Urgent.News

What's breaking now, across thousands of outlets.

AI

Building Agentic Commerce on Bitcoin, Without Trusting the Agent

Luma began with a deceptively simple question: Can an AI agent help someone buy something with Bitcoin or an RGB asset—without ever being trusted to spend their money? We think the answer is yes. And after building, breaking, and rebuilding the system across regtest, Signet, wallets, Lightning channels, merchants, and mobile hardware, we’re more convinced than ever that this is an architecture…

Building a trustless commerce system on Bitcoin, without relying on the agent holding the keys, has been a focus for Luma. After iterating through various experiments and configurations, the team has arrived at an architecture they believe is worth pursuing. The core principle of Luma is that agents can understand intent, find merchants, inspect products, read invoices, and prepare transactions, but they cannot change recipient, amount, or asset details after approval.

The node resides within the phone, ensuring that users own their keys, state, and channels. This approach allows for wallet-to-wallet transfers, agent-assisted purchasing, and human-controlled approval before payment execution. While the prototype is functional, there are still several key areas that need further development, including production-level repeatability, merchant-side receipt verification, and loss-and-restore tests.

The team is committed to sharing their findings, starting with the Alice/Bob/Carol prototype and RGB402 experiments, while keeping some aspects proprietary. The goal is to provide a clear, verifiable framework for non-custodial wallets, encouraging others to build upon their work and contribute to the advancement of trustless commerce.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

A dinner planner for my vegetarian friend that doesn't trust its own AI

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I Built My friend is vegetarian, so "what's for dinner?" always comes with a rule: no meat, no fish, and none of…

  • Safe Plate dinner planner for vegetarian friend wary of AI
  • App uses local Gemma 3 1B model from Google for meal suggestions
  • Safety check flags ingredients against user's dietary restrictions

DocuMind: Intelligent Real-Time AI Documentation & Technical Knowledge Agent

This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real Content Transforming static developer documentation and technical publications into an interactive, grounded AI…

  • DocuMind streamlines developer documentation with AI-powered conversational interface
  • Grounded AI reasoning prevents hallucinations by retrieving exact Sanity content
  • Real-time synchronization with DEV.to ensures always-up-to-date knowledge base

More from Sunday 4 October →