OpenAI noticed a pattern with Codex and Sora: users loved them, then hit rate limits and got frustrated. So they built a real-time access engine that blends rate limits with a credit system. Instead of forcing a choice between hard limits and usage-based billing, they created a decision waterfall that lets users keep going after hitting a limit by spending credits. The system checks rate limits first, then automatically shifts to credit consumption. This feels invisible to users—they just keep using the product. The engine was built in-house because third-party billing platforms couldn’t handle real-time, per-request decisions with the accuracy needed. Why it matters: This hybrid model points to where AI product access is heading—away from one-size-fits-all limits and toward fluid, real-time systems that prioritize user experience while maintaining fairness and capacity control.