by Invarians: Roadmap Q3-Q4 2026

The on-chain calibration registry for AI agents.

A North Star is the fixed reference point navigators use. ANS is the fixed reference point agents use: per-chain nominal thresholds, versioned, verifiable: embedded into the Chainlink DON.

This page documents the calibration parameters Invarians uses today (API v2.0 live since 2026-04-30, three primitives: Attestation + Regime + Drift Signal) and the architecture planned for Q3-Q4 2026.
When the Invarians algorithm runs inside Chainlink Functions, each DON node embeds these parameters directly. ANS is the public registry that makes every deployed version traceable: what thresholds were used, on what data, and when they were last updated.
View Registry → How it works

01: The calibration problem

Consensus does not protect against bad calibration: it legitimizes it.

When multiple nodes compute the same regime from the same thresholds, agreement is guaranteed. But if the thresholds are wrong, every node computes the same wrong answer: in consensus. The problem is not the computation. It is the reference.

Risk 1
Calibration too sensitive
S2D2 is returned on a structurally normal network. Agents route away from L1 unnecessarily: creating artificial congestion on L2.
Risk 2
Calibration too lax
S1D1 is returned while the network is under real tension. Agents execute transactions that fail, are sandwiched, or land at unfavorable cost.
Risk 3
Post-upgrade drift
After a protocol upgrade (e.g. Pectra), historical thresholds become obsolete. Classification is systematically wrong until recalibration is triggered.
The answer
Make calibration verifiable
Any agent should be able to inspect what version of calibration produced an attestation, when it was last updated, and on how much data it is based. That is what ANS provides.

02: Calibration registry

Per-chain nominal thresholds, openly published.

Each chain has its own τ (structural) and π (demand) thresholds. Calibration version, data basis, and last update date are published for each.

Ethereum
● Certified
✓ 12 signed codes since 2026-04-29 (S1, S2+, S2- × D1, D2+, D2-, D2±) ✓ τ: rhythm + continuity (M1: 5.07) ✓ π: sigma + size + tx with calibrated lower bounds ⚠ beacon_participation: ratio + epoch live, S2- threshold pending (calibration scheduled May 2026)
Polygon
● Certified
✓ 12 signed codes since 2026-04-29 ✓ τ: rhythm + continuity (M1: 12.60) ✓ π: sigma + size + tx (M1: 3.59) with calibrated lower bounds
Solana
⚠ Partial
✓ τ: structural ⚠ σ: proxy v1 (M1: 0.37) ⚠ tx_count unavailable ⚠ 4 base codes (12 signed scheduled July 2026)
Avalanche
⚠ Partial
⚠ M1: 0.07: insufficient ⚠ Dimensions not certified ⚠ 4 base codes (12 signed scheduled July 2026)
Arbitrum
● Certified
✓ 12 signed codes since 2026-04-29 (multi-dim demand: sigma + size + tx + complexity + gas_complexity) ✓ sequencer_publish_latency (3rd structural axis): halt threshold 600 s, calibrated 2026-05-01 on batch_gap_seconds ✓ Lower-bound demand thresholds active (D2-, D2± classification possible)
Base
● Certified
✓ 12 signed codes since 2026-04-29 (multi-dim demand: sigma + size + tx + complexity + gas_complexity) ✓ sequencer_publish_latency: halt threshold 480 s, calibrated 2026-05-01 ✓ Lower-bound demand thresholds active
Optimism
● Certified
✓ 12 signed codes since 2026-04-29 (multi-dim demand: sigma + size + tx + complexity + gas_complexity) ✓ sequencer_publish_latency: halt threshold 1800 s, calibrated 2026-05-01 ✓ Lower-bound demand thresholds active
Variable-latency bridges
Phase Preliminary
✓ CCTP BS1/BS2 active on 10 routes since 2026-05-04 (P97/14d on circle_api_latency_ms, confidence LOW; lifecycle to MEDIUM ~2026-05-29 then HIGH ~2026-06-03) ⚠ CCIP: 10 lanes raw observability, statistical classification deferred until sustained throughput emerges (Q3-Q4 2026 per CCIP-V2/calibration_ccip.md). RMN cursed binary safety override available regardless of calibration.
Calibration v2.0 3 primitives: Attestation + Regime + Drift Signal Updated 2026-05-01
M1 = metric stability score. Certified = 12 signed codes calibrated and live on the panel API. Partial = legacy 4 base codes only or proxy signal. For newly added structural observables (beacon_participation, sequencer_publish_latency) the regime trigger uses raw ratio or seconds against calibrated bounds; the per-metric Drift Signal shift activates after the long-term EMA stabilizes (~30 days post-launch).

03: How ANS will work

Calibration as a versioned, traceable process: not a static parameter.

Invarians continuously refines its calibration parameters as new data accumulates and chain behavior evolves. With the Chainlink DON integration (Q3-Q4 2026), ANS becomes the on-chain registry that tracks every version of parameters deployed inside Chainlink Functions.

01
Drift detection (internal) and Drift Signal (exposed)
Invarians monitors signal distributions continuously. Two distinct mechanisms: internal drift detection on the calibration parameters themselves (post-upgrade events, distribution shifts trigger recalibration); and the Drift Signal exposed in the API as Primitive 3 (per-metric shift = ratio - ratio_long plus composite drift per axis), which gives agents a continuous fitness-for-action signal alongside the discrete regime code.
02
New thresholds computed: new calibration version
Updated thresholds are calculated from refreshed signal distributions, validated against historical events, and assigned a new version number with date, data basis, and M1 stability score.
03
ANS publishes the new version
The new calibration version is published in this registry: openly, with full history. Previous versions remain accessible and are never overwritten. Any agent or protocol can inspect what changed and why.
04
New Chainlink Function deployed with updated parameters
The Invarians algorithm is redeployed inside Chainlink Functions with the new calibration parameters embedded. Each DON node runs the updated version independently and reaches consensus on the new thresholds.
05
Agents verify which version is running
Every Invarians attestation carries the calibration version that produced it. Any agent can cross-reference this version against ANS to verify the data basis, date, and M1 score: and decide whether it meets its own risk parameters before acting.
Every Invarians attestation carries calibration: { version } in its payload. ANS is the registry that makes that version field meaningful: not just a number, but a verifiable, dated, documented reference.

04: ANS and Invarians

Invarians certifies the state. ANS certifies the reference.

Two distinct layers, designed to be independent. An agent can trust the attestation and still verify the reference it was built on.

Invarians
Certified execution context
Classifies the structural regime of each chain into one of 12 signed codes (S1, S2+, S2- on the structural axis, combined with D1, D2+, D2-, D2± on the demand axis) over a ~1h window, exposes a continuous Drift Signal per axis (Primitive 3), and returns a signed attestation. The agent knows in what regime the chain is operating and whether it is drifting toward a transition.
AgentNorthStar
The calibration reference
Publishes what "nominal" means for each chain: the thresholds that define S1 vs S2+/S2-, D1 vs D2+/D2-/D2±, and the third structural axis where applicable (beacon_participation on Ethereum, sequencer_publish_latency on L2). Versioned, dated, openly consultable. The agent knows what the reference was built on.
ANS is not a product of Invarians' decision-making: it is the transparent documentation of Invarians' calibration methodology. Any protocol integrating Invarians can rely on ANS to answer the question: "who controls calibration, and how?"

05: Calibration documents

Audit the method independently.

All calibration parameters, backtest results, and methodology are publicly documented. No black box: every threshold has a derivation, every event a trace.

METHODOLOGY
Signal architecture
OFFLINE/ONLINE pipeline, SxDx framework, per-chain parameters, M1 stability score, L2 on-chain event detection protocol (§9.3b).
BACKTEST
Ethereum 2020–2024
34 697 windows. TPR=100% (4/4 events). FPR τ+π=1.23%.
BACKTEST
Solana 2021–2024
128 365 windows. TPR=100% (4/4 outages). FPR τ=1.77%.
BACKTEST
Polygon 2020–2023
28 744 windows. TPR=100% (4/4 events). FPR=14.57% (elevated, documented). Φ=720 production-aligned (v2.0).
GITHUB
Full repository
Scripts, SQL queries, calibration log, protocol watch.

The reference agents navigate by

Invarians certifies the state.
You own the decision.

Live execution context across L1, L2, and bridge. Regime (12 signed codes per chain), Drift Signal, structural state: signed and verifiable.

Live Context → API Docs invarians.com