Skip to content
A simple study with a central desk, a thinking board, a research shelf and an experiment bench.

William / Explore

Classic Site ↗
EN

3D is off. All content is ready to read.

01Workspace

Build something useful. Keep its boundaries clear.

I use AI to help turn ideas into software. The work spans on-device media, security analysis and local speech automation.

The projects are still evolving. I keep what is working, what is unfinished and what needs a person’s decision visible in each write-up.

View the current work
02Thinking board

Start with what the system must not do.

Before building, I map the constraints, the information that can leave the device and the decisions that need human confirmation.

An architecture is a set of choices with consequences. My workflow puts those choices next to tests and explicit acceptance criteria.

Read how I work
03Research shelf

Keep a question open long enough to test it.

I investigate local workflows, evaluation methods and the limits of automated checks. A useful research note includes what changed my mind.

These are the research records currently published on the site. An experiment can be paused; that status is part of its result.

Read the research records
04Experiment bench

Taiwan Stock Signal Lab

A deterministic post-market screening workbench built on official TPEx open data. Evaluates mechanical rules locally with zero runtime external API calls.

Research demo with local data archive accumulating. Complete screening rules, verifiable data provenance, and limitations are documented in the full case study.

View full case study