How the work is made
Research Approach
A craft-led method for using AI on Taiwan equity market data, where the source is named first, the model role is named second, and a person reads the result before it is published.
The research approach is the custom service-model page for this site. It describes how an article is actually made, so a reader can weigh the work rather than take a byline on faith.
1. Frame one question
Every piece starts with a single, bounded question about Taiwan-listed equities — for example, how order-book turnover on a TWSE-listed name changed across a quarter, or how a sector rebalance inside the TAIEX families shifted constituent weights. Broad questions are narrowed until a public dataset can answer them.
2. Pull the source data
Data is taken from public Taiwan exchange and disclosure sources: TWSE market prints, MOPS quarterly filings, and the published TAIEX index constituent files. Where a figure is derived from a raw series, the derivation is written out, not hidden behind a chart.
3. Name the model role
AI is used for one of four jobs, and the job is named in the article: sorting a large table into a readable order, summarising a long filing into a short note, scoring a panel of names against a stated rule, or drafting a first pass of prose that a person then edits. A piece never claims AI did something it did not.
4. Keep a human reading
A person sits between the model output and the published sentence. The human checks the source row, rewrites any claim the data cannot support, and removes anything that reads as a recommendation on a specific security. The model does not have the last word.
5. Publish the limits
Each article closes with a short limits note: what the dataset cannot tell you, what the time window leaves out, and where the analysis deliberately stops short of advice. The limits are part of the work, not a disclaimer bolted on at the end.
AI is a tool in the workshop, not the author of the conclusion. The model role is named in every piece.
What this approach is not
It is not a quantitative strategy, not a forecasting system, and not a tool for picking stocks. It is an editorial method for reading Taiwan equity market data carefully, with AI used where it helps and named where it is used. If you want to discuss the method in detail, the contact page is the right entry point.