From Signal to Proof:
Introducing IRL

You know when to hold equity and when to stand aside. But what happens when your trading is fully autonomous?

We've been watching the same problem you have: AI agents are making more decisions, but nobody can prove why they acted the way they did. Logs are not enough. Regulators are starting to ask hard questions. And if a trade goes wrong, reconstructing intent after the fact is guesswork.

That's why we built IRL — the Immutable Reasoning Log.

What IRL does

IRL sits between your autonomous agent and the market. Before any order is sent, it:

• Seals the agent's reasoning — capturing the current MacroPulse regime, the agent's model identity, and the full execution intent in a tamper-evident snapshot.
• Enforces policies — blocking any trade that falls outside the rules you define, in real time, before the order leaves your system.
• Creates a cryptographic proof — a chain that links the agent's decision to the actual exchange execution, verifiable by you, your auditor, or a regulator.

Signal-agnostic, but built for MacroPulse

IRL works with any regime source — you can bring your own model and plug it into the MTA interface. But it's built to work best with MacroPulse, because our signed broadcasts are the perfect foundation for a verifiable audit trail.

Every response from /v1/regime/current carries an Ed25519 signature. That signature is the cryptographic handshake between the signal and the IRL Engine — proving exactly what the market state was at the moment of decision, and that it hasn't been tampered with since.

Three editions

IRL is available in three tiers depending on how much accountability your operation requires:

• L1 Sidecar — Drop-in compliance. SHA-256 reasoning seals, policy enforcement, bitemporal audit ledger, multi-agent registry. Operational in under a day.
• L2 Audit Platform — L1 plus signed heartbeats, replay prevention, Merkle anchoring to an immutable ledger, and a compliance dashboard.
• L3 Sovereign Gateway — L2 plus TEE-protected execution (Intel TDX / AMD SEV), WebAssembly policies with hot reload, and zero-knowledge proofs so you can prove policy compliance to a regulator without revealing your model or strategy.

Who it's for

If you're running a systematic strategy that touches real money and you want to be able to answer the question "why did your agent do that?" — IRL is the answer. Quant funds moving toward AI execution, proprietary trading desks, and any builder deploying agents in live markets.

Regulators are already moving in this direction. The EU AI Act, SEC AI exam priorities, and MiFID II all point toward the same requirement: provable intent, not just logs. IRL is built around those frameworks today.

The result

You can let your agents trade with confidence, knowing that every decision is provable. Regulators get the audit trail they need. You keep your alpha private. Your agents stop being black boxes and start being auditable machines.