Latest / Models & Agents — Video Edition
Ep 147: OpenAI is testing private safety processing that keeps frontier-model interactions off…
# Models & Agents OpenAI is testing private safety processing that keeps frontier-model interactions off-limits to staff while still catching risks across long agent sessions. What You Need to Know: OpenAI previewed Private Safety Processing for frontier models to improve safety without personnel seeing raw content. Simon Willison documented an untrusted-sandbox experiment where Claude Code triggered an autonomous GitHub Actions push. New benchmarks and tokenization work on smaller models plus practical inference tuning on llama.cpp round out the day’s concrete releases. Top Story OpenAI…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.