I read HN daily and posts about features or curious behavior of Claude Code are very common.

I see no posts about OpenAI Codex. I was under the impression that it also is a very popular tool and virtually as powerful as Claude Code.

Where am I wrong? Is Codex not popular? Is it significantly worse than Claude Code? Is it being "cancelled" because of OpenAI cooperating with the US Department of War?

I have mostly seen people using claude code, never heard anyone using Codex.

the harness or the model?

I don't use the harness anymore, but it was fine. I switched to an agnostic harness that uses a multitude of models -- often the GPT ones.

Anyone that cares about quality first and foremost switches harnesses pretty fast, the features are just insane. I want to be able to dictate the models that do vision versus code, for example. I want agentic memory that I can manage. I want better permissions controls than CC or codex, I want better file editing and reversion; better harnesses out there for all of these points.

The problem with Codex is that Codex is largely only for code due to its more aggressive sandboxing. Claude can do pretty much everything for your computer out of the box. It is not insurmountable, but it is a big barrier to convince anyone to use it for anything beyond code review.