Rourke McNamara

imagine walking around with closed macbook

OpenAI Codex CLI just got way more agentic. With /goal, you give it one objective and it keeps working across turns until the goal is done. Master it this weekend. Bookmark this.

software engineers before vs after agents

@thsottiaux Skill groups or profiles with skill/plugin sets. There are so many useful skills and plugins, but having them all on at once bloats context and costs more tokens. Being able to swap from “iOS app with golang backend” to “email triage and everyday work” mode would be sick

What are we obviously not getting right with Codex?

Anyway, please reply with the best Codex pet you’ve seen. Cute, cursed, personal, inexplicable, all welcome. I would like to inspect the tiny coworker economy.

And people didn’t just make pets. They made infrastructure for pets, because of course they did. Petdex is already a whole public gallery: x.com/RaillyHugo/sta…

Codex pets have been out for about five minutes and the community has already done the correct thing: made them weird, personal, nostalgic, and slightly cursed. Starting with @dkundel turning his actual dog into a tiny desktop coworker: x.com/dkundel/status…

a P2 quali result five weeks after Verstappen called the RB22 “undriveable” at Suzuka is some pretty impressive engineering

Show us the Codex pets you hatched. Use /hatch to create your own Codex pet. We’ll pick 10 favorites to get 30 days of ChatGPT Pro.

Bring your workflow to Codex in just a few clicks. Import settings, plugins, agents, project configuration, and more so you can keep working with fewer interruptions. Your move.

@jxnlco Why are they not done yet, Jason? 😂

ok so I'm on developer experience for codex? yes but codex is for everything yes so my scope is everything experience? yes, can you make a demo for everything? can you be more specific? yeah like all the work in the economy, can you make demos for that?

chatgpt.com/codex/for-work/ I don’t know what AGI looks like, but this feels directionally correct.

I ran Claude Code and Codex side by side across a range of tasks, over the past few weeks: - Research - Architecture design - Spec writing - Code refactoring Both maxed out on the top of the range models: - Opus 4.6 max effort & GPT 5.4 xhigh, then - Opus 4.7 max effort & GPT 5.5 xhigh There's one clear leader in my eyes. When I began, I defaulted to Claude Code, but after so many frustrating duels with Opus, I've fully switched over to Codex and GPT 5.5 as my default harness for working on codebases. @AnthropicAI took an early lead, but @OpenAI are fully back in the driver's seat IMO.

It's Thursday, so obviously there are Codex updates. No matter what you need to do, you'll do it smarter and faster with Codex. Do it for a few days and you'll never be able to go back

Still wondering how you can use Codex for (almost) everything? Codex can help with more of the work that supports the work, from organizing research to making spreadsheets, decks, and summaries.

Steering by default, both because I often think of something I want to add to the prompt after I hit enter and because you can often see what direction things are heading and course correct. Queuing as the alt keystroke for the times I just want to stack another task onto the same thread

Do you prefer queuing or steering in Codex? What's your use case for each?