This is exactly the scaling problem. Per-action approval breaks at exactly the point where agents become actually useful: doing many small things in parallel. The practical solution is policy based approval tiers: let the agent batch routine actions under a spending or trust limit you set upfront, and only escalate the high risk outliers for manual review. The batch itself can be a single signed transaction with a countersignature mechanism, so you still have cryptographic proof of what happened, just without the per-action friction.
Exactly — and that's where batching is meant to lead. Per-action approval doesn't scale, but removing approval entirely isn't the answer either. The batch is the review surface for the policy: you approve the class of actions once, with the list in front of you, instead of rubber-stamping each one as it flies past. Policy decides what's allowed; the batch is where a human still sees what actually ran. Thanks for reading it closely.
Cai
This is exactly the scaling problem. Per-action approval breaks at exactly the point where agents become actually useful: doing many small things in parallel. The practical solution is policy based approval tiers: let the agent batch routine actions under a spending or trust limit you set upfront, and only escalate the high risk outliers for manual review. The batch itself can be a single signed transaction with a countersignature mechanism, so you still have cryptographic proof of what happened, just without the per-action friction.