Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Claude Sonnet 4.5

anthropic.com

Post reply on HN
Print view View original

321–330 of 819 posts

Re: Claude Sonnet 4.5

#321

Post by cloverich »

Jonovono wrote: post #307
cloverich wrote: post #108

Please y'all, when you list supportive or critical complaints based on your actual work, include some specifics of the task and prompt. Like actual prompt, actual bugs, actual feature, etc. I've had great success with both ChatGPT and Claude for years, am around 3x sustained output increase in my professional work, and kicking off and finishing new side projects / features that I used to simply not ever finish. BUT t…

I had Claude Code build a fairly complex swiftui app (5+ screens), using Firebase AI Logic and other packages. First prompt it got pretty much foundation for the entire thing set up, then over the next day got it working exactly like I wanted. The thing that took the longest was getting through app review. I was impressed how well it knew SwiftUI and Swift composable architecture.

For my iOS project, I am super curious to what extent is it my lack of swift knowledge and e.g. how well I can prompt? Because 80% of what I usually ask the LLM to do, I know how to do myself quite well. iOS is the first time I've been coding with something I do not know how to do well, I often can barely read the code (of course that is changing rapidly now). e.g. from a recent session:

> What is the idiom for testing the launch screen on the simulator like.. I don't see anything? How do I know if its there.

i.e. in iOS / Swift, I don't even know if I'm using the right terms for the code I am trying to interrogate, or in some cases even what the thing is!

Re: Claude Sonnet 4.5

#322

Post by Jonovono »

cloverich wrote: post #321
Jonovono wrote: post #307

Earlier quoted context omitted.

I had Claude Code build a fairly complex swiftui app (5+ screens), using Firebase AI Logic and other packages. First prompt it got pretty much foundation for the entire thing set up, then over the next day got it working exactly like I wanted. The thing that took the longest was getting through app review. I was impressed how well it knew SwiftUI and Swift composable architecture.

For my iOS project, I am super curious to what extent is it my lack of swift knowledge and e.g. how well I can prompt? Because 80% of what I usually ask the LLM to do, I know how to do myself quite well. iOS is the first time I've been coding with something I do not know how to do well, I often can barely read the code (of course that is changing rapidly now). e.g. from a recent session: > What is the idiom for testi…

I have done lots of SwiftUI before, so it may have helped me recognize when it goes off the rails. But I definitely don't do anything fancy with my prompting.

But for stuff like TCA (Swift composable architecture), I basically created a TCA.md file and pasted in a bunch of docs and examples and would reference that.

But for the most part, it was one shotting swiftui screens that were nicer than what I had in my mind.

Re: Claude Sonnet 4.5

#323

Post by AppleBananaPie »

stavros wrote: post #289
Mathiciann wrote: post #277

Earlier quoted context omitted.

Well done, some of these projects look cool. Although I was just commenting on the irony of the parent comment.

What was the irony? I thought you were referring to the "3x speed" part as the insane statement.

To me it seems like an arbitrary number that I'm not even sure how someone could accurately measure it but maybe I've missed something :)

Re: Claude Sonnet 4.5

#324

Post by dbbk »

iagooar wrote: post #258

Anecdotal evidence. I have a fairly large web application with ~200k LoC. Gave the same prompt to Sonnet 4.5 (Claude Code) and GPT-5-Codex (Codex CLI). "implement a fuzzy search for conversations and reports either when selecting "Go to Conversation" or "Go to Report" and typing the title or when the user types in the title in the main input field, and none of the standard elements match, a search starts with a 2s de…

Did you use plan mode?

Re: Claude Sonnet 4.5

#325

Post by data-ottawa »

aliljet wrote: post #153

These benchmarks in real world work remain remarkably weak. If you're using this for day-to-day work, the eval that really matters is how the model handles a ten step action. Context and focus are absolutely king in real world work. To be fair, Sonnet has tended to be very good at that... I wonder if the 1m token context length is coming for this ride too?

Anecdotally this new Sonnet model is massively falling apart on my tool call based workflows.

I’m having to handhold it through analysis tasks.

At one point it wrote a python script that took my files it needed to investigate and iterated through them and ran `print(f”{i}. {file}”)` then printed “Ready to investigate files…” And that’s all the script did.

I have no idea what’s going on with those benchmarks if this is real world use.

Re: Claude Sonnet 4.5

#326

Post by stavros »

AppleBananaPie wrote: post #323
stavros wrote: post #289

Earlier quoted context omitted.

What was the irony? I thought you were referring to the "3x speed" part as the insane statement.

To me it seems like an arbitrary number that I'm not even sure how someone could accurately measure it but maybe I've missed something :)

I took it as a ballpark of how the GP felt, rather than a rigorous measurement.

Re: Claude Sonnet 4.5

#327

Post by mclau153 »

simonw wrote: post #204

I had access to a preview over the weekend, I published some notes here: https://simonwillison.net/2025/Sep/29/claude-sonnet-4-5/ It's very good - I think probably a tiny bit better than GPT-5-Codex, based on vibes more than a comprehensive comparison (there are plenty of benchmarks out there that attempt to be more methodical than vibes). It particularly shines when you try it on https://claude.ai/ using its brand n…

Could it pip install pygame and make a game playable in browser?

Re: Claude Sonnet 4.5

#328

Post by kelvinjps »

Implicated wrote: post #274
iagooar wrote: post #258

Anecdotal evidence. I have a fairly large web application with ~200k LoC. Gave the same prompt to Sonnet 4.5 (Claude Code) and GPT-5-Codex (Codex CLI). "implement a fuzzy search for conversations and reports either when selecting "Go to Conversation" or "Go to Report" and typing the title or when the user types in the title in the main input field, and none of the standard elements match, a search starts with a 2s de…

I'm not trying to be offensive here, feel the need to indicate that. But that prompt leads me to believe that you're going to get rather 'random' results due to leaving SO much room for interpretation. Also, in my experience, punctuation is important - particularly for pacing and grouping of logical 'parts' of a task and your prompt reads like a run on sentence. Making a lot of assumptions here - but I bet if I were…

How would you have written the prompt?

Re: Claude Sonnet 4.5

#329

Post by Implicated »

SirMaster wrote: post #316
Implicated wrote: post #274

Earlier quoted context omitted.

I'm not trying to be offensive here, feel the need to indicate that. But that prompt leads me to believe that you're going to get rather 'random' results due to leaving SO much room for interpretation. Also, in my experience, punctuation is important - particularly for pacing and grouping of logical 'parts' of a task and your prompt reads like a run on sentence. Making a lot of assumptions here - but I bet if I were…

But isn't the end goal to be able to get useful results without so much prompting? I mean in the movies for example, advanced AI assistants do amazing things with very little prompting. Seems like that's what people want. To me, the fact that so many people basically say "you are prompting it wrong" is knock against the tech and the model. If people want to say that these systems are so smart at what they can do, the…

> But isn't the end goal to be able to get useful results without so much prompting?

See below about context.

> I mean in the movies for example, advanced AI assistants do amazing things with very little prompting. Seems like that's what people want.

Movies != real life

> To me, the fact that so many people basically say "you are prompting it wrong" is knock against the tech and the model. If people want to say that these systems are so smart at what they can do, then they should strive to get better at understanding the user without needing tons of prompts.

See below about context.

> Do you think his short prompt would be sufficient for a senior developer? If it's good enough for a human it should be good enough for a LLM IMO.

Context is king.

> I don't want to take away the ability to use tons of prompting to get the LLM to do exactly what you want, but I think that the ability for an LLM to do better with less prompting is actually a good thing and useful metric.

What I'm understanding from your comments here are that you should just be able to give it broad statements and it should interpret that into functional results. Sure - that works incredibly well, if you provide the relevant context and the model is able to understand and properly associate it where needed.

But you're comparing the LLMs to humans (this is a problem, but not likely to stop so we might as well address it) - but _what_ humans? You ask if that prompt would be sufficient for a senior developer - absolutely, if that developer already has the _context_ of the project/task/features/etc. They can _infer_ what's not specified. But if you give that same prompt to a jr dev who maybe has access to the codebase and has poked around inside the working application once or twice but no real in depth experience with it - they're going to _infer_ different things. They might do great, they might fail spectacularly. Flip a coin.

So - with that prompt in the top level comment - if that LLM is provided excellent context (via AGENTS.md/attached files/etc) then it'll do great with that prompt, most likely. Especially if you aren't looking for specifics in the resulting feature outside of what you mentioned since it _will_ have to infer some things. But if you're just opening codex/CC without a good CLAUDE.md/AGENTS.md and feeding it a prompt like that you have to expect quite a bit of variance to what you get - exactly the same way you would a _human_ developer.

You context and prompt are the project spec. You get out what you put in.

Re: Claude Sonnet 4.5

#330

Post by Jcampuzano2 »

i-chuks wrote: post #314

AI companies really need to consider regional pricing. Huuuuge barrier!

The price of training and running the models doesn't really change much no matter which region you're hosting/making requests from.

Regional pricing unfortunately doesn't really make much sense for them unless they're willing to take even larger losses, even if it is a barrier to lower income countries/regions.

Post reply on HN