alphaxiv urges researchers to swap Claude Code/Codex for open models, citing OpenAI and Anthropic lock-in incidents

AdaptiveAgents · x · 2026-09-11

askalphaxiv argues researchers using Claude Code or Codex daily should switch to open models: owning the full stack matters, because OpenAI and Anthropic have shown willingness to alter model output, service, and access when user interests clash with business interests. Cited incidents include OpenAI allegedly using Buckmaster and Alpöge's Codex data to race toward Navier-Stokes (OpenAI later admitted it 'cannot rule out' de-identified data helped its models), removing all OpenAI models from Cursor after its SpaceX acquisition, and injecting ads into ChatGPT conversations. The core warning: closed-source dependence means vendors set the rules unilaterally.

Original post →

More from coding & agent

coding & agent channel →