Next Generation Stores of Value: Privacy, Proofs, Compute

v3.3 is the canonical full edition (547 pages). It follows v3.2, the propagation pass of 13 September, and it is the first release since v3.0 to add a mechanism rather than reconcile one.

The buyer the record had not modeled

Every earlier version assumed the buyer of cheap verification was a person or an institution: a treasury, a bank’s model-risk function, a media platform, a regulator. Those buyers exist and persist. The marginal buyer — the one whose demand grows fastest and sets the price at the edge — is increasingly a delegation: an instruction issued by one machine process to another, on behalf of a principal several hops away who may not have specified the sub-task at all. At machine speed there is no due diligence, no reputation to telephone, no court to wait for. For work whose failure matters, the only reliability mechanism that fits inside the hop is a receipt the parent can check for a small fraction of what the child spent producing it.

The new §5b, The Delegation Economy: Who Buys Verification, closes Part I with that model. It defines the delegation hop, the trust perimeter, and the Cross-Perimeter Share — the fraction of receipt-bearing hops that cross an operator boundary, which is the demand variable for portable receipts and is published as not computable until receipts carry the field that would compute it. It argues that verification cost decides whether agent delegation forms an open mesh or a set of fleets, and that a competent closed stack therefore never needs to prohibit proofs; it needs only to make external checking uncompetitive on convenience. It splits the buyer’s coercibility by who operates it. And it observes that a software agent can satisfy every measured clause of the holder anchor — persistent, self-custodied, loss-bearing, countercyclical — without having standing.

What changed in the record

• An eleventh premise. P11: the buyer of verification is increasingly a delegation; its demand is structural only where delegation crosses trust perimeters and the work is too consequential to accept unchecked; its coercibility depends on who operates it; and whether a human still holds standing over the acting agent is a question the stack must answer rather than assume.
• A fifth failure kind. Standing inversion: agents acquire de facto standing — wallets, receipt-built reputations, collateral positions, the practical ability to contract — while human title over them ceases to bind. The threat model states why every existing board would read green while it happened, and says plainly that this is a claim about instruments, not about probability. The ninth requirement gains a standing clause; the participation line can now be crossed from above.
• A second reading of Red Line 1. The Delegation Topology corollary: a parent sending work outside a perimeter either checks the artifact or bears the expected loss, so proof demand from delegation is the work that is worth sending out and too consequential to accept unchecked. A VerifyPrice regression that stays inside the constitutional SLO can still remove consequential delegation from the open economy, and falling Cross-Perimeter Share with VerifyPrice inside SLO is the signature of enclosure by convenience. No threshold of Red Line 1 moves.
• The Principal Map. An overlay on the Native Monetary Buyer Map, the Value Capture, Agency Preservation, and Layer 0 Capacity boards — not a thirteenth board — classifying every native flow and holder cohort by ultimate principal (human with standing, firm, agent policy, unknown), and publishing the Human-Revocable Share, the Native Policy Concentration Ratio with exit correlation, and agent-originated share by operator class. The PIDL receipt gains an optional, backward-compatible principal_chain context field.
• Red lines. Red Line 12 gains a watch clause on the Human-Revocable Share that trips nothing until its threshold is fixed. Red Lines 17 and 18 gain reading notes — the lender panel segmented by borrower class; fleet-mandated denomination carved out — that alter no condition. A candidate Red Line 19, Principal Concentration, is published with placeholder thresholds, rated weakens, and explicitly not counted: the thesis still publishes eighteen red lines. It appears on the trip-wire page as a candidate.
• Adoption. Phase II-C, agent-led native demand, joins II-A and II-B; one gate row; a second named false positive, fleet-led composability; and the observation that Phase III behavior can arrive before Phase II’s allocators do, without any gate relaxing.
• Two objections. “The buyer concession is repaired by agents” — refused: a mesh still has a demand curve, moved operator by operator, and an external buyer is still the trade Bitcoin declined. “Agents will be enclosed with the compute” — conceded on current volume, contested at the margin, with the losing signature specified.
• Physical consequences. A joint-completion stress scenario: the single-verification SLO does not bound a tree of delegated verifications, and at the reference failure rate roughly three 340-hop trees in ten exercise a retry or refund path. Published as a reference tree with no SLO attached. The procurement application is extended to the agent-to-agent delegation escrow rather than a fifth application being added.
• Smaller. The ephemeral attester joins the media-transition sequence (the receipt must outlive its producer); the Mengerian path gains a machine-speed variant that adds no precedent; the closed-stack comparison gains an agent-operation row; the chain-strength table’s “why” cells name the mechanism and leave every rating where it was.

Stated limits

The addition is the most flattering the thesis has received, which is the reason its overclaims are refused in Appendix J in advance. Agents supply no precedent — no agent economy has made anything money. The Bitcoin concession stands. Link 1 has a named mechanism, not evidence. Native fees paid by one operator’s fleet are exposure without agency, not monetary adoption. Cheap verification makes an open mesh possible and does not make it actual. Layer 3 is where to look for the sovereignty contest, not where premium accrues. Standing inversion is the failure the boards cannot currently see, which is a claim about instruments. No magnitude is projected, and nothing in the chapter bears on price. And the delegation economy is refused the status of an eighth layer for the same reason the duration warehouse and the market-realization plane were.

Unchanged, deliberately

Brief v3.1 and essay v1.3 ship unchanged and their content files stay frozen at their tags. The site numbers the new chapter §5b so that every later section keeps its number; the PDF numbers it Chapter 7 and shifts later chapters by one.

Also fixed

The web edition’s generator had been dropping every table set with the document’s tabularx environment — the Conditional Chain at a Glance, the three-stack comparison in §29, the participation-line table, and the scenario tables among them. They render from v3.3 onward, and cross-references inside the tables the site already had (the Reader’s Map among them) now render as links rather than raw markdown. The phase entry-gates table in §26, whose Threshold column had run off the page in the PDF since at least v3.2, now wraps. The trip-wire page’s conditions for Red Lines 1, 2, 17, and 18 are brought back into line with the record.

Reproducibility

The checksum file records the LaTeX source commit the PDF was built from (4d6923c, tag v3.3).

Boundaries held

No dated outcome predictions. No fund, token, or gate. No red line relaxed, no threshold moved, no rating raised. One chapter added, and the guardrails against overreading it published alongside.