SUNIMA

No cooperation channel under any circumstance. Mathematically clean, but every major jurisdiction now treats this posture as categorically illicit. Sanctions, frozen front-ends, criminal exposure for contributors. Examples: Tornado Cash.

Strength: uncompromising. Weakness: does not survive long-term enforcement pressure.

Custodian or operator can see everything, hand over everything. Compliant by construction, but offers no real privacy: a single subpoena, a single breach, a single rogue insider exposes the entire user base. Examples: centralised exchanges, KYC-gated DeFi.

Strength: regulator-friendly. Weakness: not actually private.

Encrypted by default. Disclosure only via formally-served court order, executed by a threshold-cryptographic ceremony, not by a trusted operator. No human can unilaterally decrypt. The cooperation channel exists, but it is narrow, auditable, and cryptographically gated.

Strength: survives both privacy expectations and enforcement reality. Weakness: harder to engineer — which is why it has not existed before.

0xe5254A5E21403Ae07a6925071796d596451f7f86

Arbitrum Sepolia · verified · deposit/withdraw/transfer transacting end-to-end. 2048-bit Paillier additive HE on-chain.

PRIONPatchGateway at 0x9Fd67D98…46Fd9 · PRIONInvariants v1.0.2 at 0xA85a52CE…56C17

Arbitrum Sepolia · source-verified · authority-free upgrade gate. 3-of-5 Sentinel Committee multisig + 24h/1h/72h timelock + Founder Veto. v1.0.5 architectural batch closed 2 HIGH + 2 MEDIUM + 2 LOW + 4 INFO across 7 cold-eye iterations: clearForcedDormancy founder-rescue, 2-phase declareFounderDormant/latchFounderDormant latching, 2-phase requestReactivation/latchReactivation past-grace recovery, 7-day grace + FounderLatePing past 30d softEnd. 66/66 PatchGateway tests + full forge suite green.

0x3a0d9A52416847D984a17D0dBe849cdF6492209E

Arbitrum Sepolia · EC-ElGamal on BN254 · deposit 151K gas, withdraw 89K, encrypted transfer 222K — measured live. 12–18× reduction over the Stage 1 Paillier vault. v0.7.1: Schnorr organizer, guardian/pause, withdraw wired to TFHE range-proof attestation end-to-end.

0x9e11c9840f17d799b11c03ba129676dde30b88d6

Arbitrum Sepolia · source-verified · on-chain TFHE attestation verifier. Off-chain Rust worker (Zama tfhe-rs v1.6.0) signs EIP-191 range-proof attestations; the contract verifies and replay-protects per (consumer, ctHash, chainid, verifier, validUntil). Pause governed by an isolated guardian role — capped at 30 days with a 7-day cooldown; no custody, no drain path. Measured: verifyAndConsumeRangeProof ~33K gas (median). V2 retained for the deprecated immutably-bound vault; new consumers wire to V3 below.

Full Foundry suite passing across live stages — no regressions

Zero regressions through all live deployments. Includes adversarial suite per stage. Latest pass: full suite green after PatchGateway v1.0.5 architectural batch (66/66 PatchGateway tests + 63/63 Raise tests).

Sepolia testbed · full testbed supply minted · 600M in SunimaVesting v2 at 0xEa8A0BE8…AE9cC

Total supply equals the testbed MAX_SUPPLY (1B on Arbitrum Sepolia — the sovereign mainnet launches at 100,000,000 fixed; see Tokenomics). Founder isMinter revoked → no future mint possible. Vesting v2 redeployed 2026-05-07 to point at SunimaGovernance v1.0.2 (immutable governance pointer); 600M re-minted into the new Vesting. Six supporting contracts live (token, Vesting v2, Governance v1.0.2, RaiseV2, slashing, fee router).

Hub + CollectionPool + Privacy Vault — three-contract shielded deposit pipeline

Arbitrum Sepolia · all three source-verified. SunimaHub at 0x87ac801b…C8A7 (router · 2-day timelock · TVL aggregator) · Privacy Vault at 0x45748A…E889 (note-commitment custody, commit-reveal flow) · CollectionPool at 0xb45674…05A8 (1-hour epoch batching).

0x1840e8E2EF53e8AaA846941aa628d7bd70E27396

Arbitrum Sepolia · source-verified · derived from V2 attestation verifier. Adds verifyAndConsumeCompare (encrypted EQ/GT/GTE/LT/LTE) and verifyAndConsumeConditional (encrypted if-then-else) for future consumers. V2 remains wired to the production vault — V3 ships the richer predicate surface ahead of demand.

FROST 2-of-3 · RFC 9591 — threshold Schnorr signing across independent trust domains. Testnet milestone; operator decentralisation in flight on the Stage 5 sovereign-chain testnet.

p50 156 ms · p99 194 ms · auto-failover <200 ms · on-chain withdraw verified — tx 0x144fc744…. Numbers & falsification checklist →

2-of-3 verified end-to-end on testnet; the same protocol scales to the production k-of-N committee — FROST is threshold-ratio agnostic (same round structure, same on-chain verifier). Final committee size and threshold are determined at Stage 4 production deployment from operator-cohort coverage measurements rather than fixed at design time.

Operator onboarding pipeline tested end-to-end 2026-05-03 against the internal setup flow; threshold quorum reached and a withdraw transaction was produced.

chain-id: sunima_8081-2 · Cosmos SDK + CometBFT · Validator set live

Sovereign L1 testnet running the full Sunima chain since the clean regenesis 2026-05-20. Validator set bonded across multiple jurisdictions; operator onboarding by application during testnet phase. Operator module x/operators live in-binary with the V4.2 family — bond + probation pool + 50/50 reward intercept (V4.2-2b), staking hook auto-link (V4.2-1), governance ante denylist for probation validators (V4.2-3), and stale-link prune in BeginBlocker (V4.2-4). Court-order module x/courtorder shipped on the testnet binary as the Cosmos-native equivalent of the Solidity gateway. Public RPC: sunima.uk/chain-rpc/ · REST: sunima.uk/chain-rest/ · live network dashboard: sunima.uk/network. Operator onboarding handled by application — setup instructions issued to approved operators directly. Further testnet upgrades are in progress, hardening the chain ahead of the Stage 5 mainnet gate. Specifics are announced when each milestone ships, not before.

0x01fea626…da039

Arbitrum Sepolia · source-verified · first V3 consumer. Auditable on-chain registry implementing the legal-bridged disclosure path described in the Three Positions section above. Lifecycle: registerOrder (judge-only) → discloseConditional (anyone, requires two worker signatures: gateway-binding domain 0x05 + V3 envelope domain 0x04) → revokeOrder (judge-only). Half-open expiry interval [issuedAt, expiresAt). Cold-eye review GREEN at iteration 2 (iter 1 caught a HIGH cross-order replay, fixed in v1.1 design before deploy). 46/46 contract tests + full forge suite green. Worker endpoint /conditional_with_order_binding for the binding signature is the next milestone — the contract is live and verifiable on-chain, end-to-end disclosure flow ships when the off-chain worker emits the second signature.

0x739379D2…E5161

Arbitrum Sepolia · source-verified (Sourcify match). 8 fixes applied across cold-eye rounds 2–5 and Incubator pass 1: rotationEpoch invalidation closing the phantom-approval bug, dormancy gates, dynamic threshold, and SignerIsFounder guard preventing the founder EOA from occupying a sentinel signer slot. Incubator v1.0.2 result: 14 roles × 3 models → 0 council-confirmed findings ≥ MEDIUM. Slot 0 now occupied by an isolated signer EOA — founder is no longer a signer, only a vetoing principal.

0x809Dcbd6…7e0d3

Arbitrum Sepolia · source-verified · per-round escrow with bps-accumulator refund, vesting committed at createRound, partial-vesting scaling, pull-based fees. Walked through 9 BIOPSY iterations: 1 CRITICAL + 4 HIGH + 5 MEDIUM in the original draft → 5 architectural rewrites → 0 findings ≥ MEDIUM on the deployed bytecode. Full raise-test suite and the full forge suite pass, no regressions. Implements the milestone-gated raise model described in the whitepaper.

0xEa8A0BE8…AE9cC

Arbitrum Sepolia · source-verified · redeployed 2026-05-07 because the original Vesting committed its governance pointer immutable at construction and that pointer was the deprecated Governance v1.0.0. v2 points at SunimaGovernance v1.0.2. 600M SUNA minted into v2; total supply now equals the testbed MAX_SUPPLY (1B on Sepolia; sovereign mainnet supply is 100,000,000 fixed); founder isMinter set to false → no future mint. Original Vesting remains on-chain as an orphan (locked behind dead governance, beneficiary count 0 — does not affect circulating supply).

Repository opens to the public once the full-matryoshka deployment is live.

Whitepaper, contracts, deployment artefacts, and SECURITY.md available to investors and auditors on request until then.

Stages 1–3 are proof-of-concept research closed on Arbitrum Sepolia. Stage 1: Paillier additive HE reference vault — E(a) + E(b) = E(a+b) with no decryption. Stage 2: EC-ElGamal on BN254 — encrypted balances at 12–18× lower gas than Stage 1. Stage 3: TFHE Predicates v3 — encrypted range, comparison and conditional attestations consumable on-chain. These proved the primitives; the mainnet privacy layer is built on Halo2-KZG shielded notes.

Seven independent security layers (L1-L7) wrap the core vault. Each layer enforces a specific class of checks — rate limits, anomaly detection, sanctions filtering, threshold authorization, settlement queues. A vulnerability in one layer cannot cascade into the others. Monitored by PRION — the nervous system of the fortress, not its brain.

Ops and core are physically separated contracts: operations (DEX/lending/staking logic) hold zero funds; core holds all funds behind a FROST k-of-N threshold-MPC lock. Communication flows through a narrow signed-note channel. A full breach of operations produces zero custody loss.

Gray market respected. Black market (OFAC, terror, ransomware) disclosed only via on-chain court order petition — validator quorum, cryptographic proof, public log. No mass surveillance. No known operators. Stage 5 deliverable.

Sunima Labs operates without public founder identity, without paid advisors, without institutional VC oversight. The whitepaper, the audited contracts, and the live deployments are the credentials. Progressive decentralization: founder trust → validator-set trust → threshold cryptography.

Stages 1–4 ran on Arbitrum Sepolia (EVM) as proof-of-concept. Stage 5 is the sovereign Sunima Chain: Cosmos-SDK + CometBFT, with the mainnet privacy layer built on Halo2-KZG shielded notes — protocol-level, not EVM contracts. Testnet is live (chain-id: sunima_8081-2). Operator onboarding by application during testnet phase.

EC-ElGamal additive encryption over the BN254 curve. Same encrypted-balance interface as the Stage 1 Paillier vault, but point-addition replaces 2048-bit modular exponentiation — yielding a step-change in on-chain cost without weakening the homomorphic guarantee. v0.7.1: Schnorr organizer via ecrecover, organizer rotation secured by a 3-day timelock. Single-organizer model in this proof-of-concept; threshold MPC is the production path.

Measured cost on live Sepolia: deposit 151K gas, withdraw 89K, encrypted transfer 222K — a 12–18× reduction over the Stage 1 Paillier vault across all three primary operations. The full forge suite passes across the live-stage suite, source-verified, zero regressions through deploy.

Stage 2 Part 2 — SunimaTFHEPredicates deployed as an independent on-chain attestation verifier at 0x9e11c984…88d6 on Arbitrum Sepolia. SunimaVaultElGamal v0.7.1 and SunimaTFHEPredicates are composed end-to-end: every withdrawal requires a valid verifyAndConsumeRangeProof attestation from the TFHE worker. Derived verifier SunimaTFHEPredicatesV3 at 0x1840e8E2…7396 (source-verified, same worker and guardian) extends the EIP-191 envelope discipline with on-chain verifyAndConsumeCompare and verifyAndConsumeConditional for downstream consumers; no live consumer yet.

The original Stage 1 reference implementation. 2048-bit Paillier keys with on-chain modular addition. Superseded by the Stage 2 ElGamal vault above as the research target, but kept deployed and demoable — the public Try Vault flow on this site still routes here.

Measured cost on live Sepolia: deposit ~1.9–2.3M gas, withdraw ~1.6M, encrypted transfer up to ~3.4M. Heavy by design — Paillier proved the pattern at Stage 1; ElGamal carried it forward at Stage 2.

Six supporting contracts for the $SUNA token-generation-event pipeline, the milestone-based pre-seed raise, and PRION upgrade gating — all live and verified on Arbitrum Sepolia:

• CoreToken ($SUNA) — Arbitrum Sepolia testbed ERC-20, 1B hard cap (sovereign mainnet supply is 100,000,000 fixed — see Tokenomics); no-premint fair-launch; holder opt-in required before Layer5 can move or burn balance. Contract 0x7B7F08A0299C2833e7b73DD13afE6BeFfCEAe7fE
• SunimaVesting — On-chain enforcement of 2 % founder/team and 1.5 % investor sell-rate caps via a true rolling 90-day window, plus role-typed TGE fraction caps (10 % founder/team, 15 % investor). Contract 0xEa8A0BE81EE0f014B9FAC51aFdF00c96135AE9cC
• SunimaRaiseV2 — Milestone-based SAFT contract for the pre-seed round. Per-round stablecoin escrow, bps-accumulator refund accounting, vesting tokens committed at round-creation and pre-funded before any release, partial-vesting scaling proportional to delivered milestones, pull-based fee accrual. Founder excluded from the 2-of-3 milestone multisig. Passed internal cold-eye review through 9 BIOPSY iterations. Contract 0x809Dcbd660552FF4CAf67B7C9590c432A337e0d3
• PRIONInvariants — Hard-deny list of protected patch types (threshold-MPC parameter floor/ceiling, Sentinel-threshold floor/ceiling, kill-switch disable, invariants swap, PRION authority grant, founder-veto modify/extend). Default-DENY for unknown tags. Patch tag SHAMIR_THRESHOLD is retained at the contract level as a legacy identifier — it now guards FROST threshold-MPC parameters per the custody design evolution; renaming would require an invariants-contract redeploy. Contract 0xA85a52CE6b71cAAC335512699df3c91223956C17
• PRIONPatchGateway — 3-of-5 Sentinel Committee multisig that routes every PRION-originated patch through the invariants check and a 24 h main / 1 h embedded / 72 h rotation timelock ladder. Founder-veto holder may cancel queued ops via cancelByFounder until the deploy-time founderVetoExpiry (whitepaper target: Stage 5 + 24 months). Contract 0x9Fd67D98A05cdf38D1Fe13ff5fea3C5321a46Fd9
• SunimaGovernance — General-purpose 3-of-5 sentinel multisig for CoreToken / SunimaVesting admin ops. Same timelock ladder and founder-veto window with deploy-time founderVetoExpiry; Dead Man's Switch opens a 4-of-5 bypass on rotation ops after 30 days of founder dormancy. Contract 0x739379D2361b0A818FFadE10b10eE395496E5161

Paillier additive encryption (Stage 1). EC-ElGamal additive arithmetic on BN254 (Stage 2 — Part 1). TFHE for richer encrypted predicates (Stage 2 — Part 2) — proof-of-concept on Arbitrum Sepolia. The mainnet privacy layer is Halo2-KZG shielded notes at the protocol layer. Build private AMMs, private lending, private order books without implementing the cryptography yourself.

ERC-5564 stealth addresses + ERC-4337 account abstraction. One transaction = two visible addresses. Everything between them is invisible. Specified, not shipped: the whitepaper lists receive-side stealth addressing among the mitigations for transaction-graph observability, and no contract or chain module implements it yet.

Programmable compliance. Petitions submitted as signed on-chain messages; validator quorum votes on jurisdiction and proportionality; selective disclosure only on threshold approval. All approved disclosures publicly logged. Stage 5 deliverable.

The vault is live today. Developer documentation, interface specifications, and integration guides ship alongside Stage 2. Early builders receive $SUNA allocations from the ecosystem treasury at mainnet launch.

Interested in building? Whitepaper section 11 (Node Architecture) + section 7 (SunimaVault) describe current and near-term primitives. Full ecosystem grants framework opens with the Stage 5 mainnet.

The only surface users interact with. Receives calldata, forwards to L2. Deliberately disposable — if L1 is breached, L2–L7 behind it contain the blast.

Rate limits, anomaly detection, sanctions filtering, reentrancy locks, pattern recognition. Each layer is an independent contract with its own signer set. PRION monitors, scores, and flags — it does not act. Every flagged update passes the Sentinel Committee 3-of-5 multisig and an on-chain timelock before it touches a layer. AI flags; humans and multisig deploy.

The routing layer. Every request that reaches L5 has passed L1–L4 gates. L5 dispatches to the correct op-code (deposit, withdraw, transfer, shield) and forwards to L6/L7 for final settlement.

Honeypot detection, blacklist enforcement, counter-attack activation. L7 is where Sunima bites back — provably malicious addresses get their subsequent transactions reverted, their state frozen, their gas burned.

All user funds live here. Separated from ops by a narrow signed-note channel (the "slit"). No user-defined logic executes inside core. A full compromise of ops produces zero custody loss.

The core vault authorises a state transition only when a k-of-N FROST quorum produces an aggregated Schnorr signature. Operator shares are distributed across at least three independent vector classes — permissionless node runners, cold infrastructure on independent chains, steel/paper offline backups, and user-wallet-bound HKDF shares. Final allocation is ratified by the active validator committee and the DAO at activation, sized to the actual network topology. No single actor — including the founder — can produce a valid signature, and the master key is never reconstructed during signing. Earlier drafts described this as a Shamir 67-of-100 reconstruction lock; that model has been replaced by FROST threshold MPC. The reconstruction primitive opened a vulnerable window at the moment of disclosure when the full key briefly existed in plaintext on a single host. FROST eliminates that window — partial signatures are aggregated without ever materialising the master key.

Rust + Bloom filters. OFAC checks, known drainer lists, amount thresholds, rate limits. Sub-millisecond, zero cost. Handles the overwhelming majority of traffic without invoking the deeper review pipeline.

Lightweight models run locally on Sunima infrastructure. Graph embeddings detect Lazarus Group fingerprints, structuring, mixer hops, and known exploit sequences. Borderline transactions are escalated to Tier 3.

Multi-model verification with independent cold-context review. Same engine that audits Sunima's own source code — in internal review it has flagged candidate findings later confirmed as real bugs that single-pass self-auditing had missed.

Every PRION-originated patch passes four independent gates before production: (1) sandbox regression against the full test suite; (2) independent re-audit by a second PRION instance with a different training cut and architecture; (3) 3-of-5 signature quorum from a rotating sentinel committee, distinct from the validator set; (4) on-chain timelock — PRIONPatchGateway enforces 24 h on main-layer upgrades, 1 h on embedded-agent updates, 72 h on rotation events. Founder veto cancels queued operations during the published veto window.

Misalignment is bounded by sandbox-only training (production state never feeds classifier weights), behavioural invariants enforced at PRIONInvariants (PRION cannot modify threshold-MPC parameters, cannot reduce signature requirements, cannot disable the kill switch, cannot alter the founder-veto timeline), multi-classifier majority-of-two requirement, and a 2/3 DAO emergency kill switch that pauses all PRION-originated patches without affecting human-authored governance.

Current state (Stage 1): PRIONPatchGateway and PRIONInvariants are live on Arbitrum Sepolia and authority-free — no PRION principal is provisioned to originate patches. The full four-gate pipeline ships across Stage 4. See whitepaper §4.1 and §13.10.

A legitimate court order is submitted as a signed petition to an on-chain contract. Validators vote on jurisdiction validity, legal sufficiency, and proportionality. On quorum approval, a FROST threshold-signing ceremony across the operator cohort produces an aggregated signature authorising decryption of exactly the encrypted state named in the order — nothing else. The master key is never reconstructed; partial signatures are combined into a final signature without the key existing at any single host.

No single jurisdiction can compel compliance. No single entity can abuse the process. Mass-surveillance petitions are structurally rejected. All approved disclosures are publicly logged on-chain.

Freelance income, cross-border remittances, business privacy, savings protection in unstable jurisdictions, whale discretion. Sunima protects the ~80% of global wealth that lives in the gray market. We do not surveil. We do not flag. We do not score users — we score transactions.

OFAC sanctioned addresses, ransomware payments, terror financing per UN lists, Lazarus Group operations. Threshold for flagging is high and based on objective criteria, not subjective judgment. Only black-market flags can produce a court-order disclosure petition.

You will not find headshots, biographies, or "about us" pages here. The protocol is the team. The whitepaper is the resume. The audited contracts are the credentials.

We do not have a CMO, a head of growth, or a community manager. Every claim on this site is backed by code, cryptography, or formal verification. If you cannot prove it mathematically, it does not belong in Sunima.

Sunima is designed so that nobody — not the founders, not the validators, not the users — has unilateral control over user funds. Trust is distributed across mathematics, geography, and physics. People are interchangeable. The protocol is forever.

You want privacy without becoming legally radioactive. Stage 1 vault is live on Arbitrum Sepolia today — testnet only, but the production logic is the same: deposit, hold, transfer, withdraw with on-chain encrypted balances. No KYC at the protocol level. Disclosure happens only when a court order is formally served and validated by the threshold ceremony.

What to do: try the vault on testnet, read the disclosure model in the whitepaper, follow mainnet timing.

Try Stage 1 Vault →

You want to build private apps without owning the encryption stack. The sovereign network runs confidential transfers on Halo2-KZG shielded notes with a threshold-ElGamal disclosure committee — it carries no TFHE. The EVM research stages on Arbitrum Sepolia additionally proved homomorphic balance arithmetic on-chain and TFHE predicates via an off-chain worker; those contracts stay published as reference, not as a dependency.

Concrete numbers (measured on-chain):
• EC-ElGamal addCiphertexts: 27,000 gas on-chain (BN254 via EVM ecAdd/ecMul precompiles)
• TFHE predicates (range-proof, compare, conditional): 50–220 ms per predicate on the off-chain worker (tfhe-rs v1.6.0; CPU x86 / GPU)
• Arbitrum Sepolia block time: 1–2 s
• End-to-end encrypted transfer on the live Stage 2 vault: completes inside one block
• No client-side ZK proof generation today — privacy ops run as ordinary signed transactions, no 10–30 s proof wait

Where Sunima fits well: lending, privacy-preserving batch swaps (auction-style settlement), long-hold portfolio management, recurring payments, payroll, B2B settlement.

Where it doesn't: atomic cross-protocol arbitrage in the millisecond window, MEV-bot competition, high-frequency trading at the speed layer. Encrypted state costs time by physics: on the sovereign network a shielded transfer means building a zero-knowledge proof in the browser before the chain will take it, and in the EVM research stages TFHE predicates cost more than cleartext operations.

What to do: read architecture, read whitepaper §6 on cryptography, see live numbers in benchmarks.

GitHub · Closed source today

The bet is positional, not technical. Privacy chains exist; compliance chains exist; legal-bridged privacy — encrypted-by-default with a narrow, threshold-gated, court-orderable disclosure path — has not. Sunima is composed from established primitives (Penumbra-class shielded design over Halo2-KZG, threshold ElGamal for disclosure, FROST threshold signing; Paillier and TFHE belong to the closed EVM research stages and are not part of the sovereign network); the original contributions are at the protocol layer (matryoshka fortress, governance with founder veto encoded as an immutable on-chain expiry timestamp, slashing with mercy curves). 100M fixed supply. 77% open distribution (17% users + 26% ecosystem + 34% protocol) / 23% insiders. No institutional VC, no board seats.

What to do: read the whitepaper end-to-end, study tokenomics below, contact via the channels in the whitepaper.

Read Whitepaper v1.6 →

10,000,000 SUNA goes to the community that built this with us — Stage 1 testnet depositors, auditors, Cosmos privacy advocates, builders. 10% unlocks at TGE under a sell-rate cap, remaining 90% vests linearly over 12 months. Bot farming excluded by the Sentinel stack.

Insider allocation capped at 23% total: 15% team, 8% angel/seed investors via tokens-only terms with 5% TGE unlock after mainnet, 36-month linear vesting, zero governance rights, and no board seats. No institutional VC rounds. No "strategic partners." No anonymous shilling — community advocates engaged through transparent performance-vested ecosystem grants, fully on-chain. Benchmark: Starknet ~38% insiders, Arbitrum ~44%, Sunima 23% — roughly half the insider weight of comparable Layer-1 launches, with 0% institutional VC.

17% supply (17,000,000 SUNA) distributed linearly over 7 years to validator operators, then sustained by transaction fees. Permissionless, global, no licensing. Distribution across the validator set makes the network censorship-resistant. After Year 7, fees plus a demand-contingent APR-floor from the strategic reserve sustain validator economics.

5% supply (5,000,000 SUNA) for third-party builders constructing DEXes, lending, structured products on Sunima primitives. Milestone-gated over 3 years. Sunima builds the cryptography; the ecosystem builds what users want.

6% (6,000,000 SUNA) seeds Hyperliquid spot, Osmosis (Cosmos IBC native), and Uniswap under a 2-year lock. No centralized exchange listings by design — CEXes require KYC and custody, which conflicts with Sunima's self-custody doctrine. DAO-controlled migration allowed.

Under ~2% of supply (~2,000,000 SUNA) can be sold on TGE day. ~98% locked across vesting schedules, DAO reserves, and emission pools — team on a 4-year vest with a 1-year cliff, seed locked until mainnet. Airdrop claim window: 12 months with a sell-rate cap. Staking bonus +20% for users who lock for 12 months post-claim. Structural anti-dump.

SunimaVault with Paillier additive homomorphic encryption. 2048-bit keys, on-chain homomorphic addition, single-organizer MVP. Contract 0xe525…7f86. Try it at sunima.uk/vault.

● COMPLETE on Arbitrum Sepolia (April 2026). EC-ElGamal on BN254 for balance arithmetic via the EVM-native ecAdd/ecMul precompiles — proof-of-concept encrypted-custody primitive built on the Stage 1 Paillier reference. Vault v0.7.1 at 0x3a0d9A52416847D984a17D0dBe849cdF6492209E, source-verified. Measured live: cold deposit 151,368 gas, withdraw 88,789 gas, encrypted transfer 222,318 gas — 12–18× cheaper than the Stage 1 Paillier vault. Vault requires a valid Stage 3 range-proof attestation on every withdrawal; guardian role with circuit breaker (40% TVL / 10 min); organizer rotation secured by a 3-day timelock. The full forge suite passes across the live-stage suite, no regressions.

● COMPLETE on Arbitrum Sepolia (May 2026). On-chain TFHE attestation verifier — encrypted-by-default predicates that additive HE cannot express. SunimaTFHEPredicatesV3 at 0x1840e8E2EF53e8AaA846941aa628d7bd70E27396, source-verified, exposes verifyAndConsumeRangeProof, verifyAndConsumeCompare and verifyAndConsumeConditional on-chain. Off-chain Rust worker on Zama tfhe-rs v1.6.0 signs EIP-191 attestations (range / compare / conditional); verifier replay-protects per (consumer, ctHash, chainid, verifier, validUntil). Measured: ~33K gas median per verify-and-consume. Stage 2 vault v0.7.1 binds to v3 for live withdrawal attestation. Passed internal cold-eye review across 2 rounds. The full Foundry suite passes, no regressions.

● Privacy deposit layer live on Arbitrum Sepolia (April 2026). Note-commitment shielded deposits with commit-reveal front-run protection. Epoch-batched sweeps amortise gas across depositors and widen the anonymity set. Central SunimaHub router aggregates TVL through a thin adapter architecture — each vault type is a ~50-line adapter, audited once. Deposits are unlinkable: withdrawal claims originate from a fresh address using a secret and nullifier held only by the depositor. All contracts source-verified on Sourcify, 0 confirmed findings ≥ MEDIUM.

● Threshold-MPC PoC verified end-to-end (May 2026). FROST 2-of-3 (RFC 9591) — threshold Schnorr signing across two independent trust domains, auto-failover under 200 ms, on-chain _verifySchnorr acceptance confirmed on Arbitrum Sepolia. Numbers and falsification checklist → The PoC nodes are operator-controlled; the validator programme distributes share custody to independent operators across jurisdictions — no shared administrative access, no single party with signing power alone. Court orders flow through a two-window pipeline: 24 h review, then 24 h on-chain challenge. Final committee size set when the validator programme ships.

● Remaining: sovereign chain coordinator standing up the testnet under Sunima governance rather than borrowing Arbitrum, and validator-programme distribution of MPC share custody. Stage 4 closes when both are production-grade, then Stage 5 mainnet begins.

● Sunima Chain L1 testnet running (May 2026). chain-id: sunima_8081-2 · Cosmos SDK + CometBFT · Active validator set across distributed jurisdictions. Operator module x/operators live (V4.2 family — bond, probation pool, 50/50 reward intercept, staking-hook auto-link, ante denylist, stale-link prune). Court-order module x/courtorder live on the testnet binary. Operator onboarding by application during testnet phase — apply via docs.sunima.uk.

Mainnet: Halo2-KZG shielded notes (protocol-level, not EVM contracts). $SUNA gas token. Court Order Gateway at genesis. Ships when audits, the threshold-MPC operator cohort, and the validator set are production-ready — not before. We do not announce dates; we deploy artifacts.

Technical specifications, architecture diagrams, and cryptographic foundations are in the whitepaper. Whitepaper v1.6 →

Correct for full TFHE on every operation — which is why the sovereign network does not use it at all. Confidential transfers there are Halo2-KZG shielded notes: the proof is built in the browser and the chain only verifies it. The EVM research stages made the same point a different way: balance arithmetic on EC-ElGamal over BN254 was 12–18× cheaper than Paillier on cold deposit, withdraw and encrypted transfer (measured live on Arbitrum Sepolia), and TFHE was reserved for the predicates additive HE cannot express.

Compliance is built into the protocol, not into the team. Disclosure happens via a FROST k-of-N threshold-signing ceremony triggered by a court order in a recognised jurisdiction — no operator, anonymous or not, can unilaterally decrypt, and the master key is never reconstructed during the ceremony. The team's identity is orthogonal to the disclosure surface.

A 24-hour judicial review window plus a 24-hour public challenge window (full timeline in whitepaper §7). If the order is not withdrawn, validators participate in the threshold ceremony to recover the specific encrypted state named in the order — and only that state. Mass disclosure is mathematically impossible: there is no master key.

Stage 1–2 ship as Arbitrum L2 contracts precisely because the sovereign chain is not needed yet. Stage 5 introduces a Cosmos-SDK / CometBFT zone because shielded-pool throughput, validator-set economics, and the threshold-decryption protocol need governance-level changes that cannot be expressed inside a generic EVM contract. The chain is the consequence, not the premise.

It is not. PRION proposes patches; humans plus multisig deploy. Every PRION-originated patch passes a four-gate pipeline — sandbox regression, independent re-audit, 3-of-5 Sentinel Committee, and an on-chain timelock — before it can touch a layer. PRION's training environment is a replica fortress, never production state. Behavioural invariants on PRIONInvariants prevent it from ever modifying threshold-MPC parameters, signature requirements, the kill switch, or founder-veto timing — even by a 3-of-5 sentinel quorum. There is no autonomous patch path.

Arbitrum Sepolia served as the Solidity contract testbed and PRION (Sunima's AI auditor) proving ground — Stages 1–4 contracts deployed and source-verified there before sovereign chain migration.

Shipped (Stages 1–3 closed on Arbitrum Sepolia): Stage 1 — SunimaVault Paillier reference vault at 0xe5254A5E…7f86 (live, transacting); $SUNA TGE pipeline (six contracts: token, SunimaVesting v2, SunimaGovernance v1.0.2, SunimaRaiseV2, slashing, fee router) — full testbed supply minted = MAX_SUPPLY (1B on Sepolia; sovereign mainnet supply is 100,000,000 fixed), founder isMinter revoked; PRIONInvariants v1.0.2 and PRIONPatchGateway v1.0.5 deployed, source-verified, authority-free (no principal bound; full four-gate pipeline ships across Stage 4). Stage 2 — SunimaVaultElGamal v0.7.1 at 0x3a0d9A…209E, source-verified, 12–18× cheaper than Paillier on cold deposit, withdraw, and encrypted transfer; guardian + circuit breaker live. Stage 3 — SunimaTFHEPredicatesV3 at 0x1840e8E2…7396, source-verified, on-chain encrypted range / compare / conditional verify-and-consume at ~33K gas median; Stage 2 vault binds to v3 for production withdrawal attestation. Stage 4 Part 1 — privacy deposit layer live: SunimaHub + SunimaCollectionPool + SunimaDefWallet, note-commitment shielded deposits, epoch batching, source-verified. FROST 2-of-3 threshold-MPC PoC verified end-to-end on Arbitrum Sepolia (May 2026) — auto-failover under 200 ms, on-chain Schnorr acceptance. CourtOrderGateway v1.0.0 at 0x01fea626…da039 — first V3 consumer, the on-chain implementation of the legal-bridged disclosure path (worker binding-signature endpoint in development). The full forge suite passes, no regressions. Stage 5 testnet LIVE 2026-05-20 (clean regenesis): sovereign Cosmos EVM hybrid chain — Cosmos SDK + CometBFT consensus + canonical cosmos/evm module for EVM execution — chain-id sunima_8081-2; public RPC at sunima.uk/chain-rpc/, REST at sunima.uk/chain-rest/, network dashboard at sunima.uk/network. Roadmap: Stage 4 remainder — sovereign chain coordinator and validator-programme distribution of MPC share custody. Stage 5 remainder — hardening the Halo2-KZG privacy layer for mainnet (confidential transfers already live on the public testnet), validator decentralisation, external crypto audit before mainnet. The whitepaper marks each section's status explicitly.