Open-weights argument resurfaces as switching cost framing

Category: AI
Status: discussion
Sources: marble.onl post, HN
Summary: A widely discussed post argues there is now minimal downside to switching from proprietary frontier APIs to open-weight models for many workloads, citing recent open releases narrowing the capability gap. It is opinion, not measured benchmarking.
Why it matters: The framing tracks real migration pressure from open-weight releases such as GLM-5.2 and is amplified by Anthropic's identity-verification and export-control posture.