Last week I was testing a multi-step onchain workflow that involved moving assets across three different applications. The transaction itself was fast. The approval process wasn't.
What caught my attention wasn't execution speed. It was how much time was spent deciding whether an action should happen at all.
I tracked 27 actions during the test. Only 9 actually required value transfer. The other 18 were authorization decisions. Can this wallet interact? Can this agent execute? Is this permission still valid? Should this transaction proceed under these conditions?
That ratio felt backward.
The more automation gets introduced, the more these approval checks start piling up. One autonomous agent generated 43 transaction requests over a few hours. Most were harmless. A few needed limits. One probably should have been blocked entirely.
The interesting part was watching how Newton approached the problem.
Instead of treating authorization as a small step before execution, it felt like authorization was becoming its own operational layer. A separate decision engine sitting between intent and action.
I kept looking at the logs afterward.
The transactions themselves weren't the valuable data.
The approvals were.
Who requested what. Under which conditions. Why it was allowed. Why it was rejected.
Those records started telling a more useful story than the transfers.
If autonomous finance keeps moving toward agent-driven activity, I suspect transaction volume won't be the bottleneck people worry about.
The bottleneck might be decision volume.
When hundreds or thousands of actions need approval every day, the infrastructure that manages authorization may end up becoming more important than the infrastructure that moves assets.
Still testing that assumption. But the numbers keep pointing in the same direction.
@NewtonProtocol $NEWT #Newt .
What caught my attention wasn't execution speed. It was how much time was spent deciding whether an action should happen at all.
I tracked 27 actions during the test. Only 9 actually required value transfer. The other 18 were authorization decisions. Can this wallet interact? Can this agent execute? Is this permission still valid? Should this transaction proceed under these conditions?
That ratio felt backward.
The more automation gets introduced, the more these approval checks start piling up. One autonomous agent generated 43 transaction requests over a few hours. Most were harmless. A few needed limits. One probably should have been blocked entirely.
The interesting part was watching how Newton approached the problem.
Instead of treating authorization as a small step before execution, it felt like authorization was becoming its own operational layer. A separate decision engine sitting between intent and action.
I kept looking at the logs afterward.
The transactions themselves weren't the valuable data.
The approvals were.
Who requested what. Under which conditions. Why it was allowed. Why it was rejected.
Those records started telling a more useful story than the transfers.
If autonomous finance keeps moving toward agent-driven activity, I suspect transaction volume won't be the bottleneck people worry about.
The bottleneck might be decision volume.
When hundreds or thousands of actions need approval every day, the infrastructure that manages authorization may end up becoming more important than the infrastructure that moves assets.
Still testing that assumption. But the numbers keep pointing in the same direction.
@NewtonProtocol $NEWT #Newt .