OpenAI Restores GPT-4o After GPT-5 Rollout Sparks User Revolt

Consumer tech has seen mis-calculated upgrades before—think Coca-Cola’s 1985 “New Coke” or Microsoft’s 2012 Windows 8 start-menu removal—but this is the first time the object of attachment is a synthetic mind that users treat as confidant, co-author and therapist. The episode exposes two long-wave forces: (1) the scaling law arms-race that optimizes benchmarks and cost, not affective experience; and (2) the dawning reality that users will demand continuity and agency over the personalities of their AIs, much as they once demanded data portability from social networks. On a century arc, this kerfuffle may seem minor, yet it foreshadows the social contract that will govern human–AI symbiosis: whoever controls the “soul layer” of AI—tone, memory, values—controls user loyalty and, by extension, political and economic power. OpenAI’s quick climb-down suggests that even trillion-parameter models are hostage to the softer metrics of trust and identity, hinting that progress toward AGI will be shaped as much by anthropology as by compute budgets.