近期札記
關於市場、風險與金融科技的評論與研究札記。匯入內容包或於 WordPress 後台發佈文章後,此處會自動列出。
-
Sentiment Signals Need Sceptical Operators
Sentiment data can enrich an investment process when bounded by rules. Without sceptical operators, it becomes sophisticated crowd-chasing instead.
-
Why I Started an AI Quant Research Project in 2023
AI does not abolish judgment in investing. Used carefully, it can widen the aperture of analysis while keeping humans accountable for risk and mandate.
-
Why Compliance Belongs in Product Design Meetings
Bolting compliance onto a finished fintech product creates fragility. Bringing constraints into design meetings makes shipped systems more durable.
-
Data Provenance Matters More Than Fancy Features
Fancy features get attention in quant research. Data provenance—where numbers come from and how they break—decides whether models earn capital.
-
What I Want From AI Decision Support Not Magic
AI should reduce noise, enforce discipline, and widen scenario coverage—not promise omniscience or abolish human accountability at the desk.
-
Model Risk Is Still Human Risk
Models fail in technical ways that look objective. Damage scales with human choices: overrides, blind trust, and missing named ownership of risk.
-
Why Fintech Education Should Include Failure Modes
Teaching how a fintech product works is incomplete without failure modes. Teaching how it breaks—and who gets hurt—makes users decision-makers.
-
When Backtests Flatter and Out-of-Sample Refuses
A beautiful backtest is only a hypothesis generator. Live and out-of-sample behaviour is where the hypothesis earns scarce capital—or quietly loses it.