World Agent is a research project as well as a platform. The architecture rests on formal foundations that make the system's claims provable rather than asserted, including theorems on trust composition, origin discipline, hierarchical complexity, and the dynamics of platform-mediated place quality. Selected work is published openly; deeper technical material, including full proofs and implementation specifications, is available to qualified researchers and partners on request.
The public specification covers the platform's posture, the canonical question set, the two-layer truth model, the agent hierarchy, and an outline of the formal foundations. The full mathematical treatment, proofs, and implementation specifications live in v9 (available on request). Suitable for technical reviewers, journalists, and researchers establishing context on the system.
@techreport{lorenzini2026worldagent_public,
author = {Lorenzini, Dave},
title = {World Agent: A Spatial Intelligence Layer
for AI Systems},
subtitle = {Public Specification},
year = {2026},
number = {v1.3-public},
institution = {World Agent LLC}
}
The full technical specification including the complete mathematical foundations, theorem proofs, the agent-consumable manifest, schemas, and detailed implementation specifications. Available to qualified academic researchers, technical partners, and reviewers under confidentiality agreement.
A vertical-specific treatment showing how the World Agent architecture applies to real estate. Covers the canonical question set in real-estate idiom, vertical geocoding at property-relevant scales, multi-source consensus across real-estate data providers, cohort contestation in neighborhood experience, and the privacy posture required for mortgage-lender review. Available to qualified partners under NDA.
Upcoming talks at AWE 2026 and selected technical venues. Essays published on the spatial-intelligence-for-AI thesis as they appear. Subscribe below for research updates.
World Agent welcomes research collaboration with academic groups working on spatial AI, world models, trust and provenance, and cohort-conditioned reasoning. The framework's open problems include the empirical dynamics of platform-mediated place quality, robustness against coordinated inauthentic activity in trust pipelines, and provenance discipline in mixed-source spatial data.
Contact for research collaboration →This document is provided in confidence to qualified researchers, technical partners, and analysts. Each request is reviewed individually. Approved requesters receive the document by email along with a short confidentiality acknowledgment. Please provide accurate information; requests with incomplete or unverifiable details are not approved.
Request received.
We review requests within 2-3 business days; approved requesters will receive the document by email, along with a short confidentiality acknowledgment to confirm. Please do not resubmit; we have your request.