XLN Roadmap
This file is the strategic roadmap.
It is intentionally different from:
../todo.md — current blockers and active work
mainnet.md — the release bar for real-user-fund launch
Strategic Goal
XLN should become the default bounded-risk alternative to centralized exchange
and custodial settlement exposure.
The bet is simple:
- users and institutions want CEX-like speed and UX
- they do not want unbounded custodial risk
- bilateral provable-credit settlement is the right architecture for that gap
Strategic Constraints
These should stay visible because they shape the roadmap more than preferences do.
-
The near-term market is crypto-first, not all of global finance.
XLN can ship today on EVM-based J-machines. It does not need CBDCs to matter.
-
The first wedge is bounded-risk settlement, not universal composability.
XLN wins where bilateral state is natural: payments, settlement, treasury, and hub-mediated flows.
-
Adoption will be ratcheted by trust failures elsewhere.
CEX blowups, opaque custody, and proofless balance-sheet risk are market-making events for XLN.
Product Path
Phase 1 — Working crypto settlement network
Focus:
- make the bilateral runtime, payment flow, and recovery story real enough for serious testnet/prod-like use
- harden the runtime, J-layer integration, and operator surfaces
What success looks like:
- the system is usable by real technical users
- failures are observable and recoverable
- launch blockers are integration/safety issues, not missing core concepts
Phase 2 — Limited real-money network
Focus:
- launch with explicit boundaries
- small limits, strong observability, explicit supported assets and flows
- emphasize bounded-risk settlement rather than “replace all finance” rhetoric
What success looks like:
- early adopters can move value with a materially better risk profile than leaving funds on a CEX
- the operational model survives real usage
- trust is earned through boring reliability, not novelty claims
Phase 3 — Institutional and hub network expansion
Focus:
- more hubs
- better treasury/custody flows
- stronger entity governance and board semantics
- richer recovery and compliance-grade auditability
What success looks like:
- XLN is a credible settlement rail for exchanges, market makers, treasury operators, and other high-value actors
Phase 4 — Broader financial integration
Focus:
- extend from crypto-native settlement into wider financial interoperability when the right J-machines exist
- keep the same bilateral model; do not regress into shared-state architecture just to chase surface area
What success looks like:
- XLN becomes the obvious bilateral settlement layer wherever programmable enforcement exists
Market Sequence
1. Crypto traders and custody-sensitive users
Why first:
- the pain is already obvious
- the users already understand counterparty risk
- the alternative today is mostly “trust the venue”
Value proposition:
- same broad class of UX as a centralized venue
- better risk segmentation
- proofs and bounded exposure instead of blind omnibus trust
2. Stablecoin-heavy businesses and operators
Why next:
- they already move value on-chain
- they already feel settlement friction and operational cost
- they benefit from off-chain bilateral velocity with explicit on-chain enforcement
Value proposition:
- faster bilateral settlement
- less needless on-chain churn
- better auditability than informal custody/trader workflows
3. Inter-hub / inter-venue settlement
Why later:
- this needs credibility, uptime, and governance maturity
- the savings are large, but the trust bar is much higher
Value proposition:
- bounded bilateral exposure
- less pre-funded capital drag
- explicit proofs and dispute paths
CBDC Optionality
CBDCs are an upside branch, not the core plan.
If programmable fiat rails appear on EVM-compatible or equivalently expressive
J-machines, XLN can extend naturally.
If they do not, XLN is still valuable as crypto-native financial infrastructure.
That means the roadmap should never depend on CBDCs for near-term relevance.
What This Roadmap Does Not Try To Do
- it does not duplicate the active blocker list from
todo.md
- it does not restate the launch gates from
mainnet.md
- it does not promise timelines that outrun the current engineering reality
Post-launch engineering backlog
These items are valuable but are not allowed to obscure the active mainnet
blocker list:
- CandidateExecution/FrameDraft copy-on-write execution, Account transition
journals, cached Entity Account-section commitments and extraction of the
large consensus facades after byte-identical differential evidence.
- Event-driven dirty Account/book/deadline indexes, stable proposable queues,
one-signature representation reuse and internal verified-witness reuse.
- Runtime-level CrossJCoordinator plus separate immutable CrossJTerms and
versioned CrossJProgress reducers.
- Cross-j evidence simplification after release: one pure verified-evidence
decoder shared by consumers without sharing their authority checks; one
committed-route transition reducer; and a minimal canonical progress tuple
with UI/debug projections derived from it.
- Rebuildable cross-j lookup indexes and consolidated telemetry formatting.
Indexes remain non-authoritative, bootstrap from committed state, and are
checked against canonical state; opening/progress/close retain distinct typed
validators instead of a generic financial transaction.
- A centrally tested dispute-evidence budget derived from the canonical ABI and
Account resource limits, so future limit or ABI changes cannot make honestly
signed state exceed jurisdiction calldata bounds.
- Explicit bilateral special-confirm recovery for a finalized disputed Account,
with a new domain-separated confirmation, immutable finalized J-nonce binding,
and complete stale proof/draft cleanup. Until that protocol is designed,
audited, and activated, finalized disputed Accounts remain permanently closed;
ordinary
openAccount never replaces them and no compatibility reopen path exists.
- Rich browser physical-storage inspector, unified QA/Health cockpit,
shareable detached history viewer and additional failure-inbox UX.
- Complete BrowserVM-only catalog parity for the non-release
lock-ahb and
dispute-transformer runners before exposing either as a browser preset.
Their dispute start/finalization uses the canonical scenario clock and receipt
evidence, but later multi-entity history fanout can still reject with
J_PREFIX_LOCAL_PREFIX_MISMATCH; RPC/mainnet and the shipped hub-collapse
Scenario Player flow are independently green.
- Persist the original signed swap
maxFee and minNetReceive in terminal
swap-history projections so closed-order UI and audits retain the user’s
authorization envelope. This is audit metadata only and must not introduce
another fee formula or change execution authority.
- A future fresh typed mutable-path schema for generic oversized Entity/Book
records. There will be no compatibility reader before mainnet.
- Replace the fixed per-Entity Account-count ceiling with load-priced admission,
but only after paged Account hydration and bounded iteration are proven. The
Hub publishes a signed ten-minute policy epoch; a new Account open binds the
exact Hub, requester, bilateral pair and epoch to either adaptive Hashcash or
a bond paid through an existing Account. Existing Accounts remain usable.
Difficulty rises with Account count and measured reducer/WAL/network load;
exact retries deduplicate, while new attempts consume per-peer budgets. Track
per-Account transactions/second and deterministic machine cost, with separate
operator-set budgets for Hub-to-Hub relationships. This is long-term DoS
admission work, not a justification for a larger in-memory hard cap today.
Next product milestone — universal jurisdiction Stack Manager
- Add a dedicated Settings tab for deploying one canonical jurisdiction stack
to an arbitrary EVM RPC. Show the active BrainVault signer, chain id, native
gas balance, refresh/funding guidance, estimated deployment cost and exact
resulting contract addresses before the jurisdiction is saved locally.
- Expose basic network/name inputs first and keep admin, token catalog, dispute
defaults, fee/risk limits and deployment verification in collapsed Advanced.
- Build one typed
deployStack() orchestration library used by both the UI and
Bun CLI. Foundation deployment becomes a configuration of that same path.
Deployment manifests carry an explicit stack version (V1 for the current
contracts); versions select immutable deployed bytecode, never compatibility
behavior inside the live Runtime path.
- Persist a deployed stack only after chain-id binding, bytecode/link-reference
verification and receipt finality. Adding it to local jurisdiction discovery
must not mutate any Entity consensus configuration implicitly.
- Publish community stacks to Gossip under the deployer’s signer authority.
Publish an
official jurisdiction only with Foundation Hanko; no other signer
can announce or replace Foundation jurisdiction metadata.
- Orchestrate one company-formation action that creates the company Entity,
registers it, opens its Account with the selected hub, and proves restart
persistence. Cover both founder 1-of-1 and directors 2-of-3 companies.
- Mint separate control and dividend token classes owned initially by the
company treasury; BOARD remains encoded Hanko authority and is never a token.
The listing hub receives no ownership.
- Create CONTROL/USDT and DIVIDEND/USDT books automatically. The company may
sell treasury shares, make ordinary payments and buy shares back with ordinary
treasury bids; founders invest through ordinary buyer Accounts.
- Allow a holder proving more than 50% of CONTROL to schedule on-chain board
replacement without approval from the old board. Preserve the existing
seven-day historical-proof grace instead of invalidating old proofs early.
- Build the deterministic scenario first, then real E2E/UI coverage for issuance,
partial sales, multiple buyers, company payment, buyback, failed/successful
takeover, delayed rotation, replay and continued trading. Save and inspect
screenshots for founder, investor, company treasury and governance states.
Human-audit source context and lean LLM bundles
- Add a concise 3–4 line header to each production
core/ source file:
purpose, key authority/invariant, principal entrypoints and audit importance
/100. Do not comment obvious helpers or restate types; function-level why
comments are reserved for financial, consensus, storage, crypto and other
non-obvious authority boundaries.
- Extend
scripts/debug/gpt.cjs with a second deterministic audit profile that
includes the same mandatory Runtime/WAL/storage/Merkle/crypto manifest but
removes comments with a parser/tokenizer rather than regular expressions.
Preserve licenses and compiler directives. Keep the ordinary annotated bundle
so reviewers can compare implementation alone against documented intent.
- Gate both bundle manifests for required critical files, exact source hashes and
deterministic output. Audit comment quality separately from protocol code so
documentation changes cannot silently alter the clean-code evidence bundle.