Writing
9 August 2026
I spent a phase measuring a number that was already public. The process turned out to be in excellent statistical control — at more than double the level a graduate labour market should run at. And when I re-checked my own findings before publishing this, I caught myself breaking a rule I had written ten days earlier.
25 July 2026
Four causes survived the screen. The obvious next step was to find a country that fixed this and copy it. That is the fastest way to publish a confident, wrong answer. This is the design that stopped me, the country that broke it, and the error I caught in my own draft.
17 July 2026
I scored all 18 fishbone hypotheses for graduate underemployment against published evidence, using a rubric locked before scoring. The suspects that dominate public debate scored lowest. The four that screened in are the ones nobody campaigns about.
12 July 2026
A fishbone diagram puts every popular explanation for the 32.2% graduate underemployment defect in one frame — and holds each one to the same standard: unproven until the data says otherwise.
1 July 2026
Malaysia's graduate crisis isn't the 3.2% unemployment headline — it's the 32.2% who are employed but under-levelled. A forensic Cost of Poor Quality analysis puts the price at RM2.24 billion in lost tax revenue alone, and up to RM62 billion in drained national capacity.
27 June 2026
"Graduates can't get jobs." Everyone in Malaysia knows this. So I ran Define on it — and the same population is a 3.2% problem or a 32% problem, depending on one honest change to the definition. Including the part where the data ran out at 11pm.
19 June 2026
Sabah has 1 doctor per 1,283 people — 68% below the national target. We ran a live DMAIC investigation on it using AI. Three phases in, the data is clear: the root cause is structural, not political. Here's what we found.
16 June 2026
Two weeks ago, I didn't know what a terminal was. By the end of the evening, I had a live website — no code, no developer, no waiting. This is what actually made it work.
15 June 2026
There is a difference between acknowledging AI and reckoning with it. And that difference, I think, is everything.