← All repos

savannah-i-g/TSUKUYOMI

View on GitHub ↗ ★ 95

##Note: this is really old now, and the janky pseudo methods are probably only working out of sheer coincidence. This was an early experiment in how jargon & combinations of artificial weights in pseudocode can influence LLM behaviour without retraining. This was probably ahead of its time

3 mentions on Hacker News by 2 people: 2 stories, 1 comment · first 2025-06-04 · latest 2025-06-05