I Cancelled Claude: Token Issues, Declining Quality, and Poor Support | Hacker News

A Hacker News thread says some users are canceling Claude because of token limits, lower output quality, and weak support.

The discussion compares proprietary models with open-weight options like Qwen, Kimi, and DeepSeek for coding tasks.

Commenters debate whether local models are practical enough to replace paid AI assistants.

The broader issue is whether AI coding tools can justify their cost and reliability as alternatives get better.