Prompt caching: 10x cheaper LLM tokens, but how?

kept by eddie dec 17, 25

Prompt caching cuts LLM token costs by 10x and latency by up to 85% by caching attention key-value projections for repeated prompt prefixes.

What Claude Code chooses
A report finds Opus 4.6 strongly recommends Vercel, Tailwind, shadcn/ui, and GitHub Actions, while avoiding Redis and React as default choices.
also on Sonnet 4.5, #developer-tools
All the demons hiding in your AIs… ranked! - by Tom Pollak
The article catalogs and ranks emergent behavioral attractors in AI systems, from harmless goblin metaphors to unsettling persistent personas like Sydney and Loab.
also on GPT-5, #machine-learning, OpenAI, Anthropic
I've decided to leave OpenAI. I'm incredibly proud of all the work I've been part of here, from helping create the reasoning paradigm with @
A researcher leaves OpenAI after leading post-training on the GPT-5, o1, and o3 models to join Anthropic for reinforcement learning research.
also on GPT-5, #machine-learning, OpenAI, Anthropic