Server-authoritative AI
Designing the boundary where models interpret and propose while deterministic game-server code owns identity, legality and every shared-world side effect.
Founder and Lead Agentic AI Systems Architect
Christian Zapasnik is the founder and lead systems architect behind Lineage2AI, building server-authoritative agentic NPC, voice, memory, calibration and evaluation systems for persistent MMORPG worlds.
The work connects language models to real game systems without allowing generated output to become multiplayer truth.
Designing the boundary where models interpret and propose while deterministic game-server code owns identity, legality and every shared-world side effect.
Connecting bounded perception, planning, tool selection, persistent memory, authoritative results and measurable failure behavior.
Building voice turns that preserve player scope, current revisions and result-before-speech delivery inside a multiplayer world.
Keeping offline optimization, provider tests, boot evidence, in-game acceptance and soak evidence distinct instead of allowing live self-modification.
Christian is the project founder and a first-party source for Lineage2AI design and implementation. That relationship is disclosed because first-party experience is useful evidence, but it is not independent coverage.
AI may assist research organization, drafting and editorial visuals. Public technical claims still pass named-source, repository, editorial, technical, evidence, media and release gates. SOURCE, OFFLINE, BOOT, IN_GAME and SOAK are never treated as interchangeable.
Read the evidence-led research archiveA rigorous guide to offline calibration, online measurement and why a live MMORPG agent should not rewrite its own policy or permissions.
Read referenceA reference architecture for letting LLM agents interpret player intent without giving generated output direct control of a shared MMORPG world.
Read referenceAn evidence-based definition of agentic NPCs: what makes them different from dialogue chatbots, scripted bots and unconstrained autonomous agents.
Read reference