ZhiZhi Gewu

Can a Foundation Model Match XGBoost on Tabular Data?

Deep learning no longer loses tabular by default.

23 hrs ago • KY John

June 2026

The Last Mile From Analysis (Answers) to a Data Mart (Assets)

An analyst’s notebook is not a pipeline — and the gap is wider than it looks

Jun 21 • KY John

I Let a Neural Network Do My Feature Engineering — Then Made It Explain Itself

An experiment in automated feature discovery for credit risk: a sequential model trained on raw payment history, interrogated with explainability tools…

Jun 13 • KY John

2

May 2026

I knew the rules of Go. Writing them for a machine taught me I didn't understand them.

I’m rebuilding AlphaGo from scratch — 9×9 board, no shortcuts — to actually understand how it works instead of just running someone else’s repo.

May 30 • KY John

Why Singapore has no policy rate, the yen collapsed, and Brazil paid you 14% to hold a rising currency?

How Every Country Actually Runs Its Money

May 24 • KY John

2

What's Really Going On in Machine Learning? A Simplest Example by Stephen Wolfram

From Neural Networks to Rule Arrays: A Minimal Model of Learning

May 16 • KY John

Interchange Fee and the Universal Acceptance of Cards

Payment is fundamentally a two-sided market. Cardholders want more merchants accepting their cards. Merchants want more cardholders. Increasing the…

May 9 • KY John

April 2026

What 50 years of Singapore property prices tell us about risk

Why "average return" lies, why one quarter in 1993 still distorts a 35-year statistic, and what this means for anyone planning to buy property

Apr 25 • KY John

The Hidden Interest Rate in Your Insurance Bill

A mental shortcut for comparing payment plans.

Apr 19 • KY John

When to Use a Blacklist, and When to Use a Rule

After joining the anti-fraud team, I noticed that blacklists are used far more extensively than they were in credit underwriting.

Apr 4 • KY John

1

March 2026

If You Cannot Create It, You Don't Understand It — Even with AI"

I recently built a tiny autograd engine from scratch — just a hundred lines of Python that can compute gradients through a computation graph.

Mar 29 • KY John

Can AI Agents Win a Modeling Challenge? A Replicable Experiment

To get the most out of AI agents, we need to remove human bottlenecks and increase leverage.

Mar 25 • KY John