AI systems

Use AI in delivery without handing authority to an unreviewed agent.

The AI lane focuses on source-governed workflows, UAI-style internal memory, project handoff, file handoff, semantic retrieval, local or managed model choices, and human review gates.

Internal memory

Keep current constraints, accepted decisions, owners, next actions, and operating checks in compact AI-readable memory instead of scattering them across chats.

Project handoff

Create startup packets, receiver briefs, system profiles, context, constraints, operations, decisions, and test plans that let humans and agents restart work cleanly.

File handoff

Classify uploaded or generated artifacts into content, improvement, and archive paths so agents know what to use, what to improve, and what not to treat as current truth.

Semantic retrieval

Use search and retrieval where they help, but keep provenance, contradiction handling, and source hierarchy visible.

Review gates

Treat AI output as proposed work until source checks, human review, tests, and deployment boundaries say otherwise.

Legacy-code analysis

Use AI to summarize, map, and test legacy behavior without giving it uncontrolled write access to production logic.

UAIX memory support

Add UAI internal memory, project handoff, and file handoff to the AI delivery lane.

The microsite now has a dedicated clean route for AI Memory Handoff. It explains how Long Term Software can use UAIX-style local memory packages to keep agent work restartable, reviewable, and source-bound.