There is minimal downside to switching to open models

What happened

An essay on marble.onl argues that open-weight models are now strong enough that migrating off Claude/GPT in response to Anthropic's new ID verification requirement carries minimal real productivity cost.

Context and impact

The post became a vehicle for the broader 'cancel Claude' mood — the HN thread has 99 points and many comments share concrete setups using GLM-5.2, Qwen 3, Apertus and self-hosted inference. For Anthropic it is a sentiment signal in a sensitive pre-IPO window.

Details

• Author: marble.onl
• Core claim: open models now cover >90% of typical developer workflows
• Concrete recommendations for coding agents
• Discussion points at GLM-5.2 (MIT) and Apertus as replacements
• Connects to the wider Claude ID verification backlash