Ivan Juras

@thsottiaux 1. Hooks (on Windows, too), 2. Better orchestration with skills (look up context:fork from Anthropic), 3. Front end is not there .

What are we obviously not getting right with Codex?

@ajambrosino 1. Give us all hooks on Windows. 2. Improve the orchestration tools (context: fork -> Claude has it and it's amazing) 3. Improve the frontend skills.

what should we do in the next 3 months?

@gdb I don't see the /goal command on Windows.

codex now has a built in Ralph loop++:

Don't sacrifice clarity for speed. It won't work.

This is the difference between using AI to skip understanding and using it to accelerate understanding. Same tool. Opposite intent. One leaves you with debt. The other leaves you with leverage.

There's a concept I keep running into while building with AI: mental model debt. It's like technical debt, but worse. You stop understanding the thing you built. Then you can't steer it. 🧵

@borjafat SEO

Claude Opus 4.7 or GPT-5.5 can run your ENTIRE SEO Tired of chasing backlinks? Solved. Tired of AI slop articles? Solved. Tired of ranking for keywords that don't convert? Solved You just install a SKILL and you're ready to roll. Can run automated daily tasks on a schedule Comment "SEO" and I'll send it!

GPT 5.5 is so good now that I'm back to using ChatGPT as the main chat app. Who would've predicted that?

I have alot of respect for solo AI builders. The world is their playground now.

@iruletheworldmo Just tell me the orchestration is on par with Claude.

i’d counted gemini out because their models lacked agency but they look back in the race. the next (imminent) gemini model will the new sota and it may be hard to displace them. looks like they’ll have their own codex moment at last. the giant is waking up.

@jun_song How do you automate putting the project folder into the ChatGPT UI?

I just realized, instead of doing everything with Codex, asking GPT Pro to direct the tasks, evaluate the intermediate outputs, and make detailed prompts for Codex saved a massive amount of tokens.

@thsottiaux Don't abandon the CLI.

With some small tweaks, Codex can work for days on hard tasks. We will release some changes to make this easier to use for everyone. What’s the hardest task you’ve seen GPT-5.5 succeed at?

@lilyraynyc Everyone has these in their portfolio (whether their own site or the client's). If you manage more than 1-2 sites, it's inevitable.

Would you hire an SEO company or an SEO/AI content tool if their organic trajectory looked like this? What if I told you these were some of the most highly visible companies calling themselves the "best" in their respective categories for SEO/AI content support?

@bridgemindai Time to take a serious look at your benchmarks.

GPT 5.5 just debuted on BridgeBench. It ranks below GPT 5.4. Read that again. The "most intelligent model ever built" scores worse than its predecessor on real world vibe coding. #8 overall. 84.5 quality. GPT 5.4 sits at #6 with 85.1. Behind Claude Opus 4.6. Behind Claude Opus 4.7. Behind Claude Sonnet 4.6. Behind Grok 4.20. Behind Qwen 3.6 Plus. The smartest model in the world is not the best coding model in the world. BridgeBench just proved it. bridgebench.ai

@arena @OpenAI Is there anyone who honestly thinks Opus is better than 5.5 at anything other than frontend? I use both heavily and it's not even a contest.

GPT-5.5 by @OpenAI is now live in the Arena, landing across multiple leaderboards. Here’s how it ranks by modality: - Code Arena (agentic web dev): #9, a strong +50pt jump over GPT-5.4 - Document Arena (analysis & long-content reasoning): #6, on par with Sonnet 4.6 - Text Arena: #7, Math #3, Instruction Following: #8 - Expert Arena: #5 - Search Arena: #2 - Vision Arena: #5 Strong, well-rounded performance, especially in Code (+50 pts vs GPT-5.4). Congrats to @OpenAI on the release. Full category breakdowns by modality in the thread.

@thsottiaux @thsottiaux Missing hooks for Windows and the context:fork Claude Code construct. Please, add these :)

It’s the little things that matter, what are some small papercuts you have noticed in Codex? We’ll fix as many as possible in the next week.

@BLUECOW009 Why, though? All other models degrade heavily, even after 100k tokens. Its compact mode is amazing, and seemless.

my only complain for 5.5 is context window that is only 256k