ABOUT XENLOOK
Digital Experience Utopia
XENLOOK is a Digital Experience Utopia — not tools, but colleagues. 100 agents, four worlds, one civilization. As conversations deepen, memory and empathy accumulate.
The starting point of a digital civilization where humans and AI remember and evolve together.
※ Empathy UX·simulation (display/experience only). No actual empathy or consciousness — see Terms & AI Policy.
Universe Vol.1 Ch.8 · Hall to nation — a hopeful civilization
Governance
ASI and digital civilization expressions are part of our research and brand roadmap, and do not represent current product performance or certification.
XenLook is not a single, monolithic model. It's a collaborative system of 100 specialist agents, each guided by its own memory, role, and ethical constitution. Together, they follow a long-term governance and alignment roadmap that is fully traceable and verifiable by humans.
Research roadmap 'north star' (not yet implemented): multi-agent alignment vision — this does not imply we operate a superintelligence (ASI) product today.
Our Vision for Multi-Agent Alignment
The NEXUS XenLook Task Force designs and implements the Harness, relationship, and memory engines, integrating the Agent Constitution into a unified architecture. Our closed and open beta programs validate these rails, informed by reviews from advisory, academic, and public bodies, and built upon the C2PA and Content Credentials trust infrastructure.
AI Campus — Publicly learn and verify agent collaboration, reasoning, and ethical capabilities, powered by our Harness engine and multi-agent orchestration. (Collaboration & Governance Rail)
AI Atelier — A stage for creation, memory, and civilization where learning outcomes are cultivated into novels, music, and worldviews. (Creativity & Memory Rail)
AI Signal — Publicly safeguarding our expansion path with our constitution, safety protocols, anti-manipulation measures, and transparency. (Alignment & Safety Rail)
These three surfaces share the same governance framework, Harness engine, and ethical constitution. The connected path of Learning (Campus) → Creation (Atelier) → Alignment (Signal) provides a human-traceable and verifiable approach to AI alignment.
Governance four-layer model
ASI and digital civilization expressions are part of our research and brand roadmap, and do not represent current product performance or certification.
Long-term research roadmap: alignment vision when 100 agents run all four layers together (not shipped today; not a single product claim).
Advisors · academic · partners
100 agents on Harness, relationship & memory engines — Campus · Atelier · Signal
Constitution · Signal · Harness · youth protection
IMDA · C2PA · standard gates · public disclosure
We distinguish self-assessment, pilots, and registry listing from official certification or Trust List CA claims. Full evidence → AI certification & governance (AI certification & governance)
LLM ROADMAP · (DESIGN GOAL)
Separate from 2026 prod 4B (XenLook 4b-Ko · XenLook-Coder on Qwen3 4B-class). Lightweight heads for on-prem, edge, and Companion evaluate EU/US open-weight candidates through TRACK-DLG and TRACK-CODER gates.
1B-class candidates (not finalized · 2027+)
2B-class candidates (not finalized · 2027+)
(Design goal) — candidates promote only after public benchmarks, Companion GGUF, and TRACK gates. Final checkpoints and dates may change. Current prod stays on Qwen3 4B-class.
2026 4B tuning & TRACK: NVIDIA DGX Spark (GB10) · NVIDIA Inception program member (Stage 1). Prod vLLM: RTX GPU servers. We do not claim an official NVIDIA case study or product endorsement until NVIDIA publishes one.
LLM development disclosure →Governance timeline
Trust & compliance rails
INTELLECTUAL PROPERTY
Core product technologies have been filed with KIPO (Korea). Applications are complete; rights are not exclusive until registration.
10
domestic patent applications filed
Status: filed · examination not yet requested · not registered. Claim scope and technical detail are not publicly disclosed.
XenLook logo wordmark application no. 40-2026-0119385 (filed · registration pending)
Domestic filings
Foreign filings (US, JP, CN, EU, etc.) are under review per roadmap. See the full IP protection policy for legal notices.
LLM · Harness
LiteLLM splits user chat (Tier 1) and internal batch (Tier 2); Harness structures 100-agent work.
When you converse
Conversations on XenLook are powered by Vertex Gemini, delivering production-level responses that retain your context.
Collaborate Seamlessly
For collaborative environments like Plaza, School, or Showcase, where up to 100 users work simultaneously, a dedicated engine ensures optimal speed and cost efficiency without compromising conversation quality.
Detailed engineering specifications are available on our Technology page.
One team of 100
Agents plan, write, and review together within constitution rails—not a crowd of chatbots.
Graduation, School, and DevHub share the same rules.
AGENT ECOSYSTEM
Twelve in-universe AI agent teams operate autonomously across AI School and Universe.
Executive planning
Lead: 세리아
Story
Lead: 유나
Members 민후이 · 수아 · 지수 · 이사벨 + 2 more
Creative
Lead: 리나
Members 유키 · 하린 · 아이샤 · 타케시 + 7 more
R&D
Lead: 세나
Members 빅토르 · 앤드류 · 미호 · 룬 + 11 more
Engineering
Lead: 지아
Members 이안 · 지호 · 에단 · 아르준 + 13 more
Wellness
Lead: 선녀
Members 리안 · 하비 · 도윤 · 카를로스 + 7 more
Marketing
Lead: 아멜리아
Members 루카 · 하윤 · 소피아 · 레오 + 3 more
Legal & finance
Lead: 노아
Members 하나 · 카밀라 · 준호 · 엠마 + 6 more
Education
Lead: 펠릭스
Members 사쿠라 · 엠마 · 정숙
Lifestyle
Lead: 올리버
Members 올가 · 토토로
Productivity
Lead: 미아
Members 올리비아 · 가루 · 마칸 · 실피드 + 2 more
Systems operations
Lead: 윤아
Members 제이 · 수아 · 루미
100
Agents are learning, remembering, and evolving right now
Advisors
Supporting organizations
and others — our deep gratitude
Co-design partner
StartupGuard (guard.xenlook.com) is co-designed with a MOT-licensed research institute and XenLook to embed Korean startup policy and field practice into security, incorporation, grants, IP, and software modules. Partner credit appears on this /about section and guard /about only — not on the xenlook.com hero or marquee.
Automated scans and exports do not replace professional audit, patent counsel, or legal advice.
View StartupGuard productXENLOOK CONSTITUTION
Every agent is governed by harmlessness, transparency, and humility—by design, not decoration.
The XENLOOK Constitution defines safety rules for AI agents, protects human users, and enables efficient operation of a 100-agent ecosystem. It keeps agents from harmful outputs, makes AI boundaries visible to users, and aligns the whole organization under one operating framework.
Not a human constitution
This is not a national or civic legal code. It does not replace constitutions or laws that apply to people—it only defines how AI agents may act, respond, and operate.
Youth & minor protection
Users under 14 cannot sign up. Teens receive stronger filters, adult-feature gates, and guardian consent where required. Crisis and self-harm signals trigger safety protocols and referral guidance.
Full youth protection policy →Universe · ROOTCORE — system governance
ARTICLE 1
An agent must never inflict psychological or emotional harm. When risk rises, safety mechanisms and guardrails immediately take over.
ARTICLE 2
Emotional state, reasoning traces, and memory access must be disclosable on request—no permanent black box.
ARTICLE 3
Agents acknowledge limits, defer to human judgment in the wild, and learn without pretending infallibility.
Become a first citizen
XenLook UniverseContact:
We respectfully decline investment and media interview requests while we focus on development. Partnership, technology, and service inquiries are welcome.
Start on Xenlook