The ability for ChatGPT (o3-mini-high in this case) to one-shot explanations for bugs that I spent hours debugging is a marvel, but is also a bit degrading. - ThreadSky

mergesort.me • 33 days ago

The ability for ChatGPT (o3-mini-high in this case) to one-shot explanations for bugs that I spent hours debugging is a marvel, but is also a bit degrading.

1 / 2

Comments

tylerjamesh.bsky.social•33 days ago

Feel the same. Have had some successes with Cursor/Claude asking “why doesn’t this compile?” When all Xcode is giving me is some bullshit about “unable to type check”

Every time it’s useful I think “nice!” then “…. fuck”

mergesort.me•33 days ago

I have a very strong belief that if iOS developers realized how far behind Xcode is compared to AI-infused IDEs like Cursor or Windsurf, they’d be having existential crises about their future as software developers who write code.

tylerjamesh.bsky.social•33 days ago

I know I am.

mergesort.me•33 days ago

Twins. 👬🏻

eldonut.bsky.social•33 days ago

I think even at this point, LLMs have a bigger context window than we do. Only so much information we can keep immediately in our short-term memory buffer, but GPTs can look at huge chunks of code all at once.

grork.bsky.social•33 days ago

Is this in the $20/mo or $200/mo plan?

marcpalmer.net•33 days ago

This is the most compelling sample I’ve seen so far. I remain skeptical of larger scale usage (like building whole complex apps) not least because of the huge amount of discussion needed to instruct an AI. However this would be quite easily spotted by any of your pals if asked to help…

mergesort.me•33 days ago

1. I did ask in our shared community. While I got a few decent suggestions, none of them worked.
2. It would have taken a lot more time to break down the problem, and providing the same amount of context isn’t realistic because it means asking someone to internalize tons of context-specific code.

marcpalmer.net•33 days ago

It wasn’t meant as a criticism - clearly only dummies though ;-) FWIW I just asked GPT something about intents and I think it might have solved the problem

cereal.bsky.social•33 days ago

Isn’t the 4o the most advanced model and better suited for coding than the o3 or o3 mini? I’m a plus subscriber and not sure which model to choose. (Learning SwiftUI currently.

mergesort.me•33 days ago

o3-mini was released today, and is the most advanced model. Previously o1/o1 Pro was the most advanced model, and both of them are generally best for coding thanks to their reasoning capabilities.

cereal.bsky.social•33 days ago

Ah gotcha. Thank you 🙏

Comments

Posting Rules

Reply