Knowledge / AI

AI is a research instrument for inspectable reasoning: Labwise connects report feedback to source evidence, clinical tutors expose hypotheses, and the knowledge OS retains those traces.

On this site

Clinical Reasoning AI

Prototype: multi-agent tutor loop that scores intermediate hypotheses.

Labwise · 实验报告辅助评阅

Local teacher pilot: evidence-linked feedback, rubric review and instructor-confirmed exports.

Projects

Labwise · 实验报告辅助评阅Prototype

Evidence-led AI assistance for university lab reports: understand the source, review rubric-linked suggestions, and keep final scores and feedback in the instructor's hands. 帮助教师理解英文实验报告、核对原文证据并形成可审核的反馈。

Clinical Reasoning AIPrototype

Investigating how agent architectures can approximate diagnostic reasoning for medical learning.

Personal Knowledge OSActive

A personal operating system for accumulating, relating, and retrieving intellectual work over time.

Research BriefActive

Personal observatory pipeline — Notion Thinking Vault as input, Obsidian as memory/graph, public MUJI homepage as the projection on this site.

Intellectual focus

Labwise · Evidence-led Review — A teacher workspace for English lab reports: source evidence, rubric-aligned AI suggestions and instructor-confirmed feedback. 面向高校教师的实验报告辅助评阅。

Clinical Reasoning AI — Investigating how agent architectures can approximate diagnostic reasoning for medical learning.

Personal Knowledge OS — A personal operating system for accumulating, relating, and retrieving intellectual work over time.

Active questions

How can AI simulate clinical reasoning?How can biological mechanisms become computational models?

Related signals

Agentic tutors begin evaluating multi-step diagnostic traces — New prototypes score intermediate clinical hypotheses, not only final answers.

Personal knowledge graphs re-emerge as research infrastructure — Tooling trends favor local-first graphs with exportable relation schemas.