TEST
Calibration in progress · L1 chains live · L2 validated
Invarians is in early access. The SDK on PyPI is Alpha (pip install invarians, breaking changes possible across minor versions). The signing pipeline runs on a single node with HMAC-SHA256, on-chain anchor reserved for May 2026 deployment. SLA is best-effort until Phase 02 (hybrid model on Chainlink Functions DON, Q3-Q4 2026).
The instrument measures and certifies blockchain execution conditions in real time across L1 chains and L2 rollups, continuously. Each chain undergoes independent validation against known historical events before being considered production-ready.
Developers can test the API freely. The data is real, the signal is live. Use it to build and evaluate. Production institutional settlement is a Phase 02-03 target, not appropriate for today's centralized pipeline.
| Chain | Layer | Estimated Reliability | Status |
|---|---|---|---|
| Ethereum | L1 | ~99% | Validated |
| Solana | L1 | Structural (τ) validated · Execution profile (π) pending, ~90% overall | Partial |
| Polygon | L1 | ~95%, backtested on 4 historical events (TPR 100%, FPR 14.57%) | Validated |
| Avalanche | L1 | ~80%, internal data only · no public validation dataset available | In calibration |
| Arbitrum · Base · Optimism | L2 | Execution profile calibrated (statistical v3, 30 days) · event-detection pending (Phase D, Q2-Q3 2026) | Validated |
Reliability estimates reflect the instrument's expected precision under normal operating conditions, the probability that a declared regime corresponds to the actual infrastructure regime. These figures are updated as calibration progresses. Ethereum is the reference chain. Solana execution profile (π) and Avalanche thresholds are pending external validation, treat those signals as indicative until updated.
| Family | Count | Signals | Status |
|---|---|---|---|
| Circle CCTP | 10 routes | Per-message capture with Circle ECDSA signature (crypto.anchor: circle_ecdsa), independently verifiable against Circle's published attester public key. Real per-message latency P90/P99. |
per_message_attested |
| Chainlink CCIP | 10 lanes | Per-message capture via bytes32 messageId matching (source send ↔ destination execute). Real send-to-execute latency P90. RMN cursed override remains absolute. crypto.anchor is null today. |
per_message_attested |
Bridge classification scope: variable-latency surfaces only (Chainlink CCIP, Circle CCTP). All 20 lanes present in panel.bridges[]. CCTP entries expose state:"BS1"|"BS2" with calibrated:true (preliminary, confidence LOW). CCIP entries expose raw signals with state:null and calibrated:false. Safety overrides (RMN cursed, Iris down) fire independently of calibration.
- Dashboard, no key required
- API · 20 requests / day
- 4 L1 chains: ETH · POL · SOL · AVAX (SOL/AVAX calibrated July 2026)
- 3 L2 chains: ARB · BASE · OP (validated)
- 12 signed regime codes per chain on L1 and L2 (S1, S2+, S2- × D1, D2+, D2-, D2±)
- Axis-grouped panel:
structural(rhythm · continuity · beacon_participation on ETH · sequencer_publish_latency on L2) anddemand(sigma · size · tx · complexity · gas_complexity on L2) per chain - Delta: per-metric
shift+ compositedriftper axis (Primitive 3) - HMAC-SHA256 attestation, independently verifiable
- Higher rate limit on the V2 panel and verify endpoints
- Direct line to the maintainer for integration and feedback
- Influence on the Phase 02 hybrid model spec (DON computation surface, attestation format)
- Methodology deep dive sessions with reproducibility kit walkthrough
- Co-developed pricing once usage patterns are understood
- Suitable for production-adjacent integrations: risk dashboards, treasury rebalancing, intent fill gating
- On-premise or dedicated infrastructure
- Direct node ingestion
- Chain focus, depth over breadth
- Custom coverage and routing
- Continuity commitment, per contract
- Dedicated integration support
- Priority access to calibration updates
panel.l1[], panel.l2[], panel.bridges[] carries its own status (OK · STALE · UNAVAILABLE · UNCALIBRATED) and timestamp (computed_at for L1/L2, observed_at for bridges). When any item is STALE or UNAVAILABLE, the global oracle_status flips to DEGRADED — do not trust the signal.All decisions remain the responsibility of the operator.