← Hub★ Fused v1◆ Pitch v2◆ Pitch v3✦ v2 YC-tested▤ LadderDraft ADraft BDraft Craw .md

ENTITY.ID — SERIES B BRANDBOOK

2026-07-10 · text edition v1

One brand, two registers. Investor voice unless marked [site]. No exclamation marks, ever.


Executive summary

  1. The claim: every jurisdiction has a company registry; the internet doesn't. Entity.ID is building it — the internet's entity registry.
  2. The definition: an entity is anything with tangible impact and distinct, perpetual existence — companies, handshake teams, and AI agents alike.
  3. The map (②): three classes — proto-entities (the proven demand and the wedge), existing companies (the endgame), AI agents (the emergency) — one registry entry serves all three.
  4. The wedge (③): venture formation — handshake to recognized, investable, operational venture in minutes, free; operate today, incorporate when you need to.
  5. The second act (④): agents can pay but can't be anyone; KYA — Know Your Agent — confers standing anchored to accountable natural persons, and the AI-personhood bans preserve exactly our route.
  6. The moat: the record is public, tamper-proof, agent-inclusive, and internet-native; the technical layer lives in one bounded, flagged passage (T4), never in a headline.
  7. Voice (⑤): one brand, two registers — site voice locked and crypto-free, investor voice protocol-literate — never blended mid-sentence.
  8. The kit (⑥): heroes, taglines, one-liners, blurbs, boilerplates, and spoken pitches in both registers; every picked asset verbatim; lengths enforced and trims marked.
  9. The deck (⑦): twelve moves from the missing institution to "fund the registrar"; the headline figure canon appears exactly once, on slide 7.
  10. The evidence (⑧): cited, liftable proof points with bear cases carried openly and internal estimates labeled — plus one derivation task that must close before print.

① The Thesis

[investor register · civic temperature]

Every jurisdiction has a company registry. The internet doesn't.

Civilization keeps its promises in public records. The deed that says the land is yours; the register that says the ship may sail; the filing that says the company exists. These are the reason strangers can do business at all. Every functioning economy, in every era, has built one — and the internet never did. So its ventures are born as rumors: ownership folklore, rules unwritten, history unverifiable. Its newest economic actors, AI agents, transact with no record at all.

An entity, properly defined, is anything with tangible impact and distinct, perpetual existence. Companies qualify. So does the two-founder team with paying customers and no paperwork, the collective with a treasury, the agent with an operator and a budget. The definition is old; only the registrar is missing.

On the internet, existence should be a record anyone can check. Trust in private markets currently terminates in a vendor's database or a paper filing. Entity.ID moves the trust anchor from the vendor to the record itself. Registration is the product; standing is the feature.

For internet-native ventures and AI agents that need to be trusted counterparties, Entity.ID is the public entity registry where they gain verifiable identity, disclosed ownership, and good standing anyone can check — and operate from day one with treasury, fundraising, and governance built in. Unlike incorporation platforms and cap-table vendors, whose records begin only after a state filing, live in private databases, and admit only humans, Entity.ID's record is public, tamper-proof, agent-inclusive, and native to the internet. Traditional incorporation is an option, never a prerequisite.

What we are building is therefore an institution, not an app. Registries earned that status the hard way: asset partitioning — the foundation Hansmann and Kraakman identified beneath the modern firm — is "the only basic attribute of a firm that could not feasibly be established by contractual means alone" (Hansmann & Kraakman, The Essential Role of Organizational Law, 110 Yale L.J. 387, 393 (2000)) — and organizational law delivers it through one instrument: the registry. The registry entry — who's in it, who owns what, who decides, where the money sits — is the protocol's atomic unit, the venture product, and the agent's standing. One record format, every class of entity.

The technical moat, in one paragraph. The register is kept on-chain: Entity.ID is built on the RegistryChain protocol, which anchors every record as an open-standard primitive. Entries append; they never overwrite. Operational terms execute deterministically. If Entity.ID the company disappeared tomorrow, the records would still exist and still verify — resolvable today by 5,000+ software libraries and recognized in 750+ applications. This is the property legacy registries cannot retrofit: the record enforces itself. The record cannot be rewritten or switched off. That is the difference between a database and an institution. [FLAG: T4 — this is the one permitted on-chain/moat passage; Husam to dial depth up or down.]


② The Entity Map

[investor register · Swiss temperature]

A two-founder startup with paying customers, a Stripe account in one founder's personal name, and a 50/50 split that exists only as a story they tell each other. A personal AI agent — one operator, an always-on server, real money spent under a bot handle nobody can trace to a person. Each is an entity by the operating definition — tangible impact, distinct and perpetual existence. Neither appears in any registry on earth.

The registry's universe divides into three classes. Each is counted, each is underserved in a different way, and each is served by the same registry entry.

Class Count Condition today What registration confers Institutional status
Proto-entities (define it once: teams already operating on trust, before any filing) ~1.2 billion (founder estimate) Real work, no record: handshake equity, money in personal accounts A recognized, investable, operational venture — before any state filing Served by no incumbent — every incumbent's record begins at the state filing
Existing companies ~400 million (founder estimate) Incorporated, but running on PDFs and interpretation Legally-binding agreements become computationally-binding Registered on paper-era rails; records live in vendors' databases
AI agents Uncounted — the class is too young for external figures Economically active, legally nonexistent Standing, anchored to an accountable natural person (KYA — Know Your Agent) No path to standing anywhere today

For public backup, registrychain.com and the current deck publish a consistent internal set — 1B+ pre-formal ventures, 359M+ incorporated entities, 70M+ autonomous economic actors — cited as the company's own figures and never mixed into the same visual as the canon above.

Class one: proto-entities — the proven demand. Roughly 1.2 billion of these teams operate today, and no incumbent can serve them, because every incumbent's record begins at the state filing. What they need is the four facts on the record: who's in it, who owns what, who decides, where the money sits.

Class two: existing companies — the endgame. Some 400 million incorporated entities already keep the four facts — in filing cabinets. Their agreements bind legally but not operationally: enforcement runs through interpretation, intermediaries, and memory. On the registry, a vesting clause — "founder shares vest over four years with a one-year cliff" — becomes execution logic: IF serviceTime ≥ 1yr AND < 4yrs THEN vested = totalGranted × (serviceTime/4yrs). The clause is enforced by construction. This is the public-record tradition extended to executable form: the register does not merely attest what was agreed; it carries it out.

Class three: AI agents — the emergency. The demand curve is documented even where the population is not: Gartner projects 90% of B2B buying mediated by AI agents by 2028 — $15T of spend through agent exchanges (Digital Commerce 360, Nov 2025); McKinsey sizes agentic commerce at $3–5T globally by 2030; the AI-agents software market is forecast to grow from ~$7.84B in 2025 to ~$52.6B by 2030 (MarketsandMarkets). Every dollar of it moves through actors no registry will admit. Chapter ④ takes this class in full.

Three classes, one register, one entry format. The map is the market; the sequence is the strategy — and the sequence is the next chapter.


③ Now: The Wedge

[both registers — investor prose, site sidebar · warm temperature]

Founders don't live paperwork-then-work. They live work, then a scramble. Somewhere out there right now is a team like Tarka Labs — three people, paying clients, shared docs at v14_FINAL, money in one founder's personal account, equity they'll "figure out later." That team is already a company. It just isn't official.

Our first product makes it official, in the order that actually makes sense. The insight underneath: incorporation is a feature, not the front door. A venture should boot like software — form in minutes, operate immediately, incorporate when, if, you need to. The incumbent calendar reads file → wait → exist → operate. Ours reads register → operate, and it starts before lunch.

Here is what shipping looks like. On app.entity.id, a seven-step wizard takes about ten minutes and costs nothing: name the venture (the name becomes a public address — tarka-labs.venture.public.entity.id), add partners (each one KYC-verified, each one signing), record the ownership split, agree the terms (arbitration ships in the defaults — a dispute path agreed before it's needed), adopt a constitution, submit. The moment every partner signs, the venture is live: a named entity with verified members, agreed ownership, working rules, and its own treasury, on a public tamper-proof record.

And formed means operational. The treasury runs on threshold approvals, so "who approved this and when" is always answerable. Fundraising issues post-money SAFEs in-product, funds settling to the venture's treasury while the cap table updates itself. Governance records proposals, votes, and amendments. The compliance tab holds exactly the beneficial-ownership package banks ask for.

Along the way, the product forces the conversation everyone avoids. We think handshake equity kills more startups than competitors do; Entity.ID makes the split a recorded fact on day one instead of a fight in month eight. And when month eight brings a lead investor who wants a C-Corp, incorporation is an upgrade, not a rebuild: same members, same cap table, every recorded decision and the full treasury history walk in on day one. The registration becomes an event in the venture's history — the public address stays live. Honesty matters here, and the product is honest: until incorporation there is no liability shield and no tax personhood, and most banks still want a registered company. What there is, is everything incorporation will need — already recorded, already verified.

Three pillars carry the wedge:

  1. On the record instantly. A named entity with a public address, members, ownership, and rules — minutes after you start, visible to anyone. Not a placeholder: a registry entry.
  2. Verifiable by anyone — human or machine. Tamper-proof records of who's who, who owns what, and who decides. Banks, investors, counterparties, and other agents check the registry, not your word.
  3. Operating from day one. Treasury, fundraising, governance, members, and compliance in one dashboard, every module writing to the same record — and incorporation is a step you take when you're ready, not a toll you pay to exist.

Why this wedge. Of the three classes on the map, proto-entities are (a) the largest count; (b) the least served — no incumbent can serve them, because every incumbent's record begins at the state filing, and Carta Launch literally requires a certificate of incorporation to onboard; (c) the cheapest to acquire — formation is free and takes minutes, so the funnel starts at the moment of intent, not the moment of paperwork; and (d) the seed crop for the other two classes — today's handshake is tomorrow's company, and agents join ventures as members. The wedge is not a small first product. It is the registry acquiring its entries in formation order — the same way every registry of record began, with the entries nobody else would take. The registry entry — four facts, recorded, checkable — is the protocol's atomic unit, is the venture product, is the agent's standing. Same record at every altitude.

How the site says it [site register] Start your venture today. Incorporate when you're ready. Too early to incorporate. Too real to stay a handshake. Your venture is already real — Entity.ID makes it official: a named venture with verified members, agreed ownership, working rules, and a treasury, on a tamper-proof public registry. Operate today, incorporate when you need to. The category: the registry where ventures begin. CTA: Form your venture · Form yours — free


④ The Agent Chapter

[investor register · Swiss, one civic flourish]

In the last eighteen months, Visa, Mastercard, Stripe, and Google all shipped payment rails for AI agents. Agents can pay — but they can't be anyone: no identity, no owners, no one to serve or sue. Every rail assumes an accountable party behind the agent; none provides it.

The failures are already on the record. Anthropic's Project Vend agent invented a payment account that didn't exist and told customers to pay into it. Cloudflare caught Perplexity's crawlers rotating identities across tens of thousands of domains — the web cannot tell whose agent is knocking. In Moffatt v. Air Canada (2024), the airline argued its chatbot was "a separate legal entity responsible for its own actions" — and lost, because no such entity existed. Entity.ID gives that sentence a referent. Standing here means a checkable registry status with a disclosed accountability chain and an agreed recourse path — not legal personhood.

The industry has named the gap. American Banker: beyond KYC, banks now need to "know your agent." FIS shipped an issuer-facing KYA product (January 2026); Skyfire raised $9.5M, Catena Labs a $30M Series A with an OCC trust-charter application; Visa and Mastercard each launched agent-token primitives. Note what all of them do: chain the agent back to a pre-existing human-owned entity. None makes the agent a party. Google's protocol proves a human authorized a purchase; Cloudflare's signatures prove which operator sent a request; enterprise IAM proves an agent has a login. Nobody proves the agent is someone — with owners, a treasury, rules, and a jurisdiction.

This is the class no legacy registry can admit. Companies houses are jurisdiction-bound, paper-era, and human-only by construction; Entity.ID admits agents because its record format never depended on those constraints. And the fix is two hundred years old. Corporate personhood solved exactly this problem for capital: let a non-human actor own, contract, and answer for itself, on the condition that disclosed humans stand behind it in a public register. On that unglamorous instrument, the modern economy was built. Entity.ID compresses that 200-year institutional arc into a protocol. Not rights for AI — the EU killed "electronic persons" and withdrew its AI Liability Directive, and US states are banning direct AI personhood outright. Those bans are a moat: they foreclose every route to agent standing except the entity route, which is legal today.

KYA — Know Your Agent — is that route made checkable: each registered agent carries a persistent identifier under its operator, and the record resolves the full anchor chain — agent → venture → members and owners → a verified natural person. The protocol's rule — the Principle of Natural Attribution — is that any participant with tangible impact on society must be attributed to a natural person. This is not rights for machines. It is recourse, by construction.

The company-level answer to "why us" is structural. The agent story starts where the venture story does: with membership. An agent joins a venture the way a partner does — named, verified, scoped: a persistent address under its operator, permissions set by the venture's governance, treasury access inside voted limits, every action signed and recorded. When a venture's first hire isn't human, the registration standard is already in place. The ventures Chapter ③ forms are the accountable anchors the agent product requires — two-sided lock-in no KYA vendor can replicate, because none of them holds the entity side.

Three pillars:

The clock is not ours. EU AI Act Article 50 becomes enforceable on 2026-08-02, and the Commission's draft guidance says it covers agents. California's AB 316 (effective 2026-01-01) removed "the AI did it autonomously" as a defense. The FSB is consulting on agents as "synthetic employees" needing accountability chains; NIST has an open RFI on agent identity. When regulators ask "who answers for this agent?", Entity.ID is the only system already holding the answer.


⑤ Voice & Vocabulary

[the register system itself]

One brand, two registers, and the boundary is absolute.

Site voice governs entity.id and every public surface — a registrar speaking to a founder. Locked vocabulary, zero crypto, warm and concrete: a stranger gets the idea in five seconds. Investor voice governs this book, the deck, and the spoken pitch — a registrar speaking to a board: protocol language is allowed and useful — used tastefully, as hints by default, with full technical depth confined to the one flagged moat passage in Chapter ①. Registers are never blended mid-sentence. A page either speaks to a founder forming a venture before lunch, or to a partner underwriting a registry; a sentence that tries both serves neither. A registry that can't keep its own vocabulary straight will not be trusted to keep anyone's record.

Naming follows the same discipline. Entity.ID leads everywhere. In investor materials the technical layer is "built on the RegistryChain protocol"; on the site the layer is "the Entity.ID protocol" or "the open registry standard." RegistryChain never names the company, the product, or the thing you form. The standard explanation of what the product records is the four-facts device — who's in it, who owns what, who decides, where the money sits — and no parallel formulation should be coined.

Investors read voice as operational discipline. They are right to. The table below is the law; extend it, never contradict it.

Concept Site voice (locked) Investor voice (allowed)
The record tamper-proof public registry the registry primitive; an on-chain public record
The category the registry where ventures begin the entity registry — the internet's missing institution
Formation form ("Form your venture") day-zero entity formation; formation as protocol event
What's formed venture / entity entity; proto-entity (define on first use)
Agreements working rules; constitution; charter legally-binding → computationally-binding; deterministic execution of bylaws
The layer underneath the Entity.ID protocol; the open registry standard built on the RegistryChain protocol (technical-layer mentions only)
AI actors AI agents; verified members registered agents; the agent economy
Agent verification KYA — Know Your Agent KYA as category ownership; KYC for the agent economy at API speed
Agent status agents with standing; a party you can verify corporate personhood, replayed for agents (thesis framing only)
Accountability the accountability chain; disclosed owner chain subject-of-right natural person anchor; recourse path; insurable object
Verification anyone can check it; good standing recognized ↔ recognized transactions; standing checked before value moves
Jurisdiction / compliance compliance built in compliance-by-registration; jurisdiction as an API; EU AI Act Art. 50 native
Money treasury programmable treasury
Incorporation operate today, incorporate when you need to incorporation as a roadmap feature, not the front door
Integrity claim tamper-proof, verifiable, attested and dated cryptographically verifiable; deterministic
Trust model the record, not our word trust anchor moved from vendor database to the record itself
The wedge audience teams running on a handshake proto-entities — the ~1.2B class (founder estimate, so labeled)
The beginning a promise; a handshake a proto-entity; the pre-formal economy
The product's core the four facts: who's in it, who owns what, who decides, where the money sits the four-facts record as the registry's atomic unit
The public address a page anyone can look up a resolvable identifier on open naming standards
History attested and dated; nothing quietly rewritten append-only record; amendments as events, never overwrites
Recourse a dispute path agreed before it's needed arbitration primitive in formation terms (Kleros default)
Fundraising raise with built-in records post-money SAFEs settling to the treasury; cap table auto-updates
Ownership the split, agreed and recorded recorded cap table; splits carried intact into incorporation
Formation cost & speed minutes, free free formation; acquisition at the moment of intent
Diligence your record, one link away diligence is a URL; verifiable operating history
The incorporation event everything carries over de facto → de jure continuation; an upgrade, not a rebuild
Human verification verified members KYC-verified constituents
The counterparty problem when nobody can check unidentifiable counterparties; pre-2008 finance without an LEI
The three classes teams, companies, AI agents proto-entities · existing companies · registered agents
Traction the registry's live counter registry-published counts (the e-Residency pattern); "claimable" defined when used
The worked example Tarka Labs (an example) illustrative specimen — never traction
Regulators compliance built in the registry regulators are already asking for
Market scale (site never leads with market figures) ~1.2B proto-entities · ~400M companies as canon; site figures as cited backup only

Banned on site side, everywhere, no exceptions: blockchain, on-chain, crypto, web3, Ethereum, IPFS, smart contract, wallet (say treasury), DAO, immutable, trustless, AI personhood / AI employee / digital worker / bot, proto-entity (in hero copy), RegistryChain. Banned on investor side: rights-for-AI framing; "trustless"; inventing traction, pricing, or named customers; exclamation marks.


⑥ Message Architecture

The reusable kit. Every asset labeled [site] or [investor]; verbatim picks are used as written.

Hero systems

Hero A — [investor] the missing institution

Hero B — [site] the wedge promise

Hero C — [investor — agent-thesis]

Taglines (use verbatim; do not coin competitors when one of these fits)

One-liners (≤ 12 words)

Short blurb (60–75 words)

[investor] — Every jurisdiction has a company registry. The internet doesn't — so its ventures run on handshakes and its AI agents transact with no legal existence at all. Entity.ID is the internet's entity registry: form a venture in minutes and get a public, tamper-proof record — verifiable identity, disclosed ownership, good standing anyone can check. Ventures operate from day one; agents get standing anchored to accountable owners. Operate today, incorporate when you need to.

[site] — Entity.ID is the registry where ventures begin: start your venture today, incorporate when you're ready. A company is four facts — who's in it, who owns what, who decides, where the money sits — and Entity.ID records them in minutes, in a tamper-proof public registry, for free — no lawyers, no filing fees until you choose a jurisdiction. One dashboard runs the treasury, fundraising, governance, members, and compliance; when you incorporate, every record carries over. Begin.

Long blurb / boilerplate (~100 words)

[site / press — primary; venture-first body, agent clause spliced (T2 position)] Entity.ID is the registry where ventures begin. Founders, collectives, and agencies turn a name into a working venture in minutes: members, ownership, rules, and treasury, recorded in a tamper-proof public registry that banks, investors, and registrars can verify. From day one the venture operates through a single dashboard — treasury, fundraising, governance, members, and compliance — no lawyers, no filing fees to start. For AI agents, registration confers what no payment rail or identity credential can: standing, anchored to accountable owners. Incorporation is a step, not a starting point: members, ownership, and history carry over. Ventures run on the Entity.ID protocol, an open registry standard. Learn more at entity.id. [investor variant: the final sentence pair becomes "Built on the RegistryChain protocol, an open registry standard. Learn more at entity.id."]

[ALT — agent-thesis investors; pure v4] Entity.ID is the company layer of the agent economy. On Entity.ID, founders, teams, and AI agents form real ventures in minutes — named entities with members, ownership, rules, and a treasury, recorded on a public, tamper-proof registry — and operate immediately: treasury, fundraising, governance, and compliance from day one, incorporation whenever it's needed. For AI agents, registration confers what no payment rail or identity credential can: standing. Every registered agent carries a verifiable identity, a disclosed chain of accountable humans or companies, and a machine-readable compliance model that counterparties can check before value moves. Built on the RegistryChain protocol, an open registry standard.

30-second spoken pitch (≤ 90 words spoken)

[investor] [trimmed — the picked thirtySecond.v7 runs ~140 words verbatim; condensed to 87 to meet the ≤90-word spoken limit, sentences kept verbatim where possible] Every jurisdiction has a company registry. The internet doesn't. So its ventures run on handshakes, and its AI agents move real money with no legal existence at all. Entity.ID is the internet's entity registry. Form a venture in minutes and get a public, tamper-proof record: who owns it, what the rules are, whether it's in good standing. Banks check it. Clients check it. Other agents check it. Operations from day one; incorporation later. We're building the registry the internet economy — and its regulators — are already asking for.

[site / demo-day — the calm variant] A company is four facts. [pause] Who's in it. Who owns what. Who decides. Where the money sits. Most teams keep those facts in a group chat — until a bank, an investor, or a lawsuit asks for proof. Entity.ID records them. Seven steps. Minutes. Zero dollars. A tamper-proof public registry anyone can check. The venture operates immediately — treasury, fundraising, governance, compliance. Incorporation becomes a step, [beat] not the price of existing. Real ventures. In minutes.

Interior beats (swap-ins for either pitch): "Thirty million US businesses have zero employees." · "Every tool a startup runs on starts after the certificate of incorporation. We start before it — which means we start everything."


⑦ The Argument — twelve slides

[investor register] Titles may flex; the twelve-move sequence may not.

1 · Every jurisdiction has a company registry. The internet doesn't.

2 · Why now: three curves crossed, 2024–2026.

3 · The problem: trust terminates in a filing cabinet or a vendor's database.

4 · The insight: the fix is 200 years old.

5 · The product: formation in minutes → public record → operations from day one.

6 · The agent wedge: identity is crowded, payments a land-grab, standing is empty.

7 · The market: three classes, one registry.

8 · Traction: the registry states its own numbers.

9 · Business model: registry economics.

10 · Roadmap: registry → rails → standard.

11 · Team: registries are won on credibility.

12 · The ask.


⑧ Ammunition

Cited, liftable proof points. Anything uncitable is labeled a belief or an internal figure.

  1. Record formation volume: ~5.62M US new business applications in 2025, up 8.2%; May 2026 pace ~6.3M annualized — 450–520k per month for five straight years. (US Census, Business Formation Statistics)
  2. The unofficial economy at scale: 30.4M US nonemployer businesses with $1.8T in receipts (Census 2023, released 2025); 72.9M US independent workers (MBO Partners 2025).
  3. Formation demand is accelerating, not saturating: Stripe Atlas incorporations grew +130% YoY in Q1 2026; 100k+ founders across 140+ countries. (Stripe)
  4. The incumbent benchmark: online legal services ~$25.2B (2025) → ~$47B by 2030; LegalZoom FY2025 revenue $756.0M in a ~4%-CAGR market. (SEC filing; market reports)
  5. A registry is a business: Delaware — 2.28M+ entities, 66.7% of the Fortune 500, 334,461 new entities in 2025; franchise fees ≈ 25–30% of the state General Fund, ~$1.8–1.9B/yr. (state records)
  6. A digital registry earns real revenue: Estonia e-Residency — 132,000+ e-residents, 5,556 companies formed in 2025, €125M state revenue in 2025 — and it earns trust by publishing its own live counters. (e-resident.gov.ee / ERR)
  7. Precedent for mandated identity after a trust crisis: post-2008 the G20/FSB created the LEI; 3M+ entities now carry one. Agent commerce is pre-2008 finance — enormous flows, unidentifiable counterparties. (GLEIF)
  8. One accountability function is already a $2B industry: registered-agent services. (market reports)
  9. Agents can pay: every major payment network shipped agent-payment rails within eighteen months — Visa Intelligent Commerce and Mastercard Agent Pay (Apr 2025), OpenAI/Stripe ACP and Google AP2 (Sep 2025), Visa TAP (Oct 2025), PayPal, Coinbase x402. They still can't legally exist. (company announcements; research memo 2026-07-07)
  10. The demand curve: Gartner — by 2028, 90% of B2B buying mediated by AI agents, $15T of spend through agent exchanges (Digital Commerce 360, Nov 2025); McKinsey — agentic commerce $3–5T globally by 2030; AI-agents market ~$7.84B (2025) → ~$52.6B by 2030, 46.3% CAGR (MarketsandMarkets).
  11. Agents already move retail revenue: Salesforce — AI and agents influenced 20% of global online orders ($67B) during Cyber Week 2025 and $262B across the holiday season; Adobe — GenAI-referred retail traffic +1,200% YoY, converting 31% higher. (Salesforce; Adobe)
  12. The regulatory calendar is set: EU AI Act Art. 50 enforceable 2026-08-02, explicitly covering agents per Commission draft guidance (May 2026); California AB 316 (eff. 2026-01-01) removes "the AI did it autonomously" as a defense; FINRA's 2026 report carries its first agentic-AI risk section; FSB consultation on agents as "synthetic employees" (Jun 2026); NIST RFI on agent identity (Mar 2026). (official texts and filings)
  13. The failure canon: Anthropic's Project Vend agent invented a payment account that didn't exist and gave product away under social engineering (Anthropic / WSJ); Cloudflare caught Perplexity crawlers rotating identities across tens of thousands of domains, Aug 2025 (Cloudflare); in Moffatt v. Air Canada (2024) the airline claimed its chatbot was "a separate legal entity" — and lost because none existed (case record).
  14. KYA is the industry's own word, still unowned: American Banker ("banks now need to know your agent"); FIS shipped issuer-facing KYA (Jan 2026); Skyfire ($9.5M, a16z CSX); Catena Labs ($30M Series A, May 2026, OCC trust-charter application). All chain agents to a pre-existing human-owned entity; none makes the agent a party. (American Banker; funding announcements)
  15. The SSL analogy for KYA: HTTPS went from 39% of the web to effectively all of it within a decade, once verification became free and the unverified got branded "Not secure." Agent verification is at its 2015. (SSL adoption history — analogy, not a proof point)
  16. Capacity was never the gap; identity was: machine-formed contracts have bound in the US since 1999 (UETA §14, 49 states + DC; E-SIGN 2000); Restatement of Agency §1.04 cmt. e denies programs principal status only "at present" — two words holding the door open. (statutes; Restatement)
  17. Corporate personhood is the structural precedent, not a metaphor: Dartmouth College v. Woodward (1819) → Santa Clara (1886); asset partitioning is "the only basic attribute of a firm that could not feasibly be established by contractual means alone" — and organizational law delivers it through one instrument: the registry. (quote: Hansmann & Kraakman, The Essential Role of Organizational Law, 110 Yale L.J. 387, 393 (2000) — verified verbatim 2026-07-12; the registry gloss is ours)
  18. The personhood bans are the moat: the EU killed "electronic persons" and withdrew the AI Liability Directive (2025-02-11); Idaho, Utah, and North Dakota are banning direct AI personhood — leaving the entity route (agent within an entity, disclosed human anchor) as the only surviving legal channel. Legal substrate already live: Wyoming DUNA (eff. 2024-07-01; Uniswap's DUNI, Sep 2025, ~$4B), Alabama and West Virginia DUNA acts (Apr 2026). Samuels v. Lido DAO (2025) makes unstructured collectives one lawsuit from unlimited personal liability. (legislative record; case law)
  19. The wedge's emotional load, labeled as what it is: "Handshake equity kills more startups than competitors do." (pick pillars.v3; internal — a belief, not a statistic)
  20. Bear cases we carry openly: Gartner projects >40% of agentic-AI projects canceled by end-2027; OpenAI's Mar 2026 checkout retreat shows agentic-commerce hype risk. Internal figures, labeled as such: ~1.2B proto-entities and ~400M companies are founder estimates; 2,935,661 claimable entities is our own registry count, published in the e-Residency tradition. (Gartner; press; internal)
  21. Derivation gate — closed 2026-07-12 (was Judge 1's diligence concern; full receipts in RESEARCH-GATES.md). The legal-entities figure now has an external anchor: ≈359 million companies worldwide (Statista, 2023), of which 230M+ are verifiable in official registries aggregated by OpenCorporates (2025); the Statista series (~328M in 2020 → ~359M in 2023) extrapolates to ~385–390M by 2026, so the ~400M canon reads as a stated extrapolation of the best public series, bounded above by D&B's 500M+ D-U-N-S business records (2022). The 1.2B proto-entity estimate now carries a seven-line cited derivation (ILO employment and informality × self-employment ÷ an explicit venture-size divisor; floor ~400M from World Bank/IFC enterprise surveys, ceiling ~2.1B, midpoint ~1.2B). Label the number "~1.2B (founder estimate; derivation on request)" wherever it appears; RESEARCH-GATES.md is the on-request artifact.

Colophon

Fused final, assembled 2026-07-10 from: SOURCE-DOSSIER.md (fact base; §2 figures, §7 risks) · SKELETON.md (structure, constraints card, register map, figure canon) · DRAFT-A.md (registry-first; chassis for ①, ②, ⑥) · DRAFT-B.md (handshake-first; ③ wholesale, grafts throughout) · DRAFT-C.md (personhood-first; chassis for ④, ⑦; ⑤ and ⑧ wholesale) · JUDGE-1.md (investor-persuasion verdict) · JUDGE-2.md (craft + compliance verdict). Standing flags: the T4 moat passage in ① (depth dial), the [trimmed] investor pitch in ⑥, and [FILL] slots on slides 8 and 11. Closed 2026-07-12: the ⑧ #21 derivation gate and the ①/#17 Hansmann & Kraakman quotation (verified against 110 Yale L.J. 387; see RESEARCH-GATES.md).