While I was digging into @OpenGradient , I started thinking about something pretty basic: receipts.
When a cashier hands you exactly the cash you were expecting, almost nobody pays attention to the receipt. Most people only start looking for proof when something goes sideways.
For some reason, that comparison popped into my mind while trying to grasp OpenGradient's approach to AI verification.
At first, I assumed that inference and verification happened almost simultaneously. The model spits out a response, the proof shows up, and that's that.
But the more I thought about it, the less clear it seemed.
Markets tend to move fast. Orders get executed, positions shift, and systems make decisions in a matter of seconds. If verification comes in later, even just a moment after, who takes the hit during that gap?
This isn’t a critique. It’s a question that strikes me as interesting.
We often talk about whether a response can be verified or not, but maybe it also matters when that verification comes in and how it impacts the applications relying on it.
I used to think the crucial question was whether there was verifiable proof.
Now I'm starting to think that the time it takes to arrive could also be part of the discussion.
What do you think will be more important for adoption: the existence of verifiable proofs or the speed at which they can be generated?
@OpenGradient $OPG #OPG
When a cashier hands you exactly the cash you were expecting, almost nobody pays attention to the receipt. Most people only start looking for proof when something goes sideways.
For some reason, that comparison popped into my mind while trying to grasp OpenGradient's approach to AI verification.
At first, I assumed that inference and verification happened almost simultaneously. The model spits out a response, the proof shows up, and that's that.
But the more I thought about it, the less clear it seemed.
Markets tend to move fast. Orders get executed, positions shift, and systems make decisions in a matter of seconds. If verification comes in later, even just a moment after, who takes the hit during that gap?
This isn’t a critique. It’s a question that strikes me as interesting.
We often talk about whether a response can be verified or not, but maybe it also matters when that verification comes in and how it impacts the applications relying on it.
I used to think the crucial question was whether there was verifiable proof.
Now I'm starting to think that the time it takes to arrive could also be part of the discussion.
What do you think will be more important for adoption: the existence of verifiable proofs or the speed at which they can be generated?
@OpenGradient $OPG #OPG