GPT-5.5: The Honest Take on OpenAI's Response to Opus 4.7

OpenAI released GPT-5.5 today, exactly one week after Anthropic shipped Claude Opus 4.7. The timing is not subtle. Opus 4.7 took the SWE-Bench Verified crown at 87.6% and put Anthropic at the top of most third-party coding leaderboards; GPT-5.5 is the direct response. Worth flagging upfront: SWE-Bench Verified scores at this tier should be read with heavy skepticism. Every frontier lab has plausibly trained on or adjacent to this data, and Anthropic itself has acknowledged memorization signals o