The user's 18-core ARM64 server with 64GB RAM costs $0/hr.
GPU rental costs $15/hr. The universal encoder works on both.
Key point: the encoder is not optimized for speed. It's optimized
for universal accessibility — any CPU, any framebuffer, any host.
On the ARM64 server:
n=20: 200ms (18 cores) — quick
n=24: 3s — warm-up
n=28: 50s — coffee break
n=30: 8min — still fits in 16GB of 64GB
No GPU required. No $15/hr rental. The math runs on the hardware
you already have.
"By Thor's balls, no, it's not efficient. But it works on a potato."
Refs: FBTTY_UNIVERSAL_ENCODER.md, python/universal_encoder.html
