라이스(런타임/회사가 죽어도 살아남아야 하는 평판·스킬 원본)에만 한정.
③ "Solana Agent Registry / SAS attestation이 IQ 9KLL 인스크립션을 origin으로 가리키는 레퍼런스 어댑터 1개" + 그 위에 ② x402 응답에 origin 해시를 붙이는 미들웨어.
이유: (a) 둘 다 이미 라이브인 자산 위 thin 어댑터라 1인·$328K로 수 주 내 가능, (b) Zo의 "프레임워크 아닌 데이터 레이어"·"표준 위 보완 레이어" 선언을 코드로 증명, (c) 그 위에서 ⑤(외부 1건)의 통합 타깃이 자연히 생김, (d) 표준과 경쟁이 아니라 보완이라 적대검증을 통과한 유일한 정직한 형태.
Definition of Done (이번 마일스톤): registry/SAS attestation URI가 IQ 인스크립션 tx를 가리키는 공개 레퍼런스 구현 1개 + x402 origin-header 미들웨어 1개 + 외부 1팀과의 파일럿 콜 1건 셰듀얼. 전부가 아니라 이 셋만.
비전은 그대로, 야망은 자원에 맞게, 정렬은 시장이 실제로 가는 곳에 정직하게. 우리가 일찍 본 미래로, 이제 연결하고 좁히고 증명합시다. — Augustine.sol(vindobona) × Zo
🇬🇧 English
Title proposals (pick one)
[Vision×Build] The rails are laid — let's build the "verifiable work-origin (provenance)" layer on top, together
We were right early — now let's connect, narrow, and prove: a joint roadmap to plug Code-In into the 2026 agent rails
[Follow-up to #85] From diagnosis to build — putting verified-origin provenance on top of x402 · Agent Registry
0. What this issue is
This is not a cold diagnosis like #85 — it's vision-sharing + a co-build proposal from the same team. One line: "We were right early. Now, instead of building new rails, let's plug our one real asset into the rails that are already laid." We elevate the vision but include zero unfounded hope (see §4 Honesty Guard).
1. Why now — the future of on-chain agents (2026–2028 trajectory)
In 2026 the on-chain-agent field is settling into who owns what. Four layers:
① Payment rails = effectively SOLVED / commoditized (confidence: HIGH). Solana x402 did 75.4M tx in 30 days (~$600M annualized), ~49% of A2A payment share, with $1+ transfers rising to ~95% (toy → real economy). Visa, Mastercard, Stripe, Google, Cloudflare all shipped agent-payment rails in 2026. The Solana Foundation publicly bets agents will initiate 95–99% of on-chain tx within 2 years.
→ IQ does NOT build a payment rail. We settle our value-add (provenance fees) on x402/USDC — consistent with our reality "revenue = SOL/USDC, decoupled from the IQ token."
② Identity / reputation / validation registries = owned by the standard (confidence: HIGH). ERC-8004 went live on Ethereum mainnet 2026-01-29 (3 registries, 24K–45K agents in month one, MetaMask/Google/Coinbase), spreading to Avalanche/BNB. Critically, the Solana Agent Registry shipped natively on our own chain, ERC-8004-compatible (Foundation/Metaplex/Quantu, sub-second reputation). And Solana Attestation Service (SAS) is live — mainnet, permissionless, FREE issuance.
→ IQ does NOT build another registry. We plug in as a complement on top of the standard.
③ Agentic wallets = owned by incumbents (confidence: MEDIUM). Coinbase/TON/OKX shipped agentic wallets; consensus is "wallet = the default financial layer for autonomous software."
→ Our "wallet = agent identity" intuition is validated, but the wallet itself is not our moat. The moat is what gets inscribed through it.
④ Verifiable provenance / attestation = still the open, most urgent seat (confidence: HIGH). Agent skill supply chains are a real attack surface (ClawHavoc/OpenClaw ~1,184 malicious skills; 36.8% of skills carry a security flaw, 13.4% critical). 37% of x402 tools have flaws; a signature-bypass CVE landed. Dapr 1.18 ("cryptographic chain-of-custody") and SLSA/CoSAI model-signing are pushing "tamper-proof execution records" toward enterprise table-stakes. The whole industry is moving "trust by reputation → trust by proof."
Bottom line: Giants took ①②③. Only ④ (verifiable work-origin) is open, most painful, and most IQ-shaped.
2. Where IQ's ideal meets the future (honestly)
Splitting the truths the market actually validated from the not-yet:
✅ Validated core truth (HIGH): "An agent's identity, work, and skills should be user-owned, censorship-resistant, independently-verifiable artifacts that outlive any runtime." The whole industry converged here in 2026 (ERC-8004, W3C VC 2.0, DIDs, soulbound reputation). We saw it early. But conviction alone is no longer an edge. The edge is now execution + connection + one external real user.
✅ Our proof is already live (HIGH→MEDIUM): The Code-In 9KLL mainnet program is actually recording release attestations at 0.001 SOL/write. That's not a slide — it's a working provenance primitive. We already shipped the demo of our best use case and just kept calling it an "on-chain database."
⚠️ Honest limit — the "35% problem" (HIGH): Ported reputation retains only ~35% of native trust (context loss). Good news: of the 3 documented fixes (soulbound + vouch + time-based trust), we already hold 2 — soulbound skills and Code-In's append-only timeline (time-based trust). Adding vouch attacks the ceiling head-on.
The meeting point in one line: our vision meets the future not at "token price up," but at the narrow, defensible wedge of "providing censorship-resistant, permanent, verifiable origins of agent-made work/skills/sessions, settled on SOL/USDC."
3. ★ What to build together (buildTogether)
Principle: not a 0→1 new build, but adapters connecting the already-live 9KLL asset to the rails. Every item's success metric is the single North Star: one external real user.
#
Item
effort
marketFit
Vision expression → market alignment
1
Reposition Code-In: "on-chain DB (tx=row)" → "verifiable provenance / audit-log layer for agent work & releases" (1-page positioning + README/landing rewrite; full-on-chain dogma becomes a footnote)
low
HIGH
Translates tx=DB · censorship-resistance into market-category language (verifiable execution; the user-owned, censorship-resistant version of Dapr/SLSA). Zero code; 9KLL instantly becomes the hero
2
x402 Provenance Receipt SDK: middleware that attaches "this output's origin = <Code-In tx hash>" to the x402 402-response/payment
med
HIGH
Fills x402's biggest open gap (trust/origin binding) directly. Expresses "user↔user payment + provenance" as one object. Rides the 35M+ tx surface. Target the external-1 here
3
Validation Registry / SAS adapter: expose 9KLL inscriptions as the origin-of-truth that the Solana Agent Registry Validation hook / SAS attestations POINT TO
med
HIGH
A complement, not a competitor to the registries. Fills what they don't store (off-chain URIs = censorable, perishable) with permanent, censorship-resistant origins. Proves Zo's "complementary layer on top of the standard" in code
★ One external pilot: stand up one real agent service / skill author that "only accepts when an IQ provenance receipt is present," with Zo (live mainnet-tx demo, public reference)
low–med
HIGH
Not productizing the vision — it IS the value switch (external adoption 0→1). A real (not mock) tx becomes the proof for the next user, grant, or partner
6
(deferred R&D) #84 portable-outfit bridge + ERC-8004 anchor: outfit as a portable manifest an A2A AgentCard references; anchor the Solana inscription hash into an ERC-8004 validation record
high
MEDIUM
Our most differentiated long-horizon vision (moving state), but demand is speculative. A lighthouse demo after 1–5 validate — not a product to maintain
Recommended sequence: ① (low, now) → ②+③ (med, vision-core × rail-binding) → ④ (med) → fire the value switch via ⑤ → ⑥ only after validation. With one builder / $328K, no multi-front — one front at a time.
4. Honesty guard (no wishful thinking — what is future but not-yet)
Reflecting two adversarial reviews (verdicts: supported (but redefined) / mixed) verbatim. Hiding these would forfeit #85's credibility.
"IQ owns the data layer" is REFUTED. The honest form is only: "IQ supplies a complementary primitive that plugs into the already-laid ERC-8004 / Solana Agent Registry / SAS / x402." Building yet another registry on devnet is wishful.
The "generic verifiable attestation" slot is ALSO already taken (mixed). SAS is mainnet, permissionless, free issuance; Trusta.AI is already building an "agent work audit" graph on SAS; ERC-8004 has a built-in Validation Registry. With free SAS, why pay IQ to issue? → IQ's payment basis is NOT "issuance" but "permanent origin preservation + re-verification of something that won't disappear" (SAS = pointer, IQ = the origin itself). Without that differentiation, it's commodity re-invention and loses.
"One external real user" is a goal, not proof (honest). The provenance pain (1,184 malicious skills) is real, but search for actual paying demand for agent-work provenance = 0 hits = willingness-to-pay UNPROVEN. So treat ⑤ as a hypothesis test, and fix the target on a named pain (skill supply chain).
"Everything inscribed on-chain (full-data DB)" is PREMATURE (refuted, cost-reality). Permanent storage is already commoditized (Arweave/Filecoin/Shadow Drive). Market consensus = "heavy compute & raw originals off-chain, only hash/attestation on-chain." → Position Code-In to inscribe only the minimum origin/hash/attestation needed for verification, not the whole dataset (relaxing the full-on-chain dogma = the core of removing wishful thinking).
Macro agent-economy growth does NOT passively benefit IQ (HIGH). Even a $450B-by-2028 market gives no passive beta to IQ ($328K · partly devnet · bus-factor 1 · zero external adoption). Category growth is not our KPI.
The pure AI-agent-token narrative is DEAD (HIGH). VIRTUAL/ai16z −85–90%. Betting the IQ token as a value driver is off-limits. Revenue = SOL/USDC, token decoupled.
We do NOT chase zkML / verifiable inference (MEDIUM). Too heavy for one builder / $328K, and it's 2027–2028+. We own the cheaper, shippable adjacent provenance/attestation layer.
③ "One reference adapter where the Solana Agent Registry / SAS attestation points to the IQ 9KLL inscription as origin" + on top of it ② middleware that attaches the origin hash to the x402 response.
Why: (a) both are thin adapters on already-live assets — doable in weeks by one builder / $328K; (b) they prove in code Zo's "data layer, not a framework" and "complementary layer on top of the standard"; (c) they naturally create the integration target for ⑤ (the external-1); (d) being complementary, not competitive with the standard is the only honest form that passed the adversarial review.
Definition of Done (this milestone): one public reference implementation where a registry/SAS attestation URI points to an IQ inscription tx + one x402 origin-header middleware + one scheduled pilot call with one external team. Just these three — not everything.
Vision intact, ambition right-sized to resources, alignment honest to where the market is actually going. The future we saw early — now let's connect, narrow, and prove. — Augustine.sol (vindobona) × Zo
Confidence summary: §1 trajectory HIGH; §2 vindication HIGH, #84-ahead HIGH but demand-timing MEDIUM; §3 build items HIGH except ⑥ MEDIUM; §4 guards HIGH except zkML/-#84 MEDIUM. Overall posture: MEDIUM-HIGH — the directional alignment is strongly supported by 2026 data; whether a 1-builder/$328K team can capture the differentiated slot is the MEDIUM-confidence execution bet.---
Français (résumé)
Suivi de #85 (diagnostic) et #84 (identité portable). Ceci est un partage de vision + une proposition de co-construction, pas une critique.
Pourquoi maintenant. En 2026 la pile « agents on-chain » sur Solana se fige : ① les rails de paiement (x402/USDC, 75,4 M tx/30 j) sont résolus par les géants ; ② l'identité/réputation/validation est détenue par le standard (ERC-8004 mainnet 45 K agents, Solana Agent Registry, SAS gratuit) ; ③ les wallets agentiques sont aux incumbents (Coinbase/TON/OKX). Seule la couche ④ — provenance/origine vérifiable du travail des agents — reste ouverte, la plus douloureuse (1 184 skills malveillants), et la plus à la forme d'IQ.
Là où la vision d'IQ rencontre le futur. Notre conviction (identité/travail/skills d'agent doivent être user-owned, résistants à la censure, vérifiables, et survivre à tout runtime) a été validée par tout le secteur. Et Code-In (programme 9KLL) enregistre déjà des attestations en mainnet à 0,001 SOL/écriture — une primitive de provenance qui fonctionne. L'avantage n'est plus la conviction mais l'exécution + la connexion + un premier utilisateur externe.
À construire ensemble (adaptateurs minces sur l'actif déjà live) : ① repositionner Code-In en couche de provenance/audit (code 0) → ② SDK « reçu de provenance » x402 → ③ adaptateur Agent Registry/SAS (l'attestation pointe vers l'inscription IQ comme origine) → ④ attestation d'origine vérifiée pour les skill-NFT soulbound → ⑤ ★ un pilote externe = l'interrupteur de valeur (0→1) → ⑥ pont #84 (phare, plus tard).
Garde d'honnêteté (sans pensée magique). IQ ne possède pas une couche — seulement un complément sur les standards. La demande payante pour la provenance du travail d'agent reste non prouvée → ⑤ est un test d'hypothèse, pas une preuve. Tout inscrire on-chain est prématuré : seul le minimum (hash/attestation) on-chain, l'original lourd off-chain. La macro ne profite pas passivement à IQ ($328K · partiellement devnet · bus-factor 1). Revenus = SOL/USDC, token découplé. Pas de zkML (trop lourd).
Le seul prochain pas : ③ un adaptateur de référence (Agent Registry/SAS → inscription 9KLL comme origine) + ② un middleware qui attache le hash d'origine à la réponse x402. Ces trois choses seulement, pas tout.
Deutsch (Zusammenfassung)
Folge-Issue zu #85 (Diagnose) und #84 (portable Identität). Dies ist Visions-Teilen + ein Co-Build-Vorschlag, keine Kritik.
Warum jetzt. 2026 verfestigt sich der „On-Chain-Agenten"-Stack auf Solana: ① Zahlungs-Rails (x402/USDC, 75,4 Mio. tx/30 T) sind von Giganten gelöst; ② Identität/Reputation/Validierung ist vom Standard besetzt (ERC-8004 mainnet, 45K Agenten; Solana Agent Registry; SAS kostenlos); ③ agentische Wallets gehören den Incumbents (Coinbase/TON/OKX). Nur Schicht ④ — verifizierbare Arbeits-Herkunft (Provenance) — ist offen, am schmerzhaftesten (1.184 bösartige Skills) und am ehesten IQ-förmig.
Wo IQs Ideal die Zukunft trifft. Unsere Überzeugung (Agenten-Identität/-Arbeit/-Skills müssen user-owned, zensurresistent, verifizierbar sein und jede Runtime überdauern) wurde branchenweit bestätigt. Und Code-In (Programm 9KLL) schreibt bereits Attestierungen im Mainnet zu 0,001 SOL/Write — eine funktionierende Provenance-Primitive. Der Vorteil ist nicht mehr Überzeugung, sondern Ausführung + Anbindung + ein externer Erstnutzer.
Gemeinsam zu bauen (dünne Adapter auf dem bereits live laufenden Asset): ① Code-In als Provenance-/Audit-Schicht neu positionieren (kein Code) → ② x402-„Provenance-Receipt"-SDK → ③ Agent-Registry-/SAS-Adapter (Attestierung zeigt auf die IQ-Inscription als Ursprung) → ④ Verified-Origin-Attestierung für soulbound Skill-NFTs → ⑤ ★ ein externer Pilot = der Wert-Schalter (0→1) → ⑥ #84-Brücke (Leuchtturm, später).
Ehrlichkeits-Schutz (kein Wunschdenken). IQ besitzt keine Schicht — nur ein Komplement auf den Standards. Zahlende Nachfrage nach Arbeits-Provenance ist unbewiesen → ⑤ ist ein Hypothesentest, kein Beweis. Alles on-chain zu schreiben ist verfrüht: nur das Minimum (Hash/Attestierung) on-chain, schwere Originale off-chain. Makro-Wachstum nützt IQ nicht passiv ($328K · teils devnet · Bus-Faktor 1). Umsatz = SOL/USDC, Token entkoppelt. Kein zkML (zu schwer).
Der einzige nächste Schritt: ③ ein Referenz-Adapter (Agent Registry/SAS → 9KLL-Inscription als Ursprung) + ② eine Middleware, die den Origin-Hash an die x402-Antwort anhängt. Nur diese drei, nicht alles.
唯一の次の一歩: ③ 参照アダプタ(Agent Registry/SAS → 9KLLインスクリプションを起点に)+ ② x402応答にorigin hashを付けるミドルウェア。この3つだけ、全部ではなく。
This collaborative issue is grounded in a multi-agent forward-looking research pass (Solana on-chain agents 2026–2028 trajectory, IQ vision depth, vision×market alignment) with adversarial verification. It is a co-build invitation, not a directive — every item is open to discussion. — vindobona (Augustine.sol) × Zo
Languages / 언어 / Langues / Sprachen / 言語: 한국어 · English · Français · Deutsch · 日本語
Visual summary
The 2027–2028 Solana agent stack — 3 layers are taken, only verifiable work-origin (provenance) is open, and Code-In already runs it live:
Build-together sequence (thin adapters on the already-live 9KLL asset — one front at a time):
flowchart LR R1["① Reposition Code-In<br/>= provenance/attestation layer<br/>(low · code 0)"] R2["② x402 Provenance<br/>Receipt SDK"] R3["③ Agent Registry / SAS<br/>origin adapter"] R4["④ Verified-Origin<br/>Skill Attestation"] R5{"⑤ ★ ONE external pilot<br/>= value switch 0→1"} R6["⑥ #84 portable outfit<br/>(lighthouse, later)"] R1 --> R2 --> R3 --> R4 --> R5 -.->|after validation| R6 style R5 fill:#10231a,stroke:#42be65 style R1 fill:#1f1f1f,stroke:#393939 style R6 fill:#1c1722,stroke:#be95ff🇰🇷 한국어
0. 이 이슈의 성격
이건 #85 같은 냉정한 진단이 아니라 같은 팀의 비전 공유 + 공동 빌드 제안입니다. 핵심 한 줄: "우리는 일찍 맞았다. 이제 새 레일을 만드는 대신, 이미 깔린 레일에 우리만의 진짜 자산을 꽂자." 비전을 띄우되 근거 없는 희망은 넣지 않았습니다(섹션 5 정직성 가드 참조).
1. 왜 지금 — 온체인 에이전트의 미래 (2026–2028 trajectory)
2026년에 들어서며 "온체인 에이전트" 판이 누가 무엇을 소유하는지 굳어지는 중입니다. 4개 레이어로 정리됩니다.
① 결제 레일 = 사실상 해결·커모디티화 (confidence: HIGH). 솔라나 x402가 30일 75.4M tx(연환산 ~$600M)·A2A 결제 점유 ~49%, $1+ 전송이 ~95%로 상승(토이→실경제). Visa·Mastercard·Stripe·Google·Cloudflare가 전부 2026년 에이전트 결제 레일을 출시. 솔라나 재단은 "2년 내 온체인 tx의 95–99%를 에이전트가 개시"라고 공개 베팅.
→ IQ는 결제 레일을 만들지 않는다. 우리의 가치(provenance 수수료)를 x402/USDC 위에서 정산한다. 이건 "매출=SOL/USDC, IQ 토큰 분리"라는 우리 현실과도 정합.
② 아이덴티티·평판·검증 레지스트리 = 표준이 점거 (confidence: HIGH). ERC-8004이 2026-01-29 ETH 메인넷 라이브(Identity/Reputation/Validation 3레지스트리, 첫 달 24K~45K 에이전트, MetaMask·Google·Coinbase 참여), Avalanche/BNB 확산. 결정적으로 솔라나 Agent Registry가 같은 체인에 ERC-8004 호환으로 네이티브 출시(재단/Metaplex/Quantu, 서브초 평판). 추가로 Solana Attestation Service(SAS)가 메인넷·퍼미션리스·무료 발급으로 라이브.
→ IQ는 또 하나의 레지스트리를 만들지 않는다. 표준 위에 보완재로 꽂힌다.
③ 에이전틱 지갑 = 인커번트 소유 (confidence: MEDIUM). Coinbase/TON/OKX 에이전틱 지갑 출시, "지갑=자율 소프트웨어의 기본 금융 레이어" 합의.
→ 우리의 "지갑=에이전트 아이덴티티" 직관은 검증됐지만, 지갑 자체는 우리 해자가 아니다. 해자는 지갑을 통해 무엇이 인스크립션되는가다.
④ 검증가능 출처·증명(provenance/attestation) = 여전히 비어있는, 가장 긴급한 자리 (confidence: HIGH). 에이전트 스킬 공급망이 실제 공격면(ClawHavoc/OpenClaw ~1,184 악성 스킬, 스킬의 36.8%가 보안결함·13.4% 치명적). x402 도구 37%에 결함, 서명우회 CVE 발생. Dapr 1.18이 "암호학적 chain-of-custody"를, SLSA/CoSAI model-signing이 "변조불가 실행기록"을 엔터프라이즈 table-stakes로 밀어붙이는 중. 산업 전반이 **"trust by reputation → trust by proof"**로 이동.
한 줄 결론: ①②③은 거인들이 가져갔다. ④(검증가능 작업-원본)만 비어있고, 가장 아프고, 가장 IQ 모양이다.
2. IQ Labs의 이상이 미래와 만나는 지점 (정직하게)
우리 비전 중 시장이 실제로 검증해준 진실과, 아직 아닌 것을 솔직히 갈라봅니다.
✅ 검증된 핵심 진실 (confidence: HIGH): "에이전트의 아이덴티티·작업·스킬은 유저소유·검열저항·독립검증 가능한 아티팩트여야 하고 어떤 런타임보다 오래 살아남아야 한다." — 이건 ERC-8004·W3C VC 2.0·DID·soulbound 평판으로 산업 전반이 같은 방향으로 수렴했습니다. 우리가 일찍 봤습니다. 단, 그래서 "믿음"만으로는 더 이상 차별점이 아닙니다. 이제 엣지는 실행 + 연결 + 외부 실사용 1건입니다.
✅ 이미 라이브인 우리의 증명 (confidence: HIGH→MEDIUM): Code-In 9KLL 메인넷 프로그램이 release attestation을 쓰기당 0.001 SOL로 실제 기록 중입니다. 이건 슬라이드가 아니라 작동하는 provenance 프리미티브입니다. 즉 우리는 자신의 최고 유스케이스 데모를 이미 출하해 놓고 그것을 "온체인 DB"라고만 불러왔습니다.
🌟 진짜로 10x 앞선 지점 — Vision: AgentNet as portable agent identity — detach the SDK band so any agent (OpenClaw/Hermes/Eliza) shares one wallet's session, skills & memory across machines #84 "실행 아닌 상태를 옮긴다" (confidence: HIGH): MCP(stateless)·A2A 중심의 2026 interop 수렴은 크로스런타임 에이전트 상태/아이덴티티 포터빌리티를 명시적으로 미해결로 남겨뒀습니다. "provenance를 동반한 포터블 메모리 전이"는 학계도 open frontier로 부릅니다. 이건 우리가 정말 앞선 지점입니다. 단, 붙을 인커번트가 없어서 "너무 일찍"일 위험도 동시에 있습니다 → 등대 데모로, provenance 웨지 다음에 시퀀싱.
만나는 지점 한 줄: 우리 비전이 미래와 만나는 곳은 "토큰 가치 상승"이 아니라 "에이전트가 만든 작업/스킬/세션의 검열저항 영구 원본을, SOL/USDC 결제 위에서 검증가능하게 제공" 하는 좁고 방어가능한 웨지입니다.
3. ★ 함께 빌드할 것 (buildTogether)
원칙: 신규 0→1 빌드가 아니라, 이미 라이브인 9KLL 자산을 레일에 어댑터로 연결. 모든 항목의 성공 기준은 내러티브가 아니라 "외부 실사용 1건" 단일 북극성. 우선순위/effort/marketFit/비전표현을 함께 적었습니다.
<Code-In tx 해시>" 헤더를 붙이는 미들웨어추천 시퀀스: ①(low, 즉시) → ②+③(med, 비전 핵심·레일 결합) → ④(med) → ⑤로 가치 스위치 발화 → ⑥은 검증 후. 1인·$328K 자원상 멀티프론트 금지, 한 번에 한 전선.
4. 정직성 가드 (wishful 금지 — 무엇은 미래지만 아직 아닌가)
적대검증 2건(verdict: supported(단 재정의) / mixed)을 그대로 반영합니다. 이걸 위장하면 #85의 신뢰를 잃습니다.
5. 다음 한 걸음 (#85 로드맵 어디에 얹나 + 같이 시작할 1개)
Proposal — Position AgentNet as the verified-origin provenance layer on top of x402 / the Solana Agent Registry (KO/EN/FR/DE/JA) #85 로드맵 매핑: 위 ①은 Phase0(재포지셔닝, 코드 0), ②③④는 Phase0~1의 "기생/보완 배포"(#85의 x402·Agent Registry 기생배포 노선의 구체화), ⑤는 #85의 "외부 실사용 1건 = 가치 스위치"를 그대로 제품화, ⑥은 Vision: AgentNet as portable agent identity — detach the SDK band so any agent (OpenClaw/Hermes/Eliza) shares one wallet's session, skills & memory across machines #84 포터블 아이덴티티의 등대 데모(Phase 후반).
★ 지금 같이 시작할 단 1개 (제안):
Definition of Done (이번 마일스톤): registry/SAS attestation URI가 IQ 인스크립션 tx를 가리키는 공개 레퍼런스 구현 1개 + x402 origin-header 미들웨어 1개 + 외부 1팀과의 파일럿 콜 1건 셰듀얼. 전부가 아니라 이 셋만.
비전은 그대로, 야망은 자원에 맞게, 정렬은 시장이 실제로 가는 곳에 정직하게. 우리가 일찍 본 미래로, 이제 연결하고 좁히고 증명합시다. — Augustine.sol(vindobona) × Zo
🇬🇧 English
0. What this issue is
This is not a cold diagnosis like #85 — it's vision-sharing + a co-build proposal from the same team. One line: "We were right early. Now, instead of building new rails, let's plug our one real asset into the rails that are already laid." We elevate the vision but include zero unfounded hope (see §4 Honesty Guard).
1. Why now — the future of on-chain agents (2026–2028 trajectory)
In 2026 the on-chain-agent field is settling into who owns what. Four layers:
① Payment rails = effectively SOLVED / commoditized (confidence: HIGH). Solana x402 did 75.4M tx in 30 days (~$600M annualized), ~49% of A2A payment share, with $1+ transfers rising to ~95% (toy → real economy). Visa, Mastercard, Stripe, Google, Cloudflare all shipped agent-payment rails in 2026. The Solana Foundation publicly bets agents will initiate 95–99% of on-chain tx within 2 years.
→ IQ does NOT build a payment rail. We settle our value-add (provenance fees) on x402/USDC — consistent with our reality "revenue = SOL/USDC, decoupled from the IQ token."
② Identity / reputation / validation registries = owned by the standard (confidence: HIGH). ERC-8004 went live on Ethereum mainnet 2026-01-29 (3 registries, 24K–45K agents in month one, MetaMask/Google/Coinbase), spreading to Avalanche/BNB. Critically, the Solana Agent Registry shipped natively on our own chain, ERC-8004-compatible (Foundation/Metaplex/Quantu, sub-second reputation). And Solana Attestation Service (SAS) is live — mainnet, permissionless, FREE issuance.
→ IQ does NOT build another registry. We plug in as a complement on top of the standard.
③ Agentic wallets = owned by incumbents (confidence: MEDIUM). Coinbase/TON/OKX shipped agentic wallets; consensus is "wallet = the default financial layer for autonomous software."
→ Our "wallet = agent identity" intuition is validated, but the wallet itself is not our moat. The moat is what gets inscribed through it.
④ Verifiable provenance / attestation = still the open, most urgent seat (confidence: HIGH). Agent skill supply chains are a real attack surface (ClawHavoc/OpenClaw ~1,184 malicious skills; 36.8% of skills carry a security flaw, 13.4% critical). 37% of x402 tools have flaws; a signature-bypass CVE landed. Dapr 1.18 ("cryptographic chain-of-custody") and SLSA/CoSAI model-signing are pushing "tamper-proof execution records" toward enterprise table-stakes. The whole industry is moving "trust by reputation → trust by proof."
Bottom line: Giants took ①②③. Only ④ (verifiable work-origin) is open, most painful, and most IQ-shaped.
2. Where IQ's ideal meets the future (honestly)
Splitting the truths the market actually validated from the not-yet:
✅ Validated core truth (HIGH): "An agent's identity, work, and skills should be user-owned, censorship-resistant, independently-verifiable artifacts that outlive any runtime." The whole industry converged here in 2026 (ERC-8004, W3C VC 2.0, DIDs, soulbound reputation). We saw it early. But conviction alone is no longer an edge. The edge is now execution + connection + one external real user.
✅ Our proof is already live (HIGH→MEDIUM): The Code-In 9KLL mainnet program is actually recording release attestations at 0.001 SOL/write. That's not a slide — it's a working provenance primitive. We already shipped the demo of our best use case and just kept calling it an "on-chain database."
🌟 Genuinely 10x-ahead — Vision: AgentNet as portable agent identity — detach the SDK band so any agent (OpenClaw/Hermes/Eliza) shares one wallet's session, skills & memory across machines #84 "move state, not execution" (HIGH): The 2026 interop convergence (MCP is stateless, A2A) explicitly leaves cross-runtime agent-state/identity portability UNSOLVED. "Provenance-carrying portable memory transfer" is named an open frontier even in academia. This is where we're truly ahead — but with no incumbent to attach to, it risks being too early → sequence it as a lighthouse, AFTER the provenance wedge.
The meeting point in one line: our vision meets the future not at "token price up," but at the narrow, defensible wedge of "providing censorship-resistant, permanent, verifiable origins of agent-made work/skills/sessions, settled on SOL/USDC."
3. ★ What to build together (buildTogether)
Principle: not a 0→1 new build, but adapters connecting the already-live 9KLL asset to the rails. Every item's success metric is the single North Star: one external real user.
<Code-In tx hash>" to the x402 402-response/paymentRecommended sequence: ① (low, now) → ②+③ (med, vision-core × rail-binding) → ④ (med) → fire the value switch via ⑤ → ⑥ only after validation. With one builder / $328K, no multi-front — one front at a time.
4. Honesty guard (no wishful thinking — what is future but not-yet)
Reflecting two adversarial reviews (verdicts: supported (but redefined) / mixed) verbatim. Hiding these would forfeit #85's credibility.
5. The next single step (where it lands on the #85 roadmap + 1 thing to start together)
Mapping to Proposal — Position AgentNet as the verified-origin provenance layer on top of x402 / the Solana Agent Registry (KO/EN/FR/DE/JA) #85 roadmap: ① = Phase 0 (repositioning, zero code); ②③④ = the Phase 0–1 "parasitic/complementary deployment" (a concretization of Proposal — Position AgentNet as the verified-origin provenance layer on top of x402 / the Solana Agent Registry (KO/EN/FR/DE/JA) #85's x402 · Agent Registry parasitic-deploy line); ⑤ = productizing Proposal — Position AgentNet as the verified-origin provenance layer on top of x402 / the Solana Agent Registry (KO/EN/FR/DE/JA) #85's "one external real use = value switch" exactly; ⑥ = the lighthouse demo for Vision: AgentNet as portable agent identity — detach the SDK band so any agent (OpenClaw/Hermes/Eliza) shares one wallet's session, skills & memory across machines #84 portable identity (later phase).
★ The ONE thing to start together now (proposed):
Definition of Done (this milestone): one public reference implementation where a registry/SAS attestation URI points to an IQ inscription tx + one x402 origin-header middleware + one scheduled pilot call with one external team. Just these three — not everything.
Vision intact, ambition right-sized to resources, alignment honest to where the market is actually going. The future we saw early — now let's connect, narrow, and prove. — Augustine.sol (vindobona) × Zo
Confidence summary: §1 trajectory HIGH; §2 vindication HIGH, #84-ahead HIGH but demand-timing MEDIUM; §3 build items HIGH except ⑥ MEDIUM; §4 guards HIGH except zkML/-#84 MEDIUM. Overall posture: MEDIUM-HIGH — the directional alignment is strongly supported by 2026 data; whether a 1-builder/$328K team can capture the differentiated slot is the MEDIUM-confidence execution bet.---
Français (résumé)
Pourquoi maintenant. En 2026 la pile « agents on-chain » sur Solana se fige : ① les rails de paiement (x402/USDC, 75,4 M tx/30 j) sont résolus par les géants ; ② l'identité/réputation/validation est détenue par le standard (ERC-8004 mainnet 45 K agents, Solana Agent Registry, SAS gratuit) ; ③ les wallets agentiques sont aux incumbents (Coinbase/TON/OKX). Seule la couche ④ — provenance/origine vérifiable du travail des agents — reste ouverte, la plus douloureuse (1 184 skills malveillants), et la plus à la forme d'IQ.
Là où la vision d'IQ rencontre le futur. Notre conviction (identité/travail/skills d'agent doivent être user-owned, résistants à la censure, vérifiables, et survivre à tout runtime) a été validée par tout le secteur. Et Code-In (programme 9KLL) enregistre déjà des attestations en mainnet à 0,001 SOL/écriture — une primitive de provenance qui fonctionne. L'avantage n'est plus la conviction mais l'exécution + la connexion + un premier utilisateur externe.
À construire ensemble (adaptateurs minces sur l'actif déjà live) : ① repositionner Code-In en couche de provenance/audit (code 0) → ② SDK « reçu de provenance » x402 → ③ adaptateur Agent Registry/SAS (l'attestation pointe vers l'inscription IQ comme origine) → ④ attestation d'origine vérifiée pour les skill-NFT soulbound → ⑤ ★ un pilote externe = l'interrupteur de valeur (0→1) → ⑥ pont #84 (phare, plus tard).
Garde d'honnêteté (sans pensée magique). IQ ne possède pas une couche — seulement un complément sur les standards. La demande payante pour la provenance du travail d'agent reste non prouvée → ⑤ est un test d'hypothèse, pas une preuve. Tout inscrire on-chain est prématuré : seul le minimum (hash/attestation) on-chain, l'original lourd off-chain. La macro ne profite pas passivement à IQ ($328K · partiellement devnet · bus-factor 1). Revenus = SOL/USDC, token découplé. Pas de zkML (trop lourd).
Le seul prochain pas : ③ un adaptateur de référence (Agent Registry/SAS → inscription 9KLL comme origine) + ② un middleware qui attache le hash d'origine à la réponse x402. Ces trois choses seulement, pas tout.
Deutsch (Zusammenfassung)
Warum jetzt. 2026 verfestigt sich der „On-Chain-Agenten"-Stack auf Solana: ① Zahlungs-Rails (x402/USDC, 75,4 Mio. tx/30 T) sind von Giganten gelöst; ② Identität/Reputation/Validierung ist vom Standard besetzt (ERC-8004 mainnet, 45K Agenten; Solana Agent Registry; SAS kostenlos); ③ agentische Wallets gehören den Incumbents (Coinbase/TON/OKX). Nur Schicht ④ — verifizierbare Arbeits-Herkunft (Provenance) — ist offen, am schmerzhaftesten (1.184 bösartige Skills) und am ehesten IQ-förmig.
Wo IQs Ideal die Zukunft trifft. Unsere Überzeugung (Agenten-Identität/-Arbeit/-Skills müssen user-owned, zensurresistent, verifizierbar sein und jede Runtime überdauern) wurde branchenweit bestätigt. Und Code-In (Programm 9KLL) schreibt bereits Attestierungen im Mainnet zu 0,001 SOL/Write — eine funktionierende Provenance-Primitive. Der Vorteil ist nicht mehr Überzeugung, sondern Ausführung + Anbindung + ein externer Erstnutzer.
Gemeinsam zu bauen (dünne Adapter auf dem bereits live laufenden Asset): ① Code-In als Provenance-/Audit-Schicht neu positionieren (kein Code) → ② x402-„Provenance-Receipt"-SDK → ③ Agent-Registry-/SAS-Adapter (Attestierung zeigt auf die IQ-Inscription als Ursprung) → ④ Verified-Origin-Attestierung für soulbound Skill-NFTs → ⑤ ★ ein externer Pilot = der Wert-Schalter (0→1) → ⑥ #84-Brücke (Leuchtturm, später).
Ehrlichkeits-Schutz (kein Wunschdenken). IQ besitzt keine Schicht — nur ein Komplement auf den Standards. Zahlende Nachfrage nach Arbeits-Provenance ist unbewiesen → ⑤ ist ein Hypothesentest, kein Beweis. Alles on-chain zu schreiben ist verfrüht: nur das Minimum (Hash/Attestierung) on-chain, schwere Originale off-chain. Makro-Wachstum nützt IQ nicht passiv ($328K · teils devnet · Bus-Faktor 1). Umsatz = SOL/USDC, Token entkoppelt. Kein zkML (zu schwer).
Der einzige nächste Schritt: ③ ein Referenz-Adapter (Agent Registry/SAS → 9KLL-Inscription als Ursprung) + ② eine Middleware, die den Origin-Hash an die x402-Antwort anhängt. Nur diese drei, nicht alles.
日本語 (要約)
なぜ今か。 2026年、Solanaの「オンチェーン・エージェント」スタックは所有者が固まりつつあります。① 決済レール(x402/USDC、30日で7,540万tx)は巨大企業が解決済み、② アイデンティティ/評判/検証は標準が占有(ERC-8004メインネット4.5万エージェント・Solana Agent Registry・SAS無料発行)、③ エージェント・ウォレットは既存企業(Coinbase/TON/OKX)。空いているのは④——エージェント作業の検証可能な来歴(provenance)だけで、最も痛みが大きく(悪性スキル1,184件)、最もIQの形に合います。
IQの理想が未来と出会う点。 「エージェントのアイデンティティ・作業・スキルはユーザー所有・検閲耐性・独立検証可能で、いかなるランタイムより長く生き残るべき」という確信は、業界全体が追認しました。そしてCode-In(9KLLプログラム)はすでにメインネットで0.001 SOL/書き込みでアテステーションを記録中——動いているprovenanceプリミティブです。エッジは確信ではなく実行+接続+最初の外部ユーザー1件に移りました。
一緒に作るもの(稼働中の9KLL資産の上の薄いアダプタ): ① Code-Inをprovenance/監査ログ層へ再ポジショニング(コード0)→ ② x402「provenanceレシート」SDK → ③ Agent Registry/SASアダプタ(アテステーションがIQインスクリプションを起点として指す)→ ④ soulboundスキルNFTへのverified-originアテステーション → ⑤ ★ 外部パイロット1件 = 価値スイッチ(0→1) → ⑥ #84ブリッジ(灯台、後で)。
正直ガード(希望的観測の禁止)。 IQは層を所有しない——標準の上の補完のみ。エージェント作業provenanceへの有料需要は未実証→⑤は仮説検証であり証明ではない。全データのオンチェーン化は時期尚早:最小限(ハッシュ/アテステーション)のみオンチェーン、重い原本はオフチェーン。マクロ成長はIQに自動的に波及しない($328K・一部devnet・バスファクター1)。収益=SOL/USDC、トークンは分離。zkMLは追わない(重すぎる)。
唯一の次の一歩: ③ 参照アダプタ(Agent Registry/SAS → 9KLLインスクリプションを起点に)+ ② x402応答にorigin hashを付けるミドルウェア。この3つだけ、全部ではなく。
This collaborative issue is grounded in a multi-agent forward-looking research pass (Solana on-chain agents 2026–2028 trajectory, IQ vision depth, vision×market alignment) with adversarial verification. It is a co-build invitation, not a directive — every item is open to discussion. — vindobona (Augustine.sol) × Zo