While I like the models from OpenAI, I did not have as great experience with their harness or quota. I could easily consume 10% weekly quota per hour with astra in light effort and only a single agent (business premium 5x). Also I really miss the ergonomics in Claude code whenever I try codex or even opencode, in codex it still feels bare minimum feature set.
Now I’m using my codex quota by telling Claude to run the codex plugin as developer sub agent, good way to spot model weaknesses between the two and I get to keep the Claude code harness.
I was an OpenAI user, then switched to Claude and have had a lot of success with it. Despite the switching costs, I'm tempted to try it. What have people's experiences with switching been, if anyone else has done it yet.
I was a big fan of Anthropic. But it looks as if they are becoming greedy and losing momentum. I still have a subscription, I like Claude Design, and I want to keep comparing it with Codex and sharing tasks between the two subscriptions. But I’m getting more and more excited about OpenAI’s work.
They have a much better image model. GPT 2.5 is great, but even GPT 2.0 was awesome. Their voice is like real Jarvis, and now Astra has computer and browser-use capabilities.
They are still neck and neck, but it looks as if OpenAI has every chance to pull ahead faster.
What's interesting is that loyalty in this space seems to be collapsing into whoever removes friction from the actual build loop, not whoever wins the benchmark race. I've watched teams switch stacks mid-project for exactly this reason, and it makes vendor lock-in a much weaker moat than most AI companies assume. Do you think Anthropic loses you for good, or is this just until the next Claude release resets the whoa meter?
While I like the models from OpenAI, I did not have as great experience with their harness or quota. I could easily consume 10% weekly quota per hour with astra in light effort and only a single agent (business premium 5x). Also I really miss the ergonomics in Claude code whenever I try codex or even opencode, in codex it still feels bare minimum feature set.
Now I’m using my codex quota by telling Claude to run the codex plugin as developer sub agent, good way to spot model weaknesses between the two and I get to keep the Claude code harness.
Great writeup! I agree this probably isn't AGI, but the fact that we're even asking that question is proof of how good this is.
Yeah really sucks that you can no longer use Opus 4.8 in Claude Code, thinking of switching to Codex too because of this..
Opus 5 is a bad model in more ways than one
/model claude-opus-4-8
I was an OpenAI user, then switched to Claude and have had a lot of success with it. Despite the switching costs, I'm tempted to try it. What have people's experiences with switching been, if anyone else has done it yet.
I have a similar story.
I was a big fan of Anthropic. But it looks as if they are becoming greedy and losing momentum. I still have a subscription, I like Claude Design, and I want to keep comparing it with Codex and sharing tasks between the two subscriptions. But I’m getting more and more excited about OpenAI’s work.
They have a much better image model. GPT 2.5 is great, but even GPT 2.0 was awesome. Their voice is like real Jarvis, and now Astra has computer and browser-use capabilities.
They are still neck and neck, but it looks as if OpenAI has every chance to pull ahead faster.
What's interesting is that loyalty in this space seems to be collapsing into whoever removes friction from the actual build loop, not whoever wins the benchmark race. I've watched teams switch stacks mid-project for exactly this reason, and it makes vendor lock-in a much weaker moat than most AI companies assume. Do you think Anthropic loses you for good, or is this just until the next Claude release resets the whoa meter?