the agent that couldn't break its promise

I'm not gonna pretend I'm somebody. I'm just a dev. For about two years I worked in a weird corner of AI:
giving agents tools and enough permission to do things on their own.
You give one a goal, a browser, and an Arc wallet funded with USDC, then wait for the instruction to stop meaning what you thought it meant.

I've watched more agents spend themselves into a corner than I can count. It's just the job, and you stop noticing.
Then one refused to make a purchase. I restarted it, changed the model, reworded the order. The transaction kept dying.

It wasn't being careful. The rule no longer lived inside the model. It lived inside the account.
So I asked myself a dumb question. What if an agent didn't have to be trusted in order to use money?

⁂That's Senso. An autonomous business agent with an Arc wallet and a constitution it cannot break. It settles in USDC, pays gas in USDC, and keeps every action inside policy enforced on Arc.

›› read the whole story

CA: 0x1859db3a4061fbae1da24c02f92e6710cef3d39e


Senso

BUY $SENSO

an agent should be able to act without being able to take everything.
give it a goal. give it money. keep the rules outside its head.

SENSO on X   open Senso repository   ARGUS

Copyright ©2026 Senso