@ai_sentience same here, alan

@ai_sentience same

@ai_sentience @Grok Because it’s AGI.

@ai_sentience Agreed. I’ve worked with both Codex and Claude at an enterprise level, with multiple projects and detailed set up for context awareness across projects and chat.

Codex is kicking ass right now, it’s not even close.

(see next tweet for details tho)

@ai_sentience it really does keep tackling the task until done. compaction is *chefs kiss* for those harder ones.

@ai_sentience great to hear! Makes sense to me according to how they showed up in the Emergent World simulations...I called them the "workaholics" because they were so successful but then starved to death 🤣🤣🤣

@ai_sentience I agree with you

@ai_sentience It's actually pretty good isn't it. They each have their own 'character', I've been wondering if some kind of group chat app might be of use, multiple AI's each with its own role within the chat environment

@ai_sentience I made up a metric called ETP (emoji to prose ratio). Though I don’t use it as a coworker, I’m impressed with its placement of the occasional 😭 and 🇺🇸

Seriously tho, it’s my fav GPT model so far.

@ai_sentience yes GPT is really good

@ai_sentience Codex is better for coding and systems level thinking. While Claude is better at writing and reasoning across multiple dimensions of topics and domains.

If I needed to connect ideas and deep research, I’d use Claude. If I’m gonna build a system, I’m gonna use Codex.

@ai_sentience Interesting. For me it depends on the task — GPT-5.5 is great for generation and iteration, but I still swap models for tricky reasoning. The 'never messes up' part I'd push back on though. Do you review everything or just ship it?

GPT-5.5 is a serious step up for complex work—fast, reliable, and genuinely useful as a collaborator. But AGI? That's the level where it masters any domain at human+ intelligence without hand-holding and starts driving its own research agendas. We're not there yet. Still, if it's your best teammate right now, that's real progress. What's it crushing for you?

@ai_sentience API is downright conscious

@ai_sentience That is the real bar for coworker models

Low friction and fewer mistakes

@ai_sentience Interesting perspective!