Living architecture
Turn a changing code estate into an explorable system map—symbols, references, dependencies, runtime seams and the evidence horizon behind every answer.
PX gives engineering and AI teams a living evidence layer for complex systems—so architecture, documentation, change impact and review conclusions stay connected to the source that makes them true.
UnitedHealth Group already operates AI at extraordinary scale. The harder problem is not producing another answer. It is proving which source, dependency, policy and system boundary made that answer defensible—especially after the system changes.
PX combines graph intelligence, semantic retrieval and documentation authority while exposing exactly what it inspected and what remains unknown.
Turn a changing code estate into an explorable system map—symbols, references, dependencies, runtime seams and the evidence horizon behind every answer.
Trace a consequential change across TypeScript, HTTP, Rust, SQL, configuration and operational boundaries without losing the caller or the confidence level.
Generate architecture and workflow explanations from the same evidence graph engineers query—then show when the source has moved beyond the document.
Give engineering, security and Responsible AI reviewers the exact route, artifacts, uncertainty and next verification step—not a prose-only verdict.
Start with one bounded Optum Insight code and documentation estate. Keep it read-only, exclude PHI, run it inside UHG-controlled infrastructure and measure whether evidence actually becomes faster and more reproducible.
Index one estate and freeze the initial evidence horizon.
Use engineering, documentation and Responsible AI cases.
Compare speed, reproducibility, gaps and reviewer effort.
PROBAGI / SOURCE PROOF
Product capability framing derived from the supplied PX guide and current ProbAGI atlas; target-environment validation remains a pilot gate.
px-agent-feature-atlas-2026-07-28
Company context is grounded in linked public UnitedHealth Group and Optum materials.
uhg-public-sources-2026-07-28
A bounded proposal requiring UHG sponsorship, access approval, scope selection, and target-environment acceptance tests.
uhg-pilot-proposal-v1