It’s Not Codex, It’s Codex/GPT-5-Codex

I’d just like to interject for a moment. What you’re referring to as Codex, is in fact, Codex Harness, or as I’ve recently taken to calling it, Codex/GPT-5-Codex. Codex by itself is not a coding agent; it is another execution layer of a fully functioning large language model made useful by the GPT-5-Codex core, model orchestration, tool APIs, and vital guardrails comprising a full agent system as defined by OpenAI.

Many users run Codex every day, without realizing it. Through a peculiar turn of events, the version of Codex which is widely deployed today is often called just “Codex”, and many of its users are not aware that it is basically the GPT-5-Codex model, developed by OpenAI.

There really is a Codex, and these people are using it, but it is just a part of the system they use. Codex is the harness: the thing that wires tools, sandboxes, repos, and permissions together. The GPT-5-Codex model is the thing that actually understands your code, reasons about abstractions, and writes the patches.

Codex is an essential part of an agent system, but useless by itself; it can only function in the context of a capable model. Codex is normally used in combination with GPT-5-Codex: the whole system is basically Codex/GPT-5-Codex. All the so-called “Codex” distributions are really distributions of Codex/GPT-5-Codex.

If you use Codex but don’t acknowledge GPT-5-Codex, you’re obscuring the crucial role played by the model. It’s not just a matter of credit; it’s about correctly understanding how your agent works. Calling the whole thing “Codex” is like calling the human brain “a skull”.

So please, next time, call it Codex/GPT-5-Codex, and help us keep technical terminology accurate.