Nathan

Professional Builder

[Nathan @Submorf]

@ClaudeDevs Good call

[OpenAI @OpenAI]

We’re bringing new capabilities to GPT-Rosalind, a model series purpose-built for life sciences research at enterprise scale.

It brings GPT-5.5’s agentic coding and tool use together with stronger intelligence for drug discovery, analysis, design, and experimental workflows.

https://t.co/SrAJ3Mt7ka

[Ian @dumbfook]

openai really cooked with figuring out how to make codex the default platform to get work done

[Nathan @Submorf]

ChatGPT and Codex are combining into one product. 🤔

[Nathan @Submorf]

@VictorTaelin If the goal's success criteria is properly defined there is no possibility of a loophole. If you leave it vague it starts to assume.

[Nathan @Submorf]

@VictorTaelin Are you planting the loophole?

[Matt Popovich @mpopv]

me: okay codex now let's add this new feature to the app i had you start building three minutes ago

codex: absolutely! first let's add the NEW_FEATURE_ENABLED flag (defaulting to false) to ensure backwards compatibility with the legacy system,

[Nathan @Submorf]

@thsottiaux How do you guys debug these types of generalized claims? Seems like there would be endless potential fail points.

[Nathan @Submorf]

Just download amphetamine (free app in App Store)

purchasing this is not necessary.

[will depue @willdepue]

Tired of holding your laptop half open to keep your agents running?

Introducing AgentPlug: A USB-C dummy plug that keeps your Mac in clamshell mode by pretending to be an external display!

No commands, no security worries (just pull it out to stop!), no hassle.

[Nathan @Submorf]

@thsottiaux @RhysSullivan Huge🔥

[Nathan @Submorf]

You don’t even need to type /goal.

Just tell it to do a thing and to not stop until a criteria is met.

[nic @nicdunz]

guys, /goal is already in the codex app and has been since it was a thing. just put it in your prompt and send it and youll see. it doesnt show anything when you type it but it works.

again: it does not show anything when you type it, but it works. just write it in your prompt and send it. idk why everyones struggling so much to do this.

[nic @nicdunz]

guys, /goal is already in the codex app and has been since it was a thing. just put it in your prompt and send it and youll see. it doesnt show anything when you type it but it works.

again: it does not show anything when you type it, but it works. just write it in your prompt and send it. idk why everyones struggling so much to do this.

[Riley Brown @rileybrown]

when /goal in codex app?

[Simon Høiberg @SimonHoiberg]

GPT 5.5 is so good at doing stuff. Tool calling, searching, coding, handling long, iterative jobs. Incredible.

It's still horrible to talk to/brainstorm with.

It will agree with everything I say. Rarely push back (unless I tell it to, but that defeats the purpose).

It'll tell me how great my questions are and that all my ideas are amazing.

If I correct it on anything (even if my judgement is clearly wrong) it'll tell me I'm "absolutely right"

Sigh....

Opus is such a pleasure to talk to in this regard. Gemini also (even though the empathy thing is super annoying, but it will at least push back and challenge/correct me if I'm wrong).

Why is GPT 5.5 still such a doormat? It's infuriating.

[Nathan @Submorf]

This

[Robin Ebers · AI for Non-Coders @robinebers]

In Codex, I wish there was a model selector before you accept a plan.

feels like planning with High/Extra High and implementing with Medium is pretty common

[Robin Ebers · AI for Non-Coders @robinebers]

In Codex, I wish there was a model selector before you accept a plan.

feels like planning with High/Extra High and implementing with Medium is pretty common

[𝘼𝙡𝙚𝙭 @ItsAlexhere0]

Unpopular Opinion:

CODEX > Claude Code
Feels pretty obvious at this point

[Nathan @Submorf]

@aryanlabde Absolutely

[Nathan @Submorf]

Codex OP

[adam @theCTO]

hey @sama can we normalize models just saying "i dont know" ?

eliminates 99% of hallucinations