Hi, I'm Shiho Yoshino 👋

Founder of An-soku LABO
Building reliable infrastructure for AI systems.

I work on independent research, OSS utilities, AI-agent governance, reproducibility, and product/runtime observations.

Focus

• AI infrastructure & reliability
• Agent authority / means / human-approval boundaries
• Deterministic and verifiable utility tools
• Reproducible bug reports and product observations
• AI workflow / artifact provenance
• Load Minimization Theory (LMT)

Selected public projects

File Integrity Toolkit

Deterministic file inspection and integrity verification utilities.

https://github.com/kouteichan/file-integrity-toolkit-oss

Deterministic Calculation Toolkit

A deterministic calculation toolkit designed for reliable AI/tool workflows.

https://github.com/kouteichan/Deterministic-Calculation-Toolkit-OSS

LMT Code Generation Protocol

Public work related to LMT-based code generation and engineering protocols.

https://github.com/kouteichan/LMT-Code-Generation-Protocol

OSS / Product Contributions

I also contribute reproducible bug reports, evidence, and workflow/design feedback to AI infrastructure projects.

Recent work includes observations and proposals around:

• cross-device conversation-state consistency
• file/source accessibility across AI runtime boundaries
• tool-call cancellation and terminal-state behavior
• ChatGPT Project ↔ Codex artifact/task handoff
• provenance, human approval, and repository mutation boundaries

My approach is:

Observe → reproduce → separate confirmed facts from unknowns → preserve evidence → contribute

An-soku LABO

Reliable Infrastructure for AI Systems

Research and engineering focused on making AI systems easier to verify, govern, reproduce, and use reliably.