JB

AI approves, pays, decides and ships. Now every action can come with proof. Here's what I've been building 👇 nexart.io nitter.gq/Nexart_io/status/21057…

The most expensive AI failure is the one that leaves no record. An agent screens a payment and raises no alert. A KYC check comes back clean. A claims agent finds no reason to escalate. Each of those “nothing” outcomes is a decision. Most stacks record it as an absence. When the payment turns out to be fraud, the first question isn’t what the agent decided. It’s whether it actually looked. Did the sanctions lookup run, or time out? Did the model get the full transaction, or a truncated one? Did a fallback model answer because the primary was rate-limited? A log of successes can’t tell you. A silent failure and a clean pass look identical. The fix is unglamorous: every run produces a record, including the ones where a source refused, a call failed or the agent did nothing. And that record is sealed when it’s made, so “checked and found nothing” can’t be written in later. When one of your agents says “nothing to report”, what would you show to prove it actually checked? That’s the question I’m building NexArt around: records of what an AI system actually executed, verifiable by someone who doesn’t have to trust you.

How to turn Amazon Bedrock Converse calls into Certified Execution Records, verify them locally, protect sensitive values and add optional Node attestation nitter.gq/Nexart_io/status/21128…

EN 18286 Is Now the First Harmonised AI Act Standard. Where Execution Evidence Fits in Your QMS

I'm building with Builders of X 🚀 Find builders from around the world: buildersofx.com/builder/arro…

NexArt is the AI evidence layer, but asking teams to rebuild their stack isn’t realistic. So I used @Replit to build bridge.nexart.io: bring your existing JSON, map it, and turn it into a verifiable Certified Execution Record. Replit helped me ship faster. Thanks @amasad!

Most teams already have the data to prove what their systems did. It's just stuck in logs nobody outside can trust. Evidence Bridge fixes that without asking anyone to rebuild anything. Really proud of this one.

Really happy to have @lmfresneda and HAGGIN on @Nexart_io The fun part of certifying a deal you're in: I can't edit it either. Not Luis, not me, not NexArt No SDK, no rebuild. Every handshake goes through the NexArt Bridge and comes out sealed A handshake used to be a promise. Now it's also proof Here's to the next thousand

Started another round of training ….

Just adjusted my 5 years plan to a 1 year plan for @Nexart_io

Logs describe. CERs prove. Tamper-evident records of what your AI actually executed, verifiable without an account. Nexart.io @Nexart_io