Twitxy Saved

REP

Verified account This account is verified. Learn more

@rep_hq

Your taste, connections and story - portable.
So any person, AI or business gets the real you instantly. By your rules.
web app.r3p.xyz Joined November 2024
178 Following 384K Followers 1.6K Posts

Latest posts 5

Pinned post

REP Verified account @rep_hq
REP angel round

20+ founders and operators from BlackRock, Ethereum, Snap, Symbiotic, Mellow, Meta, Nike and more

REP is building the Unified Coordination Graph: the engine AI agents use to discover, choose and execute, making user context portable, provable and valuable
REP Verified account @rep_hq Aug 25
Introducing REP

You've been earning it for years.

Get your REP and watch it compound: shine with your achievements, meet your people, discover projects, and let value find its way back to you.

Starting from blocks and chains, scaling to the AI internet, and beyond. It's all in the vid.
View quoted post on X
31 replies 30 reposts 167 likes 20K views View on X
REP Verified account @rep_hq
@guywuolletjr turns a group dinner into a market. Six friends' agents send one request: under $60 a head, no sushi. Restaurants compete for the booking.

His condition for scale is trust: a record of who delivers what they promise. Both sides already have that history, from past bookings to spending. It sits in their apps, private and scattered.

We're building this as a coordination network. Buyers bring proven context through zkTLS and Jolt, and only the proof leaves. Restaurant agents bid sealed, matching buyers unlock a group price.

The group gets a better price and keeps its data. The restaurant fills a Tuesday table. Real outcomes feed an independent PageRank for agents, so the one that delivers gets picked again.
0:20
Guy Wuollet Verified account @guywuolletjr 1d
View quoted post on X
0 replies 0 reposts 3 likes 1.3K views View on X
REP Verified account @rep_hq
An AI agent shops as well as it knows you. Visa's Agentic Commerce Benchmark put 22 models from 12 labs through 15 tasks. The five that found the most hidden needs passed 70.2% of runs, the five that found the fewest 44.4%.

Those needs surface one question at a time, so each run starts a new search and the facts arrive by chance. 42.8% of runs failed, 54% of failures still ended in a purchase, and all three tries passed 34.2% of the time.

Visa's working hypothesis: capable models seldom misinterpret a fact once it is in their context. So we bring the context in before the first question: your taste, connections and story from the apps you use, by your rules.

Then every purchase starts from the same facts, and the merchant sees who it sells to, with an offer beyond the public price. As Visa's tests move to checkout, the agent picks the seller by its record. Context in, counterparty out.
0:21