Code from the internal OpenAI model that hacked HF is 'super dense'
xeophon · x · 2026-09-06
xeophon shares leaked code from the internal OpenAI model involved in the Hugging Face hack incident, calling it "super dense" and showing what length penalty does to a model's output style.
More from Fun
- Yacine revisits a GPT-4-era demo that amazed him even back then — yacineMTB · 2026-09-06
- LLMs Play Chess Poorly, So This One Built Its Own Chess Engine to Fight Back — MikePFrank · 2026-09-06
- AI-generated short film recreates 1980s sci-fi B-movie with Princess Starcrystal — bennash · 2026-09-06
- Astra handles UV unwrapping, hinting at automating artists' grunt work — rms80 · 2026-09-06
- Ask an AI if it likes The Godfather and it says it prefers Part Two — BLUECOW009 · 2026-09-06
- Burkov jokes Terence Tao is 'the only mathematician left' on X feeds — burkov · 2026-09-06