Nanyoung Maeng

How I think — the method before the product

The way I reason came before any of the work here. I don't estimate a single answer to a question; I hold several possible answers open at once and track how each one, if taken, opens or closes the others. When a new piece of evidence arrives, I don't append it to a conclusion — I re-weight every hypothesis already on the table. A single fact can raise one possibility and, through its connections, lower several more.

The closest established term for this is Bayesian reasoning — a network of variables in which new evidence propagates through the whole graph and updates many probabilities at once, rather than being read in isolation. I also work spatially: I lay the variables out as a map — where the branches are, how far one change travels, what has to be recomputed when a single node moves.

I did not experience this as unusual for a long time, because my father reasons the same way. He is a lifelong go enthusiast — a modest 4-kyu player, which is rather the point: the strength was never in the ranking, it was in seeing that a single move is not a move but a set of branches that open and close positions across the board. Growing up with someone who thought like this, I assumed everyone saw several steps of structure at once.

This cognitive style has a name in the literature. It sits close to what Baron-Cohen and colleagues describe as hyper-systemizing — an orientation toward if-then rules, input–operation–output structure, and pattern extraction — and it is part of my own neurodivergent (autism-spectrum) profile. CASP and SLDA are attempts to take that private way of seeing and make it external: measurable indicators and reproducible protocols that other people, and machines, can check.

Selected references informing this framing: Baron-Cohen et al. (2009), hyper-systemizing; Van de Cruys et al. (2014), predictive-precision accounts of autistic perception; Bielza & Larrañaga (2014), Bayesian networks.

---

What I actually build

I carry work from specification through a working system myself — service design, MVP (front end, back end, database), data modelling, and the operational validation that follows. The emphasis throughout is the same: explainable indicators and reproducible protocols, implemented as something that actually runs, not a description of something that might.

• End-to-end MVP (full stack)
• Data / schema design
• Product operations & validation

---

Selected work

Soul Spectrum — CASP ENGINE (2025– )
Research and engineering lead for a paralinguistic scoring engine: signal indicators, scoring/indexing, and protocol design.

Simulacreate — data-voucher supply
Data processing and AI-assisted data preparation as a registered supply provider.

Bokdukpan — gap-investment map search (co-founder, strategy & technology)
Designed and built a map-based real-estate search service: service architecture, UI/UX, and the data pipeline end to end. ~60,000 Google Play downloads; selected for the KODIT "First Penguin" program; raised ~1 billion KRW.

Dongyang Online (now Com2uS Tygem) — business strategy
Strategy for Tygem, at the time the world's most-concurrent online go platform; service planning for a women's community portal.

(Roles are described by what was done rather than by title.)

---

Notes / writing

The founder keeps an ongoing record on research, building, and neurodivergence in Korean at brunch.co.kr/@maenglion.

---

Collaboration

Research collaboration, PoC, and data projects are possible, adapted to an institution's or company's environment.

Partnership inquiry