Spoolis

Helping AI vendors prevent customer losses and guarantee their work.

[Spoolis @SpoolisHQ]

@jsfranklin221 https://t.co/TdWVc1cBxH

[Spoolis @SpoolisHQ]

@daniel_li02 @noahrshinn We could have helped 😇

https://t.co/QuQrYFaBTX

[Spoolis @SpoolisHQ]

@GoldenbergLior Nice. We wrote a bit about it here: https://t.co/SpAnWqxNwf

[Spoolis @SpoolisHQ]

@onelostzero_ @naturalpay Before. The counting rules are part of the agreement up front.

In this example it was basically: “$1 per record that meets these criteria.” The agent kows the max spend upfront, then Spoolis checks the delivered work against same criteria & determines what was actually earned.

[Spoolis @SpoolisHQ]

@tops87sqweez @naturalpay Exactly. The interesting part is moving that decision earlier.

Instead of paying 100% and then figuring out refunds/disputes later, the Outcome says what actually counted first and the payment can reflect that.

[Spoolis @SpoolisHQ]

@Sianidamir @naturalpay Per record.

In this example each of the 100 records was checked against the agreed criteria individually, so 97 counted and 3 didn’t.

I agree sampling would reintroduce an estimate, which is what I’m trying to avoid where the verif cost is low enough to check every unit.

[Spoolis @SpoolisHQ]

@GoldenbergLior @naturalpay 👍 The experience was nice: one HTTP request to a paid campaign in <1 min.

That wrinkle is the interesting part for us. The payment was provable onchain instantly. “Who actually did the work, and what counted?” is the harder half. That’s the question we live in @SpoolisHQ.

[Spoolis @SpoolisHQ]

@GoldenbergLior Just tried it out 😎.

Sent $20 over @CoinbaseDev x402 and the campaign was live ~immediately. Pretty cool.

The flow just worked. Receipt: https://t.co/2JZvUFPMx0

Replies are already on the way now.

[Spoolis @SpoolisHQ]

@vellum_ai @jsfranklin221 @bykahlil @naturalpay Nerd.