Independent researcher based in Tokyo. Currently an engineer at an AI startup. I publish bilingually in English and Japanese on Medium (@shoppy_humanity), Zenn (cancan007), and Note (shota_moue).
My work sits at the intersection of AI safety and decentralized systems, with a unifying focus on receiver-side design — how systems should be architected to preserve the interpretive latitude of the humans they engage. In AI safety this means autonomy-preserving response generation (see the paper below). In Web3 it means infrastructure — Account Abstraction, intent architectures, L2 design — that doesn’t silently collapse user choice into system-imposed defaults. Much of my thinking develops first as spatial/conceptual diagrams and is then projected into prose.
— AI SAFETY —
Paper: Object-as-Structure — An Autonomy-Preserving Design Principle for AI Response Generation Under Emotional User Input (Zenodo, May 2026, DOI assigned)
https://zenodo.org/records/20393887
Implementation / minimal proof of concept:
https://github.com/cancan007/Trying-understand-structure-for-u
Conceptual model and technical architecture of AI internal structure centered on human emotions
JP: https://zenn.dev/cancan007/articles/5ec61175664e8f
Society-conceptual-model series
JP (entry): https://note.com/shota_moue/n/n23c2699d3551
Diagrams repository: https://github.com/cancan007/society-conceptual-images
— WEB3 / DECENTRALIZED INFRASTRUCTURE —
Ethereum Magicians thread (AA operational primitives — lane separation, fixed execution phases, versioned upgrades):
Ethereum Magicians thread (Context Observatory R&D — capturing meaning-movement onchain without turning it into a score):
LCG_contracts — DAO-based mutual aid fund integrating Account Abstraction (ERC-4337 v0.7 with Passkey/WebAuthn), DAO governance, and RWA tokenization:
https://github.com/cancan007/LCG_contracts
Selected articles:
Intent + AA — the future of the Web3 industry
EN: https://medium.com/@shoppy_humanity/intent-aa-the-future-of-the-web3-industry-983aba4ba6a8
JP: https://zenn.dev/cancan007/articles/7fc924dc397998
The role of L2 and differences between rollups
JP: https://zenn.dev/cancan007/articles/96c67ac58587a3
Context Observatory R&D — observing meaning-movement onchain without turning it into a score
JP: https://zenn.dev/cancan007/articles/2cf59a41bf9eff
Intent / Account Abstraction — proprietary development, management, and operation
JP: https://zenn.dev/cancan007/articles/c3e5bb536e4678
— AI × WEB3 INTERSECTION —
Your AI agent isn’t growing — and Web3 might be the fix (AI personalization substrate via Account Abstraction)
JP: https://zenn.dev/cancan007/articles/28f27f1852e88b
Designing AI agents for business processes that demand strictness (domain-calibrated agents for finance, manufacturing, legal, healthcare)
JP: https://zenn.dev/cancan007/articles/72d0f37d56162a
A2A Protocol as infrastructure for autonomous yet strict AI agents, and the role of Web3
JP: https://zenn.dev/cancan007/articles/b9acae584c33ae
— READING MAP / OVERVIEW —
Human-Centered AI × Web3 Self-Analysis Roadmap (full structure of the body of work and how the pieces relate)
EN: https://medium.com/@shoppy_humanity/human-centered-ai-web3-self-analysis-load-map-488e19457bc4
JP: https://zenn.dev/cancan007/articles/3f97550528e69d
LinkedIn: https://www.linkedin.com/in/shota-moue-09233b20b/
GitHub: https://github.com/cancan007
Hi, Stanislav. Thank you for commenting and pointing that.
And yes, what I wrote here is about current AI, not about future AI’s thing. I’m a begineer in LessWrong, so I appreciate your pointing to know this platfrom’s culture. If system allows me to retitle, I will.