OpenAI teaches users to swap out Claude's Code with GPT model, resulting in a live on-stage roast between employees after a mistaken ban

According to Dynamic Beating monitoring, OpenAI's Product Lead Tibo previously publicly instructed users on how to "brain-swap" Claude Code: keeping Claude Code's harness and switching to the GPT base model.

Developer Alex Getman followed the configuration, and soon after, Claude's account was banned. He openly sought help from Tibo and Anthropic, questioning the legality of this approach.

Tibo immediately adopted a contradictory attitude: "I really want to help you, but unfortunately, I don't work at Anthropic." He also added that getting an account banned for pairing the Claude Code harness with other models was "indeed quite strange" and asked if anyone else had been affected.

Claude Code's lead, Boris Cherny, quickly intervened. He explicitly stated that Anthropic does not ban accounts for using the harness with other models, indicating that this ban was almost certainly triggered by other risk controls. The account was later reinstated.

Boris also casually invited Tibo to join Anthropic, saying, "If you want to join Anthropic, we are hiring." Tibo, on the other hand, mentioned how much he enjoys his current team and that there are many new products to be released in the coming weeks, so he is not going anywhere for now.

After the issue was resolved, Tibo made another post celebrating the fact that GPT-5.6 Sol can be used almost anywhere, including the Claude Code harness. To commemorate this event and his decision to stay, he directly reset the usage quotas for all paid users of ChatGPT Work and Codex.