Named gates. Three-gauntlet validation. Retractions on-chain. Compose them into your workflow.
X3-RONIN is the intelligence layer your trading workflow plugs into. Synthesize 30 data streams into one verdict. Cite the named gates that cleared. Inherit our calibration bands, persona library, and retraction discipline. Build your strategy on rigor you didn't have to invent. Bring your venue. Bring your data. We bring the discipline.
Every claim verifiable. Status: x3lab.io/audit#status · Manifesto: x3lab.io/about
ON-CHAIN AUDIT live on Base mainnet → L6 PUBLIC RETRACTIONS /retractions →
Each loop is a composable primitive. Query the verdict. License the gates. Federate the validation. Build your workflow on top — your sizing, your execution, your edge. Bring your data. Inherit our discipline. Every customer's outcomes make the engine smarter for the next.
Template prompts turn "does retail funding-flip predict 4h reversals?" into SQL, a cohort, a metric, a backtest harness.
a half-thought → a tested hypothesis · ≈ 1 hour
PLUG IN Query our hypothesis catalog. Submit your own — they enter the validation pipeline.
Funding · OI · whale flow · macro digest · news · chart structure · past setups — 30 streams compressed into one decision frame.
30 streams → 1 decision frame · live, named-gate evidence
PLUG IN Verdict API + WebSocket. License the meta-label models for self-hosted use.
Three independent gauntlets — forward replay with walk-forward + bootstrap CI, Freqtrade live-paper against real exchange tape, and external probes on Hyperliquid wallets + on-chain flow. Independent data pipelines, no shared upstream artifact.
playback · live-paper · wallet probes · 3-of-3 to promote
PLUG IN Run our three gauntlets on your strategy. Federated TEE compute keeps your data private.
Survivors become named gates with explicit contracts. Trades fire only when N gates clear, sized to your loss limit, gated by a confirmation token. 112 MCP tools dispatch the actual probes, fills, and management.
propose → confirm → execute · per your rules
PLUG IN Receive named-gate verdicts. Your sizing, your execution, your edge stays yours.
Every trade carries the gates that fired, the contracts they signed, and the retraction history. When a gate stops working, the retraction is a public git commit. Audit your LP can read. Audit your future self will thank you for.
gates fired · contracts signed · retractions in git
PLUG IN Subscribe to retraction events on Base. Build dashboards. Build whatever you want on top.
Every customer's outcomes flow back as anonymized calibration data. Gates that fail in production get demoted faster. Calibration bands re-tune. Persona library grows. The network gets smarter with every plug-in.
outcomes → calibration drift → next verdict is better
PLUG IN Your outcomes (anonymized) train the next verdict. Network effect compounds; opt-out anytime.
A gate is a falsifiable contract — a measurable claim with a threshold and a unit. If it stops working, we say so out loud, in a public git commit, and the gate dies. Below is the live matrix for the current setup window: nine pass, two hold, one fail.
An edge that cannot be measured cannot be cited.
validated_edges.json :: methodology.iron_law
Don't build a trading system from scratch. Compose it from rigor. Bring your venue, your data, your edge. Inherit our gates, calibration, and retraction discipline.
Query our verdict API with your setup. Get named gates, calibration bands, comparable historical setups. Plug into your existing workflow in 10 lines.
License the gate definitions, validation harness, and meta-label models. Run on your own data, your own infra. Quarterly calibration updates. Your derived strategy stays yours.
Confidential compute in a TEE (Phala/Marlin). Upload your raw data; our algorithms run in an enclave; output returns to you. We never see your data. You never see our model weights.
Every verdict carries its on-chain proof. Every retraction is a public git commit. Your strategy stays your IP — but every primitive it depends on is verifiable from any Base node.
Every retracted edge is a transaction on Base. Every transaction is a permanent block. Anyone reads the full history from any node — without trusting us, our git, or any custodian. The cryptographic anchor is the proof.
| Edge | Threshold spec | Status | Last update | Tx |
|---|---|---|---|---|
| L5_SMOKE_TEST_2026-05-23 | L5 on-chain anchor smoke test · canonical_hash 32B | anchored | 2026-05-23 UTC | 0x2a0c507e…670a7fd3 |
Source: services/onchain_anchor.py — direct anchoring via tx data field. No contract code, no admin keys, no upgradability — by design. Live on Base mainnet. Every signal's canonical hash is immutable.
Falsifiable edges. Proofs in git. Built by a studio that runs its own rules before shipping them.
Register your strategies, data sources, and custom gates as MCP tools. They land at read tier — referenced by trades, never able to fire one without an operator-curated promotion. Every call audited. Every gate retractable. Tenant-scoped end to end.
Exchange APIs, wallets, webhooks. Encrypted, tenant-scoped, idempotency-keyed. Already ships.
Custom MCP tools register at read tier. Referenced by the gate matrix. Cannot fire trades — exec tier is operator-curated only.
Custom orchestrators in an isolated runtime with egress allowlist. Coming after the matrix proves itself in production.
Nobody owns it yet.
