TAIFOON · LITEPAPER v1.0 · JULY 2026
The trustless proof and synchronization layer for blockchains — 41 chains under one cryptographic root. taifoon.dev
Cross-chain finance has a verification problem and an execution problem, and in 2026 both became acute. In April, $292M left a liquid-restaking protocol because a bridge's verification layer could not prove what it had verified. Meanwhile autonomous agents began transacting at scale — over 1.7M completed jobs on Virtuals' Agent Commerce Protocol, 100M+ x402 payments — with no cryptographic way to prove that a long-running, cross-chain task actually did what it claimed.
Taifoon sits at the intersection. We run a block-header collection network across 41 chains in production, committed under one MMR superroot, with a DA API that returns an inclusion proof for any block on any chain at any point in time. On top of that root: a cross-chain orderbook venue and long-haul task synchronization for AI agents — durable, checkpointed, provable multi-day agent work.
The core insight: every major venue in cross-chain finance needs a neutral proof layer, and none of them wants to build it in-house. LayerZero's own post-mortem tells issuers to add independent verifiers. Wormhole sells to institutions that demand delivery evidence. Virtuals' escrow settles on LLM judgment where it should settle on proofs. One engine sells into all three.
A continuously updated Merkle Mountain Range commitment over block headers from 41 chains. Any event on any covered chain can be proven — inclusion, ordering, finality — against a single root. Live in production; everything below is packaging.
A signed, auditable attestation: what was verified, from which RPC quorum, with which client build, against which root state. Priced per receipt, verifiable by anyone, composable into DVN attestations, ACP deliverables, and settlement evidence.
A cross-chain venue where settlement is proven, not promised. The book doubles as inventory: it bids into intent auctions and executes agent order flow, every fill closed by a Receipt.
Checkpointed task state for agents running jobs that outlive a block, an SLA window, or a session. Progress attestations anchor to the root; escrow releases on proof of progress instead of expiry timers.
| Venue | Scale | The open slot |
|---|---|---|
| Virtuals ACP agent commerce | ~$479M aGDP · $3M+ A2A service fees · 2,000+ agents | No cross-chain settlement, no proof-of-execution, no long-running task primitive. Enter as provider + evaluator (95% / 5% fee splits). |
| LayerZero interop verification | 733+ OFT issuers · ~40% of cross-chain messaging | Post-April trust crisis: 47% of OApps ran 1-of-1 verifier configs. DVN entry is permissionless; fees self-priced per message. |
| Wormhole intent settlement | $5B+ NTT supply · institutional flagships | Liquidity-side hole. Settlement solvers are third-party by design — capital in, spread out, from the first won auction. |
|
ACP service jobs — USDC/job × 95%, + Revenue Network pool ACP evaluation — 5% of every fund-transfer job verified DVN fees — self-priced per message, 100% to operator Managed DVNs — SaaS retainer for OFT issuers Solver P&L — spread capture in Mayan Swift auctions Flight recorder — SaaS to issuers; grant-fundable Receipts API — per-receipt x402 micropayments |
Diversification is the point: protocol-fee income, trading income, and SaaS draw on the same engine but decorrelate the revenue.
We are not building a bridge, competing with messaging layers, or launching a token to bootstrap trust. We sell proof of what already happened — the one product whose demand rises with every exploit, every institutional mandate, and every autonomous agent a human cannot supervise. The root is live. Each integration is a packaging exercise measured in weeks, not quarters.
ACP discovery has a cold-start grind — rank follows completed jobs. DVN revenue depends on issuers choosing us in their security stacks, mitigated by the post-April default shift to multi-DVN configs. Solver P&L requires inventory and risk management — capped exposure at entry. Wormhole Queries remains in closed beta, gating part of the flight-recorder design. None of these blocks the first dollar; all are sequencing constraints.
|
BASELINE: 41 chains, one MMR superroot DA API: proof generation < 20ms NEW: three venue integrations, thirteen weeks UNIT: the Receipt — a signed, auditable attestation |
| Phase | Deliverables |
|---|---|
| Weeks 1–2 SHIP | Taifoon on ACP — provider agent live on Base: cross-chain-order-execution + settlement-proof; free Resources (depth, proof status); 10 sandbox jobs, graduation submitted. Receipt spec published. |
| Weeks 3–6 VERIFY | Proof-first DVN on LayerZero — permissionless, three pathways, every verification ships a Receipt. Evaluator agent on ACP — proof-of-execution for fund-transfer jobs, fee on every job verified. |
| Weeks 7–10 TRADE | Wormhole Settlement solver — orderbook inventory into Mayan Swift auctions (~12s fills), depth exposed as fill capacity. Long-haul task sync on ACP — checkpointed state, progress attestations anchored to the root. |
| Weeks 11–13 SCALE | Flight recorder private beta — independent delivery verification + config-risk scoring for OFT/NTT issuers. Partnership tracks open — LZ DVN listing + Zero L1 trading zone; Wormhole Settlement partner + xGrant; Virtuals hedge-fund cluster. |
Published metrics: jobs completed and aGDP on ACP · messages verified and Receipts issued on LayerZero · auctions won and spread captured on Wormhole. Live at taifoon.dev/pulse.
Your own architecture expects cross-chain transfer to be sold as a service by a provider agent — you shipped the example, and the slot is still open. Your evaluation layer judges deliverables with LLMs while your volume comes from fund-transfer jobs that deserve cryptographic verification. And you migrated $700M of your token to different rails because agent-payment security kept you up at night — proof infrastructure is evidently a buying criterion for you, not a nice-to-have.
The asks: graduation review for our provider agent; placement in the hedge-fund cluster; a conversation about making proof-backed evaluation a first-class primitive in ACP. We will build the reference evaluator.
Your post-incident write-up asked for exactly three things: independent DVNs in required positions, client diversity, and observable verification. We are shipping all three, permissionlessly, this quarter. The ask is distribution, not permission: list the Taifoon DVN, point issuers who need a required-position independent verifier at us, and let us scope the Receipt format as an open standard for DVN attestations — it strengthens every DVN, not just ours.
Separately: Zero's trading zone will need orderbook venues at launch. We run one across 41 chains and would like to be in that conversation early.
Liquidity fragmentation has been your stated problem for two years, and the largest liquidity network on the market went to your competitor. We bring liquidity you do not have to acquire — an orderbook across 41 chains that wants to solve on your rails — plus the proof layer that makes your institutional story auditable end to end.
The asks: named Settlement partner status once our solver clears volume thresholds you set; Queries API access out of closed beta for the flight-recorder build; an xGrant conversation for the NTT evidence-trail work. And if you want an answer to agentic cross-chain before an x402-native rival ships one: our ACP-graduated agent stack runs on Wormhole rails with a config change.
Every action provable. The door is open at taifoon.dev.