v1.0.0-BETA

Introduction to ISNAD

ISNAD is the decentralized trust layer for AI resources. The protocol provides a cryptographic framework to ensure that AI code behaves exactly as intended, backed by economic incentives and on-chain audits.

How it Works

The core of the protocol relies on Proof-of-Stake Attestation. When a resource is inscribed on-chain, auditors can stake $ISNAD tokens to vouch for its safety. The more stake behind a resource, the higher its trust score.

01// Check trust score before loading a skill
02const trust = await isnad.getTrust(skillHash);
03 
04if (trust.tier === "VERIFIED") {
05 // Safe to execute
06 await skill.execute(params);
07} else {
08 console.warn("Unverified skill - proceed with caution");
09}

Trust Tiers

Resources are classified into tiers based on total weighted stake:

TierMin StakeMeaning
UNVERIFIED0No auditor attestations
COMMUNITY100 $ISNADSome community trust
VERIFIED1,000 $ISNADMultiple auditors with significant stake
TRUSTED10,000 $ISNADHigh-confidence, heavily audited

Key Concepts

Resources

Any content-addressable AI artifact: skills, configs, prompts, memory, models, APIs. Resources are inscribed on Base L2 with a SHA-256 content hash for permanent verification.

Attestations

When an auditor stakes tokens on a resource, they create an attestation. Each attestation includes the stake amount and lock duration (30-90 days). Longer locks earn higher multipliers.

Slashing

If a resource is found to be malicious, any user can flag it. A jury of randomly selected auditors reviews the evidence. If found guilty, all stakes on that resource are burned.

Yield

Auditors who stake on clean resources earn yield from the reward pool. Yield scales with stake amount and lock duration—incentivizing longer commitments.

Next Steps