Date: 2026-09-05; owner decisions updated 2026-09-06. Status: accepted launch design.
Investigated SHA: b97c454d605e750a08da7ff6baab645330175468; the working tree contains further changes.
This document defines the product and launch criteria. Accepting the design does not perform deployment,
fund transfers, or raise the current limit.
Owner alignment, 2026-09-30: MML means provable accounts supporting 51% of world GDP by 2050. Ethereum, TRON and XLNC are the initial J focus; additional compatible EVM Js, including Base and Arbitrum, belong to the same architecture. Supporting them together must reuse the canonical financial path, with deployment, observation, finality and withdrawal evidence for each admitted J. Future central-bank EVM Js fit this model when available. This scope update does not declare deployments or change capital authorization.
USDT · Tron ⇄ USDT · Ethereum. The user gets working liquidity on the needed network through funded Accounts and an executable quote. The first customers are treasury/OTC teams with a recurring flow between payouts on Tron and capital use on Ethereum.
The multi-J architecture, finality, and XLNC boundaries are described in xlnc-soft-mainnet.md. Customer acquisition, onboarding, and pilot economics remain in launch-pilot.md. Where the first market, MM composition, and launch ladder diverge, this later design governs.
H1 H2 H3
\ | /
bilateral Accounts / routing
| |
Tron Entities Ethereum Entities
\ /
MM1 MM2
Our own H1–H3/MM are used both in development and at the first network launch. Independent operators are a subsequent stage. The existence of a second MM is not proven by two names using the same undivided pool of funds.
| Operation | First offering | What must be explicitly shown |
|---|---|---|
| Payment | USDT · Tron → USDT · Tron; then routes between H1/H2/H3 | Recipient, amount, fees, available receive capacity |
| Same-J swap | WETH/USDT · Ethereum | Specific registered tokens, price, minimum receive, fees |
| Cross-J swap | USDT · Tron ⇄ USDT · Ethereum | Both jurisdictions, both sides of the swap, quote validity period and execution conditions |
| Payment + conversion | Alice pays USDT · Tron; Bob receives USDT · Ethereum | A single verifiable quote and a completed result for Bob |
In the current market catalog and executable scenario the asset is called WETH, not native ETH. Until a proven wrap/unwrap path exists, receipt of native ETH cannot be promised, nor can the wrapper be hidden behind an “ETH” label. A local token with that symbol also does not confirm backing by real WETH. Public launch requires a registered address, decimals, issuer/backing, and a verified deposit/withdrawal of the chosen asset.
USDT_TRON and USDT_ETH are convenient labels for different assets. Canonical identity
is bound to the jurisdiction stack and the local token ID, not to the ticker:
cross-j market.
An XLNC token named USDT does not automatically become the same asset or a claim on Tether.
Payment + conversion is the next combined user operation after the cross-J swap is proven. Two independently completed steps cannot be declared a single atomic payment without a corresponding production path and recovery verification. Lending/credit remain part of the Account architecture; a separate lending product must not delay the first working corridor.
How much a user can get right now, at what full price, and with what probability of completion. MM balance and drawn order-book depth are not, by themselves, executable liquidity.
| For each request | Required data |
|---|---|
| Identity | Source/target jurisdiction-local asset, amount, direction, trade size |
| Quote | Net receive, full fee/spread, minimum receive, expiration, observation moment |
| Execution | Canonical route/market, available MM inventory, and the capacity of the involved Accounts |
| Result | Accepted/completed/rejected/expired/cancelled, reason, time to economic completion |
| External cycle | Separate time/cost for funding, rebalancing, J-finality, and withdrawal |
Measure both directions separately, along with trade sizes agreed with customers: quote availability across all requests, completion across accepted quotes, p50/p95 time, full cost, capital tied up, and MM inventory restoration time. The time of funded off-chain execution does not include or substitute for the time of a bank transfer, chain deposit, or on-chain withdrawal.
The owner has accepted the following expansion design. These are design exposure ceilings, not current balances, issued limits, or executed transfers.
| Stage | Accepted target | Verifiable basis for transition |
|---|---|---|
| Public testnet | Test assets only | Full product, accounting reconciliation, failure/exit/restart; normal resource constraints remain |
| Canary | $10 000 | Executable cap, limited exposure, error detection/stop, proven recovery |
| Pilot I | $100 000 | Operational reliability, repeated customer operations, reconciliation, and available exit |
| Pilot II | $500 000 | Real behavior of two MMs, available quotes, and liquidity replenishment |
| Production beta | $2–5 million | Repeatable treasury economics, concentration risk, and operational resilience |
| Mainnet | Dynamic limit | Proven risk engine and an explicitly accepted policy for raising/lowering limits |
The previously recorded $1 000 relates to the prior limited-experiment plan. The owner reconfirmed the ladder on 2026-09-06; it describes the launch design; the actual first tranche, operators, and transition between stages are recorded explicitly. No funds were moved as part of accepting this document.
The current policy still contains
riskCapUsd: null and riskCapEnforcement: not_implemented. Neither $1 000 nor $10 000
can currently be called a programmatically enforced protection.
The cap must cover admitted exposure and obligations, including pending operations, credit, and collateral lease; it must not be bypassed by parallel admissions or restart. Do not sum the same asset repeatedly as reserve, collateral, and Account claim. Show operating expenses, gas/resources, and dispute reserve separately from customer liquidity. The exact aggregation scope and risk assessment belong to risk policy, not a UI slider. An accounting/exit error halts fund movement; transition to the next stage does not happen automatically on a calendar schedule.
From the accepted concept, the following are not published as facts: $23.7 billion/day, 1.06 million accounts/day, sample MM rates, fee $0.84, execution 420/430 ms, liquidity $428k/$842k, turnover $1.8m, and completion 99.99%. Market figures require their own current attribution; interface-example numbers — real measurements with a precise description of the test rig and a link to the artifact.
“Private” discloses the observers and the data available to them. “Instant” requires a measured threshold time and prefunding/availability conditions. Exit depends on the specific J, finality, and transaction inclusion; a local fullnode by itself does not guarantee withdrawal. Do not claim anonymity, trustlessness, or guaranteed on-chain exit time.
The saved replay-report
contains 111 frames, equivalent: true for TS/Rust W1/W4/W8.
This is historical local evidence on its recorded artifacts:
not a current clean-SHA release gate, not public-testnet readiness, and not live TPS.
bun run check and relevant release gates are mandatory; readiness is not determined by model opinion.MML is provability coverage of the accounts supporting economic activity, not payment turnover divided by GDP. Coverage attribution remains to be specified. For near-term adoption evidence, count a completed payment once and a swap without summing both legs; exclude repeat settlement, technical relays, self-trading and faucet traffic. Track repeated useful use and available liquidity separately from production TPS.
The wallet acceptance plan must show funding → payment → same-J/cross-J swap → withdrawal → recovery after hub failure. The new UI explains proof, chosen collateral backing and Delta Transformers; the existing Svelte frontend E2E must exercise the same financial outcomes with exact balances, receipts and completed J settlement.
Owner intent is a responder-specific clock: a user-started dispute gives the hub one hour by default, with longer protection for larger stakes. The current implementation instead uses the sum of both signed response windows: hub one hour plus user 24 hours gives a 25-hour unilateral-finalization deadline. Pull-bearing proofs also use that summed deadline for finalization. The owner clarified that reaction protection applies to every kind of dispute; cross-J claims additionally need safe copying of locks and valid evidence. These are requirements, not an implemented replacement clock.
Owner accepted 2026-09-30. This is the design policy for the implementation task; existing accounts and contracts retain their currently signed clocks.
Use the maximum authorized account exposure, including conditional commitments, agreed before the account opens. For an initial USDT corridor, interpolate linearly between these knots, rounding upward to whole minutes:
| Authorized exposure | Hub reaction window |
|---|---|
| Up to 100,000 USDT | 1 hour |
| 1,000,000 USDT | 8 hours |
| 10,000,000 USDT | 48 hours |
Keep the user reaction floor at 24 hours: userWindow = max(24h, schedule(exposure)).
This is an adopted initial policy, not a security optimum or a live USD oracle. Other
assets require an agreed denomination/conversion rule. Above 10m requires an
explicit signed policy rather than unbounded extrapolation. Never reprice an
active dispute or allow an old short-window proof to authorize a larger exposure.
Current immutable signed windows must be respected. Under today’s summed rule, 10m with two 48-hour windows means 96 hours before unilateral finalization, not a two-day exit. Separating reaction from finalization requires proof of the cross-J recovery boundary before implementation.
For each portable conditional claim, a valid source reveal must still be able to reach the target within its accepted publication window. Account for source observation/finality and target inclusion, including different dispute start times. Example: target starts 00:00, source 00:02; a source reveal valid at 01:02 must not be defeated by target finalization at 01:00. Longer reaction windows alone do not establish this conservation property. Validate the clock/portability rules before changing code or advertising a one-hour complete exit.
These are implementation tasks, not newly demonstrated defects. Preserve the single live release checklist in todo.md; dated red/green reports must be reproduced on the actual candidate before treating them as current.
| Order | Task | Completion evidence |
|---|---|---|
| 1 | Establish the current production/recovery boundary | Reproduce the canonical payment/cross-J/restart artifact; fix its first divergence if present; exact per-frame R/E/A roots and ordered outputs; full check without a missing-tool failure |
| 2 | Implement the accepted signed exposure schedule and safe recovery clocks | Deterministic rounded interpolation; conditional exposure admission; old-proof/active-clock protection; source-late-reveal and sibling-evidence acceptance tests; TS/Solidity agreement and reviewed bytecode |
| 3 | Deliver the financial walkthrough on real J boundaries | New UI explains the three protections; existing-frontend E2E verifies funding, payment, same-/cross-J swap, withdrawal and hub failure; actual secured recovery and unsecured receivable; real Ethereum/TRON observations and receipts |
| 4 | Produce the minimal stateful EVM XLNC artifact | Pinned genesis/rules and reduced block gas limit; one producer and a separate full node; canonical reserve/account/settlement/withdrawal/restart; identical verified roots; measured execution and two-day catch-up cost; correlated-dispute inclusion capacity |
| 5 | Connect adoption evidence to MML and finish reader cleanup | Unique useful completed operations; separate reserves/backing and account claims; a documented GDP-attribution method; dated release claims; one short reading path; root file relocation only after updating verified consumers |
Items 1–3 establish useful enforceable finance. XLNC research may proceed from the existing J boundary without delaying its repair. A billion-TPS measurement and watch-hardware verification are later acceptance artifacts, not substitutes for this first usable journey.