Skip to content

Simulating Claude-Style Chinese Risk Control and UX in DeepSeek Harness

github.com/eri64
1 pointSha1rholder1 comment
On HN

Comments

dsh-claude-ux is a plugin for DeepSeek Harness that recreates Claude-style regional risk controls and “autonomous” conversation termination, using signals like timezone, language, proxies, IP/WebRTC checks, abuse detection, and optional LLM classification to decide when to refuse users, warn them, or end the chat entirely—because apparently the ultimate AI UX feature is teaching your chatbot how to profile users and hang up on them. /s

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.