Engineering · 11 min
Cat AI is going to be slop. Here’s how we tried not to be.
A 4-tier voice quality gate, a 15-mood daily lottery, pop-culture voice modes, and 17 rounds of audit. The architecture behind a cat AI that tries not to sound generic.
Applied research · 7 min
From Natsume to Bridgerton: how an HCI paper changed our cat AI
An HCI paper validated defamiliarization as the lever for non-mundane LLM cat voice. Their reference set was 1906 Japanese literary fiction. We translated it to pop culture our users actually recognise.
Build log · 9 min
I shipped a cat AI app in 14 days with Claude as pair programmer. Here’s the playbook.
Solo dev, Singapore-based, zero outside funding. From create-expo-app to Google Play production in fourteen days. The workflow, the audit loop, what Claude was good at, what it wasn’t.