Live data from Hacker News

Claude Sonnet 4.5

anthropic.com

331–340 of 819 posts

Re: Claude Sonnet 4.5

#331
post #202
post #40

Earlier quoted context omitted.

In my experience Gemini 2.5 Pro is the star when it comes to complex codebases. Give it a single xml from repomix and make sure to use the one at the aistudio.

Gemini 2.5 Pro = Long context king, image input king GPT-5 = Overengineering/complexity/"enterprise" king Claude = "Get straightforwaed shit done efficiently" king

On the plus side, GPT5 is very malleable, so you CAN prompt it away from that, whereas it's very hard to prompt Claude into producing hard code: even with a nearly file by file breakdown of a task, it'll occasionally run into an obstacle and just give up and make a mock or top implementation, basically diverge from the entire plan, then do its own version.

Re: Claude Sonnet 4.5

#332
post #272

Earlier quoted context omitted.

> It’s a simple substitution request where I provide a Lint error that suggests the correct change. All the models fail. I could ask someone with no development experience to do this change and they could. I don't understand why this kind of thing is useful. Do the thing yourself and move on. For every one problem like this, AI can do 10 better/faster than I can.

How can I trust it to do the complicated task well when it fails to do the simple thing?

It's not about simple vs. complex. It's about the types of tasks the AI has been trained on: pattern-matching, thinking, reasoning, research.

Tasks like linting and formatting a block of code are pretty simple, but also very specialized. You're much better off using formatters/linters than an AI.

Re: Claude Sonnet 4.5

#333

I need to try Claude - haven't gotten to it. I use AI for different things, though, including proofreading posts on political topics. I have run into situations where ChatGPT just freezes and refuses. Example: discussing the recent rape case involving a 12-year-old in Austria. I assume its guardrails detect "sex + kid" and give a hard "no" regardless of the actual context or content. That is unacceptable. That's like…

In my (admittedly very limited) experience with trying to talk about "controversial" topics, Claude seems to be much stricter about shutting down the conversation fast.

Re: Claude Sonnet 4.5

#334

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…

How would you have written the prompt?

tbh, I don't really understand it well enough to be able to give a response here. But here's a real prompt I just used on a project copy/pasted:

```

Something that seems to have been a consistent gotcha when working with llm's on this project is that there's no specific `placement` column on the table that holds the 'results' data. Our race_class_section_results table has it's rows created in placement order - so it's inferred via the order relative to other records in the same race_class_section. But this seems to complicate things quite a bit at times when we have a specific record/entry and want to know it's placement - we have to query the rest of them and/or include joins and other complications if we want to filter results by the placements, etc.

  Can you take a look at how this is handled, both with the querying of existing data by views/livewire components/etc and how we're storing/creating the records via the import processes and give me a
   determination on whether you think it should be refactored to include a column on the database for `placement`? I think right now we've got 140,000 or so records on that table and it's got nearly
  20 years worth of race records so I don't think we need to be too concerned with the performance of the table or added storage or anything. Think very hard, understand that this would be a rather
  major refactor of the codebase (I assume, since it's likely used/referenced in _many_ places - thankfully though that most of the complicated queries it would be found in would be easily identified
  by just doing a search of the codebase for the race_class_section_results table) and determine if that would be worth it for the ease of use/query simplification moving forward.
```

This comes with a rather developed CLAUDE.md that includes references to other .md documents that outline various important aspects of the application that should be brought into context when working in those areas.

This prompt was made in planning mode - the LLM will then dig into the code/application to understand things and, if needed, ask questions and give options to weigh before return with a 'plan' on how to approach. I then iterate on that plan with it before eventually accepting a plan that it will then begin work on.

Re: Claude Sonnet 4.5

#335
post #89
post #76

Earlier quoted context omitted.

Compact?

/compress or something like that, basically taking the context and summarizing it.

Cursor does this automatically, although I wish there was a command for it as well. All AIs start shitting the bed once their context goes above 80% or so.

Re: Claude Sonnet 4.5

#336

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…

You're critiquing OP for not playing with how the models currently work (bad at gathering context on their own). Sure, if you bend over backwards and hop on one foot, you can get them to do what you want.

OP is critiquing the model as a product vs. the marketing promises. The model should be smart enough to gather context about the project to implement features properly on their own, if they are ever going to 'write 90% of all code THIS YEAR' as people like the founder of Anthropic claim.

Re: Claude Sonnet 4.5

#337
Unfortunately also disappointed with it in Cursor vs GPT-5-Codex. I asked it to add a test for a specific edge case, it hallucinated some parameters and didn't use existing harnesses. GPT-5-Codex with the same prompt got everything right.

Re: Claude Sonnet 4.5

#338

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…

How do you measure 3x sustained output increase? Is it number of lines? Tickets closed? PRs opened or merged? Number of happy customers?

Merged PRs. We typically plan out our work, break up into e.g. JIRA tasks, then when we create PR's _very generally_ they should be tied to actual JIRA tickets i.e. pre-planned work. A ticket is usually a requested feature or bug (as reported by an actual user). So my PR rate, or perhaps less controversially my JIRA close rate, is around 3x higher for the last few months. That's also reflected more generally in my feedback productivity wise (i.e. people that are looking at the project as a whole rather than e.g. how many commits I've made). I exclude from 3x side projects and CLI tools, which are weird to quantify - they are typically things that would usually have been ideas in my head I never did at all. I guess I also generally exclude refactoring although I do that more. For example I had claude fix a bug that was dogging our typescript compilation. I couldn't figure out what was so slow about it (>60s to compile). Turned out it was a specific recursive type pulled in by a specific version of a library mixed by usage from one file! It actually took it a while to figure it out, it kept proposing solutions and I had to re-direct it a bunch, using mostly just intuition as opposed to experience. e.g. "No, re-run the diagnostics and look at the debug output, give me three examples of area / commands you could look at and how" and then I'd pick one. I just did that task on the side, I'd go back and look at it output once every day or two, then prompt it with something else, then just go do my usual tasks as though that didn't exist. That type of work given our pace / deadlines / etc, might never have gotten done at least not anytime soon. But I do stuff like that all the time now, I just don't often measure it.

Is that helpful?

Re: Claude Sonnet 4.5

#339
post #92

Earlier quoted context omitted.

That’s such a great analogy. I always say GPT is like the genius that completely lacks common sense. One of my favorite things is when I asked it why the WiFi wasn’t working, and showed it a photo of our wiring. It said that I should tell support: > “My media panel has a Cat6 patch panel but no visible ONT or labeled RJ45 hand-off. Please locate/activate the Ethernet hand-off for my unit and tell me which jack in the…

I'm curious what you would have expected it to reply given the input you provided?

Er, I said it in my post, but calling support and saying “can you set up the WiFi” would have been fine.

Re: Claude Sonnet 4.5

#340

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…

How do you measure 3x sustained output increase? Is it number of lines? Tickets closed? PRs opened or merged? Number of happy customers?

I think it's just a meaningless sentence.
Post reply on HN