Wes Roth

Claude Just Changed Overnight

The video details Anthropic’s abrupt decision to block OpenClaw, an open‑source AI‑agent framework, from using Claude Max subscriptions. The move effectively ends the cheap, flat‑rate token subsidy that many developers relied on to run large‑scale agent swarms, sparking a heated debate across the AI community. Key insights reveal that Anthropic cited excessive costs from unoptimized third‑party usage and low cache‑hit rates as justification. Their internal tools, Claude Code and Claude Co‑Work, are engineered to reuse prompt caches, dramatically reducing compute expenses. OpenClaw’s architecture bypassed these efficiencies, making it far more expensive for Anthropic to serve, prompting the rollout of an “Extra Use” plan that charges per‑token at a discount only after substantial spend. Prominent voices such as OpenClaw creator Peter Steinberger and Cloud Code founder Boris Churnney underscore the controversy. Steinberger accuses Anthropic of a “copy‑then‑close” strategy—mirroring open‑source innovations in their own products before cutting off the community. Usage data cited in the video shows a single week of $200 API costs on Claude Sonnet 4.6, illustrating how quickly tokens can be consumed under the previous model. The fallout signals a potential shift in the AI tooling ecosystem. Startups and hobbyists who built products on Anthropic’s affordable tier now face higher, metered costs or must migrate to competing platforms like OpenAI. The episode raises broader questions about platform control, developer trust, and the sustainability of open‑source contributions when large AI firms prioritize proprietary ecosystems over community collaboration.