Stanford dumps 1,000+ system prompts into the open
Researchers at Stanford have published a public collection of more than 1,000 system prompts scraped from applications built on ChatGPT, Claude and other large language models.
The dataset accompanies AISPA, a paper proposing a method for auditing those prompts from the user’s side. Erik Brynjolfsson and Alex Pentland are among the authors.
The preprint is on arXiv as 2607.28617, 2026.
Up next
Spend-limit errors now name the limit and the reset time
Claude Code 2.1.225 adds gateway spend limits: hit one and the message now names the limit, the reset time, and your operator’s custom text. It needs the gateway on 2.1.225. Also fixed: a 401 that swapped CLAUDE_CODE_OAUTH_TOKEN for a short-lived token and broke headless sessions.
Related stories
- Coding-agent prompt files grow 226% and never shrink
- Claude memory: say what it’s for, let dreaming fix notes
- Tübingen researchers want AI to propose many models
- Claude Fable 5 tripled Opus 4.6’s EVE Online trading gains
- Scientists use Claude least where they say it matters most
- METR clocks Anthropic benchmark gains at 1.9x old pace
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.