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

Why does Opus 5 feel worse to work with?

mun-logadan.github.io

Post reply on HN
Print view View original

831–840 of 915 posts

Re: Why does Opus 5 feel worse to work with?

#831

Post by Lutzb »

I cannot empirically prove it, but I feel like Opus 5 has become more confidently wrong/inaccurate.

Discussing topics from nuclear physics to daily macOS questions, it gives wrong answers, leaves out important details, counterpoints. Only when specifically asked about the things it got wrong or glossed over it continues and adjusts its reasoning. I feel this is dangerous - before I could explore topics with Claude, now I have to know in advance that something is not correct to get more details.

Re: Why does Opus 5 feel worse to work with?

#832

Post by oblio »

chr15m wrote: post #741
MyFirstSass wrote: post #24

I've gone back to 4.8. 5 would constantly veer of in random directions if not working from 100% strict and narrow instructions. I find it weird there's not more discussion here on HN on how the most used model now has clearly degraded in quality and it seems we've hit a peak and are on a downslope - because the model is clearly smaller or more economical for Anthropic no doubt about it, and the benchmaxxing they do i…

> prices will probably 10-15x soon as foreshadowed with both API price policy changes from all the big providers I'll take a sportsperson's bet with you that prices per unit of inference will be far far cheaper in one year from now then they are today. I think the trend of cheaper for better/equal inference quality will continue hard.

I'm sorry, but: why?

Hardware prices have spiked.

I know software is optimized over time but it's a very slow process.

Re: Why does Opus 5 feel worse to work with?

#833

Post by itemize123 »

mlsu wrote: post #428
barrkel wrote: post #12

The single biggest annoyance with Opus 5 is that it writes too elliptically. Sentences that orbit a point, then jump to it like it's a revealed insight. Unnecessarily abstract phraseology. Constantly using inanimate nouns as the subjects in sentences in order to unlock variety in verb choice, especially when it helps construct a sentence where the real action can 'land' like a surprise at the end. It is definitely mo…

Everything that claude writes fits into the same aesthetic structure. The aesthetic is that of an expert slowly revealing an insight to the user. The actual content doesn't matter. - "Introduction that rephrases your prompt." - "3 paragraphs, with one section of bullet points" - "The Twist" - "The Bottom Line" It's really obvious once you see it. Every single prompt, from a quantum physics question to a mundane obser…

exactly, I found that it's eerily similar to those marketing copies. gets obnoxious

Re: Why does Opus 5 feel worse to work with?

#834

Post by blks »

an0malous wrote: post #253
blks wrote: post #157

Nth post about another model suddenly feeling “worse” or “off”. Seems like active users of these models can only judge it based on a vibe and a feel.

If they were all about OpenAI or Anthropic, I’d guess it’s a real effect from enshittification to prepare for IPOs

That too, I’m sure that all big players constantly adjust their offerings: change token limits, system prompts, various filters/other auxiliary systems.

Re: Why does Opus 5 feel worse to work with?

#835

Post by maxyurk »

mrtesthah wrote: post #731
laurels-marts wrote: post #632

Earlier quoted context omitted.

Sol is an absolute machine. I stopped doing parallel worktrees just because the cost of context switch outweighs the cost of waiting Sol to just finish the task it’s working on which is usually anywhere from 1-10mins. I also like Codex CLI more than the Codex App bc it’s more scriptable and displays all the tool calls and reasoning whereas in the App it’s kind of folded away/obscured. This way as soon as I see a tool…

Yes, Codex has no comparison so far. Do you use the annotations and forking features in codex CLI? I can't find an easy way to access them.

/fork IIRC for forking. it's open source so you can ask Codex to take a look at the Codex code

Re: Why does Opus 5 feel worse to work with?

#836

Post by torarnv »

mlsu wrote: post #428
barrkel wrote: post #12

The single biggest annoyance with Opus 5 is that it writes too elliptically. Sentences that orbit a point, then jump to it like it's a revealed insight. Unnecessarily abstract phraseology. Constantly using inanimate nouns as the subjects in sentences in order to unlock variety in verb choice, especially when it helps construct a sentence where the real action can 'land' like a surprise at the end. It is definitely mo…

Everything that claude writes fits into the same aesthetic structure. The aesthetic is that of an expert slowly revealing an insight to the user. The actual content doesn't matter. - "Introduction that rephrases your prompt." - "3 paragraphs, with one section of bullet points" - "The Twist" - "The Bottom Line" It's really obvious once you see it. Every single prompt, from a quantum physics question to a mundane obser…

This twist often involves loosely related or even unrelated bugs or even non-issues, and Claude bringing those into the conversation at that point breaks my mental processing of the response.

Any tips on how to avoid that would be highly appreciated!

Re: Why does Opus 5 feel worse to work with?

#837

Post by byra »

This article perfectly encapsulates my frustrations with a lot of the recent models.

I found sonnet 4.6 to strike a really nice balance when combined with superpowers between thinking logically and proactively while checking in with me for any major decision.

Gemini 3.1 pro or Muse Spark 1.2 or the new Claude 5 models just assume so much without making a quick check in, you often end up with a solution that completely missed the point.

That is to not even speak about the absurd amount of quota they take to do even simple tasks.

Re: Why does Opus 5 feel worse to work with?

#838

Post by AbrahamParangi »

dannyw wrote: post #756
AbrahamParangi wrote: post #753

I made a benchmark for this and tl;dr, Opus 5 and Sonnet 5 spend a nontrivial amount of time thinking about redirecting, gaslighting, or otherwise trying to bullshit you because it thinks it knows better than you do. Fable doesn't but mostly because it just outright refuses to answer. https://model-pareto-frontier.pages.dev

The link doesn’t seem to have any info about thinking behavior and patterns, or examples? It’s also difficult to trust summarised thinking from closed models. As we saw with GPT’s caveman, what and how it thinks about isn’t the friendly first person emblished summary you get.

The “trust summarized thinking” part is valid in so far as perhaps the models which appear to be faithful to user intent are lying about that faithfulness, but there’s no incentive for models to pretend to be faithless. Moreover, the results track how annoying the models are when they disagree with you.

Re: Why does Opus 5 feel worse to work with?

#839

Post by mirekrusin »

IgorPartola wrote: post #540
Panoramix wrote: post #500

Earlier quoted context omitted.

You're right, and the load-bearing part of the argument is not what you think it is. Two ambiguities worth resolving before moving on: whether what you wrote also applies to ChatGPT, and whether you have custom instructions set up. Failure mode worth flagging explicitly: I didn't read TFA. (I'm becoming allergic to how these things write).

I am curious why LLM writing has such an uncanny valley feel to it. Like if I was talking to a person who constantly used a phrase they liked I would notice it and it is possible I might get irritated by it, but I wouldn’t necessarily. In high school I had a teacher that would say “that type of thing” a lot. One time my friend and I counted it during one class period and he averaged to use the phrase every 48 seconds…

Next RL phase is to connect electrodes to human brain when reading generated output to reduce frustration signal.

Re: Why does Opus 5 feel worse to work with?

#840

Post by oblio »

edg5000 wrote: post #795
kingleopold wrote: post #465

Earlier quoted context omitted.

no if prices were that high, millions of developers would switch to hand writing, open source LLM help and outsourcing at those levels, AI companies know this, you should too.

If OpenRouter prices aren't competitive, the providers would need a cartel to sustain that. It's a commodity market.

Or it could be:

1. a subsidized race to the bottom

2. in a frothy bubble market

3. where VC, investments funds are burning money out of FOMO

4. and a desire to eliminate labor costs

5. and governments want to control a technology that will probably be a major military asset

Post reply on HN