Indonesia · Credit Risk · OJK Data · 2026

Two portraits of one dataset.

Two real credit research reports — SME and consumer non-performing loans, built from the regulator's own numbers — rendered as two completely different websites. Same figures, opposite visual languages: one is a city at night, the other a printed report. Designed and coded end-to-end, autonomously, by Claude Fable 5.

4.86%SME NPL · Apr-26
6.44%Jawa Barat, worst SME big book
3.28%mortgage NPL
CITY
consumer credit · night city
Rp 2,084 T of household credit as a WebGL city. Every window is a loan; the red ones stopped paying. Scroll flies you district to district.
Enter the city →
GROUND
SME credit · Swiss print
The annual-report cut: paper, ink, one signal red, and a drag-to-rotate spike map of all 34 provinces on real geography.
Read the ground truth →

Same numbers, every time.

The dataset behind all six: OJK's Statistik Perbankan Indonesia plus a direct pull of the regulator's public portal cube (April 2026), reconstructed and validated against published aggregates to the rounding digit, in two research reports written on the PKA desk.

Both sites hold the same discipline: no stock images, no chart libraries, no templates, no build step — every mark generated from the data arrays in the page source. Each carries its own /guide explaining exactly how it was made.