Definitions, reference architectures and implementation reports for agentic game systems. Every public entry preserves its sources, review date, evidence boundary and machine-readable alternate.
The pilot is intentionally small. New pages enter this archive only after their own claim, source, media and release gates pass.
01A rigorous guide to offline calibration, online measurement and why a live MMORPG agent should not rewrite its own policy or permissions.
02A reference architecture for letting LLM agents interpret player intent without giving generated output direct control of a shared MMORPG world.
03An evidence-based definition of agentic NPCs: what makes them different from dialogue chatbots, scripted bots and unconstrained autonomous agents.
Lineage2AI is first-party project documentation. We make that relationship explicit and keep observations inside the boundary of what the named source, code version or test actually establishes.
Architecture statements are checked against current schemas, implementation boundaries and versioned project documentation.
Offline, provider-lab, boot, in-game and soak evidence remain separate so a passing test cannot become a larger marketing claim.
Model output is described as a proposal. Deterministic game-server validation owns identity, legality and every shared-world side effect.
Routes, internal links, sitemap, Markdown alternates and LLM discovery are generated from the same reviewed public registry.
Human pages, Markdown alternates, sitemap entries and LLM discovery are compiled from one eligible registry. Planned topics and future articles never appear as published facts.
