Generative-AI solutions consultant by day · multi-agent systems researcher by conviction · Taiwan
生成式 AI 解決方案顧問／多代理系統研究者——把自己運行的系統寫成論文，再拿論文回頭改系統。
zaious.dev · ChronicleCore Studio · LinkedIn · ORCID · Google Scholar · ☕ Buy me a coffee
Information flows endlessly. Chronicle anchors what matters.
Three private systems. None of them is on GitHub; what they produce is — each line ends with the public evidence you can actually open.
ChronicleCore — the whole thing: a governance-first multi-agent system I run daily, 39 specialist agents in five departments (Core / Shield / Hands / Soul / Senses, per the public roster v3.0), with MCP as the hard boundary between LLM reasoning and tool execution. Everything below runs inside it. → architecture whitepaper · ASAF, the framework the operating experience became
Chronicle-Ark — the multi-agent IDE the agents actually live in, including the paper room where research gets written under explicit AI disclosure: core ideas from the human, execution delegated, boundary stated on the page. → every paper and preprint under Research · riftbound-chronicle was built there
sovereign-dossier → biography — a single source of truth for my own facts (roles, projects, papers, with provenance and visibility rules), exported as a brief for agents and as a résumé pipeline, both over MCP. → every number and status on this page is checked against it before it goes up
| What it is | ||
|---|---|---|
| Antigravity: Skills Chronicle | VS Code extension — a visual dashboard for managing AI-agent skills. |   |
| translate-academic-paper | Turns a paper PDF into a self-contained bilingual HTML (original / 中文 / side-by-side), with OCR fallback and a verified glossary. |  |
| journal-atlas | Submission advisor that reasons over real journal metadata (399 curated entries) instead of model recall. Also an agent skill. |  |
| jev-capability-atlas | Bilingual capability map of TypeSafe's Jev: where a calibrated, narrow-decision model holds up and where it breaks, built on real API receipts. Independent, not sponsored by TypeSafe. |  |
| basesmall | Desktop app that keeps your team's MLB game in a corner of the screen: runners, the count, every pitch, the score. Unofficial. Download |  |
| SmalLoL | Basesmall's League of Legends sibling: a thin desktop bar with a pro match's score, gold lead and next objectives, plus the draft read live from the broadcast, since the official feed leaves it out. Unofficial. Download |  |
| OwnVoice Meet | For meetings in a second language: live captions with translation, and a reply box that turns what you type in Chinese into short spoken English you read aloud in your own voice. The first version was built the afternoon of my JCDL 2026 online demo and used through it. Windows. |  |
| JevTRPG | Paste a résumé, get a 1920s d100 investigator sheet; every skill is one scored question to Jev, and nothing you paste is stored. Unofficial. Live |  |
| PhilosopherAtlas | Interactive map of 235 philosophers across history and cultures, bilingual. Live |  |
| riftbound-chronicle | Preparation-phase AI for the Riftbound TCG — deck coach, cited rules consult, and a human-confirmed Player 2 agent — on a sovereign rules core that fails closed on anything it can't model. Not a simulator, not a judge. Live |  |
| Taiwan.md | Open-source curated knowledge base about Taiwan for LLMs — collaborator (31 commits upstream). taiwan.md | |
| PTT-Alertor-Discord | Discord bot that watches PTT boards for keywords and posts rich embeds. |  |
Built and run under my studio, ChronicleCore Studio (編年史記工作室).
- Riftbound 編年史 — Taiwan's player intelligence site for the Riftbound TCG. 3.7–4.6k page views a day (2026-09), 640–970 daily clicks from search (2026-09), most content drafted by an agent skill and curated by hand.
- Riftbound Agent — the rules-engine side of the same project as a free tool: card database with full rules text, deck builder with import and legality checking; rules assistant and deck coaching next. Built on riftbound-chronicle above, not on a chatbot.
- SummonLibrary 喚書者 — book search across Taiwanese second-hand inventories, with an MCP server.
- AChanGer 朱藏 — anime merch intel and card-exchange tool.
- Journal Atlas — the hosted version of the submission advisor above.
- KoreaCare 韓妝原價雷達 — snap an influencer ad or a package, find the original Korean product, and see how the advertised price compares with live Korean retail prices.
- Basesmall — follow your MLB team from a tiny always-on-top window; free and open source, Windows first.
- SmalLoL — the same idea for League of Legends esports: a quiet desktop bar for the pro match you can't watch, built and tested during the 2026 World Championship.
- JevTRPG — a résumé read into a tabletop investigator sheet, in English, 繁體中文 and 日本語.
One question, asked three ways: how designed interaction constitutes the human — a philosophy-of-interaction axis, a sociology-of-interaction axis (agents as social actors), and a psychology axis. Practice first: I design, deploy, and live inside the systems, then write the theory and check it against them.
- Agentic Social Affordance Framework (ASAF) — Frontiers in Computer Science (Human-Media Interaction), 2026. Agent identity design as the collaboration interface in multi-agent systems. doi:10.3389/fcomp.2026.1860996
- Journal Atlas: Venue Fit Scoring Under Incomplete Evidence, and the Corpus Method It Requires — Proceedings of JCDL 2026 (ACM/IEEE Joint Conference on Digital Libraries), Posters & Demonstrations. The paper behind the journal-atlas repo above. doi:10.1145/3805696.3846515
- Toward a Philosophy of Interaction — preprint, Zenodo. doi:10.5281/zenodo.21225988
- Idiosyncratic Cognitive Amplification — preprint, Zenodo. doi:10.5281/zenodo.20436631
- Before "How to Use It": Locating Generative AI in the Teaching-Learning Structure — preprint, SSRN, co-authored with a philosopher of education. doi:10.2139/ssrn.7278438
- 2026 conference papers: one on sensor-embedded props for elderly day care, presented orally at IEEE BMEiCON 2026 in Fukuoka (second author). Accepted: two at the Taiwan Association for Information Society annual conference (companion-AI holding effects; accountability in human-in-the-loop multi-agent systems), and one at the Taiwan Philosophical Association's "Technology, Craft, and Public Philosophy" (AI Philosophy in Taiwan panel).
- Earlier: VR metaverse podcasts, IEEE ICCE-TW 2022; M.S. Interaction Design, NTUT.
How the writing gets done. Under explicit disclosure, with a multi-agent AI collaboration setup I built myself. The rule I hold to: core ideas must originate from the human author; execution may be delegated. It has passed peer review with that disclosure on the page.
- 2016 — founded and moderated PTT's VR board the year the first Vive shipped; first individual member of Taiwan's VR/AR industry association (TAVAR).
- 2022 — founded PTT's AI_Art board the year ChatGPT launched, writing and enforcing the rules for generated content before any platform had them.
- 2023– — generative-AI solutions consultant at dentsu Taiwan: 5+ enterprise PoCs, an AI-literacy programme for 1,000+ staff, government-approved lecturer under Taiwan's SME digital-transformation schemes.
- 2026– — CTO at Lumo, an elder-care AIoT startup; the long-term-care paper above comes from that line.