It's screwing up even in very simple rebases. I got a bug where a value wasn't being retrieved correctly, and Claude's solution was to create an endpoint and use an HTTP GET from within the same back-end! Now it feels worse than Sonnet.
All the engineers I asked today have said the same thing. Something is not right.
A model or new model version X is released, everyone is really impressed.
3 months later, "Did they nerf X?"
It's been this way since the original chatGPT release.
The answer is typically no, it's just your expectations have risen. What was previously mind-blowing improvement is now expected, and any mis-steps feel amplified.
What we need is an open and independent way of testing LLMs and stricter regulation on the disclosure of a product change when it is paid under a subscription or prepaid plan.
Unfortunately, it's paywalled most of the historical data since I last looked at it, but interesting that opus has dipped below sonnet on overall performance.
I mean, that's part of the problem: as far as I know, no claim of "this model has gotten worse since release!" has ever been validated by benchmarks. Obviously benchmarking models is an extremely hard problem, and you can try and make the case that the regressions aren't being captured by the benchmarks somehow, but until we have a repeatable benchmark which shows the regression, none of these companies are going to give you a refund based on your vibes.
This is not the same thing as a "omg vibes are off", it's reproducible, I am using the same prompts and files, and getting way worse results than any other model.
It has a habit of trusting documentation over the actual code itself, causing no end of trouble.
Check your claude.md files (both local and ~user ) too, there could be something lurking there.
Or maybe it has horribly regressed, but that hasn't been my experience, certainly not back to Sonnet levels of needing constant babysitting.
If LLMs have a 90% chance of working, there will be some who have only success and some who have only failure.
People are really failing to understand the probabilistic nature of all of this.
"You have a radically different experience with the same model" is perfectly possible with less than hundreds of thousands of interactions, even when you both interact in comparable ways.
Or maybe when usage is high they tweak a setting that use cache when it shouldn't.
For all we know they do whatever experiment the want, to demonstrate theoretical better margin, to analyse user patterns when a performance drop occur.
Given what is done in other industries which don't face an existential issue, it wouldn't surprise me some whistle blowers in a few years tell us what's been going on.
An upcoming IPO increases pressure to make financials look prettier.
In fact as my prompts and documents get better it seems it does increasingly better.
Still, it can't replace a human, I really need to correct it at all, and if I try to one shot a feature I always end up spending more time refactoring it few days later.
Still, it's a huge boost to productivity, but the time it can take over without detailed info and oversight is far away.
My largest gripe with Claude Code, and with encouraging my team to use it, is that checkpoints/rollbacks are still not implemented in the VS Code GUI, leading to a wildly inconsistent experience between terminal and GUI users: https://github.com/anthropics/claude-code/issues/10352
Rollbacks have been broken for me in the terminal for over a month. It just didn’t roll back the code most of the time. I’ve totally stopped using the feature and instead just rely on git. Is this this case for others?
Not discounting at all that you might "hold it" differently and have a different experience. E.g. I basically avoid claude code having any interaction with the VCS at all - and I could easily VCS interaction being a source of bugs with this sort of feature.
It worked when first released but hasn’t for ages now.
I literally just posted in another comment that people shouldn't be worried about killing their current session/context window. I used to get worried about compaction and losing context, but now when I feel like things are slipping I kill it quick and start a new session.
For the last month I've been working on a relatively big feature in a larger project.
I often compact the session when starting a new feature, often have to remind claude to read the claude.md etc. I still use it as if it was a new session regularly, it frequently doesn't remember what it did an hour ago, etc.
But the compact seems to work which is a very different experience than the one of the GP, who kills the session when it reaches the context limit and writes explicit summary files.
However when I try to log in via CLI it takes me to a webpage with an “Authorize” button. Clicking the button does nothing. An error is logged to the console but nothing displays in the UI.
We reached out to support who have not helped.
Not a great first impression
For the claude.ai UI, I've never had a single deep research properly transition (and I've done probably 50 or so) to its finished state. I just know to refresh the page after ~10mins to make the report show up.
I recently put a little money on the API for my personal account. I seem to burn more tokens on my personal account than my day job, in spite of using AI for 4x as long at work, and I’m trying to figure out why.
Just a pro sub - not max.
Most of the time it gives me a heads up that I'm at 90% but a lot of the times it just failed, no warning, and I assumed it was I hit max.
I just signed up as a paying customer, only to find that Claude is totally unusable for my purposes at the moment. There's also no support (shocker), despite their claims that you'll be E-mailed by the support team if you file a report.
What symptoms do you see? There are some command line parameters for reinstall / update that might be worth trying.
Cursor, Claude code, Claude in the browser, and don't even get me started on Gemini.
When I investigated I found the docs and implementation are completely out of sync, but the implementation doesn’t work anyway. Then I went poking on GitHub and found a vibed fix diff that changed the behavior in a totally new direction (it did not update the documentation).
Seems like everyone over there is vibing and no one is rationalizing the whole.
I can’t understand how people would run agents 24/7. The agent is producing mediocre code and is bottlenecked on my review & fixes. I think I’m only marginally faster than I was without LLMs.
And specifically: Lots of checks for impossible error conditions - often then supplying an incorrect "default value" in the case of those error conditions which would result in completely wrong behavior that would be really hard to debug if a future change ever makes those branches actually reachable.
I don’t know where the LLMs are picking up this paranoid tendency to handle every single error case. It’s worth knowing about the error cases, but it requires a lot more knowledge and reasoning about the current state of the program to think about how they should be handled. Not something you can figure out just by looking at a snippet.
At the same time, the amount of anti-patterns the LLM generates is higher than I am able to manage. No Claude.md and Skills.md have not fixed the issue.
Building a production grade system using Claude has been a fools errand for me. Whatever time/energy i save by not writing code - I end up paying back when I read code that I did not write and fixing anti-patterns left and right.
I rationalized by a bit - deflecting by saying this is AI's code not mine. But no - this is my code and it's bad.
This is starting to drive me insane. I was working on a Rust cli that depends on docker and Opus decided to just… keep the cli going with a warning “Docker is not installed” before jumping into a pile of garbage code that looks like it was written by a lobotomized kangaroo because it tries to use an Option<Docker> everywhere instead of making sure its installed and quitting with an error if it isn’t.
What do I even write in a CLAUDE.md file? The behavior is so stupid I don’t even know how to prompt against it.
Think about it, they have to work in a very limited context window. Like, just the immediate file where the change is taking place, essentially. Having broader knowledge of how the application deals with particular errors (catch them here and wrap? Let them bubble up? Catch and log but don't bubble up?) is outside its purview.
I can hear it now, "well just codify those rules in CLAUDE.md." Yeah but there's always edge cases to the edge cases and you're using English, with all the drawbacks that entails.
In particular writing tests that do nothing, writing tests and then skipping them to resolve test failures, and everybody's favorite: writing a test that greps the source code for a string (which is just insane, how did it get this idea?)
The assumption is that your test is right. That's TDD. Then you write your code to conform to the tests. Otherwise what's the point of the tests if you're just trying to rewrite them until they pass?
Claude Code creator literally brags about running 10 agents in parallel 24/7. It doesn't just seems like it, they confirmed like it is the most positive thing ever.
Full disclosure - I am a heavy codex user and I review and understand every line of code. I manually fight spurious tests it tries to add by pointing a similar one already exists and we can get coverage with +1 LOC vs +50. It's exhausting, but personal productivity is still way up.
I think the future is bright because training / fine-tuning taste, dialing down agentic frameworks, introducing adversarial agents, and increasing model context windows all seem attainable and stackable.
I'm definitely faster, but there's a lot of LLM overhead to get things done right. I think if you're just using a single agent/session you're missing out on some of the speed gains.
I think a lot of the gains I get using an LLM is because I can have the multiple different agent sessions work on different projects at the same time.
That is not an uncommon occurrence in human-written code as well :-\
> Automation doesn't just allow you to create/fix things faster. It also allows you to break things faster.
https://news.ycombinator.com/item?id=13775966
Edit: found the original comment from NikolaeVarius
The degradation is palpable.
I have been using vscode github copilot chat with mostly the claude opus 4.5 model. The underlying code for vscode github copilot chat has turned to shit. It will continuously make mistakes no matter what for 20 minutes. This morning I was researching Claude Code and pricing thinking about switching however this post sounds like it has turned to shit also. I don't mind spending $300-$500 a month for a tool that was a month ago accomplishing in a day what would take me 3-4 days to code. However, the days since the last update have been shit.
Clearly the AI companies can't afford to run these models at profit. Do I buy puts?
Then again, the google home page was broken on FF on Android for how long?
Doesn't mean it's not a useful tool - if you read and think about the output you can keep it in check. But the "100% of my contributions to Claude Code were written by Claude Code" claim by the creator makes me doubt this is being done.
Shaping of a codebase is the name of the game - this has always been, and still, is difficult. Build something, add to it, refactor, abstraction doesn’t sit right, refactor, semantics change, refactor, etc, etc.
I’m surprised at how so few seem to get this. Working enterprise code, many codebases 10-20 years old could just as well have been produced by LLMs.
We’ve never been good at paying debt and you kind of need a bit of OCD to keep a code base in check. LLM exacerbates a lack of continuous moulding as iterations can be massive and quick.
Not that old big non-AI software doesn't have similar maintainability issues (I keep posting this example, but I don't actually want to callthat company out specifically, the problem is widespread: https://news.ycombinator.com/item?id=18442941).
That's why I'm reluctant to complain about the AI code issues too much. The problem of how software is written, on the higher level, the teams, the decisions, the rotating programmers, may be bigger than that of any particular technology or person actually writing the code.
I remember a company where I looked at a contractor job, they wanted me to fix a lot of code they had received from their Eastern European programmers. They complained about them a lot in our meeting. However, after hearing them out I was convinced the problem was not the people generating the code, but the ones above them who failed to provide them with accurate specs and clear guidance, and got surprised at the very end that it did not work as expected.
Similar with AI. It may be hard to disentangle what is project management, what is actually the fault of the AI. I found that you can live with pockets of suboptimal but mostly working code well enough, even adding features and fixing bugs easily, if the overall architecture is solid, and components are well isolated.
That is why I don't worry too much about the complaints here about bad error checks and other small stuff. Even if it is bad, you will have lots of such issues in typical large corporate projects, even with competent people. That's because programmers keep changing, management focuses on features over anything else (usually customers, internal or external, don't pay for code reorg, only for new features). The layers above the low level code are more important in deciding if the project is and remains viable.
From what the commenters say, it seems to me the problem starts much higher than the Claude code, so it is hard to say how much at fault AI generated code actually is IMHO. Whether you have inexperienced juniors or an AI producing code, you need solid project lead and architecture layers above the lines of code first of all.
I'd much rather make plans based on reality
Other AI agents, I guess. Call Claude in to clean up code written by Gemini, then ChatGPT to clean up the bugs introduced by Claude, then start the cycle over again.
I've had one llm one-shot a codebase. Then I use another one to review (with a pretty explicit prompt). I take that review and feed it to another agent to refactor. Repeat that a bunch of times.
If the code is cheap (and it certainly is), then tossing it out and replacing it can also be cheap.
Similarly, Human-in-the-loop utilization of AI/ML tooling in software development is expected and in fact encouraged.
Any IP that is monetizable and requires significant transformation will continue to see humans-in-the-loop.
Weak hiring in the tech industry is for other reasons (macro changes, crappy/overpriced "talent", foreign subsidies, demanding remote work).
AI+Competent Developer paid $300k TC > Competent Developer paid $400k TC >>> AI+Average Developer paid $30k TC >> Average Developer paid $40k TC >>>>> Average Developer paid $200k TC
Huh?
A Coding copilot subscription paired with a competent developer dramatically speeds up product and feature delivery, and also significantly upskills less competent developers.
That said, truly competent developers are few and far between, and the fact that developers in (eg.) Durham or remote are demanding a SF circa 2023 base makes the math to offshore more cost effective - even if the delivered quality is subpar (which isn't neccesarily true), it's good enough to release, and can be refactored at a later date.
What differentiates a "competent" developer from an "average" developer is the learning mindset. Plenty of people on HN kvetch about being forced to learn K8s, Golang, Cloud Primitives, Prompt Engineering, etc or not working in a hub, and then bemoan the job market.
If we are paying you IB Associate level salaries with a fraction of the pedigree and vetting needed to get those roles, upskilling is the least you can do.
We aren't paying mid 6 figure TC for a code monkey - at that point we may as well entirely use AI and an associate at Infosys - we are paying for critical and abstract thinking.
As such, AI in the hands of a truly competent engineer is legitimately transformative.
Tl;dr - Mo' money, Mo' expectations
Edit: And 3 minutes later it is back...
I run multiple agents in separate sessions. It starts with one agent, building out features or working on a task/bug fix. Once it gets some progress, I spin up another session and have it just review the code. I explicitly tell it things to look out for. I tell it to let me know about things I'm not thinking of and to make me aware of any blind spots. Whatever it reviews I send back to the agent building out features (I used to also review what the review agent told me about, but now I probably only review it like 20% of the time). I'll also have an agent session started just for writing tests, I tell it to look at the code and see if it's testable, find duplicate code, stale/dead code. And so on and so forth.
Between all of that + deterministic testing it's hard for shit to end up in the code base.
You can assert that something you want to happen is actually happening
How do you assert all the things it shouldn't be doing? They're endless. And AI WILL mess up
It's enough if you're actively reviewing the code in depth.. but if you're vibe coding? Good luck
It's not a world where everything produced is immediately verified.
If a human consistently only produced the quality of work Claude Opus 4.5 is capable of I would expect them to be fired from just about any job in short order. Yes, they'd get some stuff done, but they'd do too much damage to be worth it. Of course humans are much more expensive than LLMs to manage so this doesn't mean it can't be a useful tool... just it's not that useful a tool yet.
This can be abused because the programmer is both judge and jury, but people tend to handle this paradox much better than LLMs.
1. Competent humans architecting and leading the system who understand the specs, business needs, have critical thinking skills and are good at their job
2. Automated tests
3. Competent human reviewers
4. QA
5. Angry users
Cutting out 1 and 3 in favor of more tests isn't gunna work
I'm personally not advocating for not having humans in the loop. I don't know of anybody using llm tools or advocating for them that are saying there shouldn't be humans in the loop. "vibe coding" seems to mean different things to different people.
The amount of times I have to "yell" at the llm for adding #[allow] statements to silence the linter instead of fixing the code is crazy and when I point it out they go "Oops, you caught me, let me fix it the proper way".
So the tests don't necessarily make them produce proper code.
I spent 20 minutes between guiding it because it was putting the translation in the wrong cells, asking it not to convert the cells to a fancy table, and finally, convincing it that it really had access to alter the document, because at some point it denied it. I wasn't being rude, but it seems I somehow made it evasive.
I had to ask it to translate in the chat, and manually copy-pasted the translations in the proper cells myself. Bonus points because it only translated like ten cells at a time, and truncated the reply with a "More cells translated" message.
I can't imagine how hard it would be to handhold an LLM while working in a complex code base. I guess they are a godsend for prototypes and proofs of concept, but they can't beat a competent engineer yet. It's like that joke where a student answers that 2+2=5, and when questioned, he replies that his strength is speed, not accuracy.
Anyways, I replied because I had something else I wanted to say.
I was using Gemini in a google worksheet a while back. I had to cross reference a website and input stuff into a cell. I got Gemini to do it, had it do the first row, then the second, then I told it to do a batch of 10, then 20. It had a hiccup at 20, would take too long I guess. So I had it go back to 10. But then Gemini tells me it can't read my worksheet. I convince it that it can, but then it tells me it can't edit my worksheet. I argue with it, "you've been changing the worksheet wtf?" I convinced it that it could and it started again, but then after doing a couple it told me it couldn't again. We went back and forth a bit, I'd get it working, it would break, repeat. I think it was after the third time I just couldn't get it to do it again.
I looked up the docs, searched online, and I was concerned that I found Google didn't allow Gemini to do a lot of stuff to worksheets/docs/other google workspace stuff. They said they didn't allow it to do a ton of stuff that I definitely had Gemini doing.
Then a week or two went by and google announced they're allowing gemini to directly edit worksheets.
So wtf how did I get it to do it before it could do it???
Manage that yourself! If you have hooks throwing errors then feed the error back into the llm.
So I have a different experience with Claude Code, but I'm not trying to say you're holding it wrong, just adding a data point, and then, maybe I got lucky.
And this is not tied to the LLMs. It's that to EVERYTHING we do. There are limits everywhere.
And for humans the context window might be smaller, but at least we have developed methods of abstracting different context windows, by making libraries.
Now, as a trade-off of trying to go super-fast, changes need to be made in response to your current prompts, and there is no time validate behavior in cases you haven't considered.
And regardless of whether you have abstractions in libraries, or whether you have inlined code everywhere, you're gonna have issues.
With libraries changes in behavior are going to impact code in places you don't want, but also, you don't necessarily know, as you haven't tested all paths.
With inlined code everywhere you're probably going to miss instances, or code goes on to live its own life and you lose track of it.
They built a skyscraper while shifting out foundational pieces. And now a part of the skyscraper is on the foundation of your backyard shed.
folks have created software by "vibe coding". It is now time to "face the music" when doing so for production grade software at scale.
That's a big, slow, and expensive process though.
Will Anthropic actually do that or will they keep throwing AI at it and hope the AI figures this approach out? We shall see...
LLMs do understand codebases and I've been able to get them to make reactors and clean up code without them breaking anything due to them understanding what they are doing.
Bugs are being solved faster than before. Crashes from production can directly be collected and fixed by a LLM with no engineering time needed other than a review.
I like cli tools, and claude is generally considered a very good option for that.
I have a coworker who likes codex better.
Start prompting it for annoying shit "Set up a project layout for X", then write things yourself inside that - the fun stuff or stuff you care about.
Then use it for refactors or extrapolation "I wrote this thing that works, but this old file is still in old format, do what I did there"
It's very good for helping with design of just above layperson knowledge. "I have this problem organizing xyz, what's a good pattern for this?"
or just "I want to do a project that does xyz, but dont know where to start, let's chat about it"
Some of these 'chatty' queries can be done in web, but having it on CLI is great b/c it'll just say "Can I do this for you" and you can easily delegate parts of the plan.
Give it a shot. That's pretty low level agentic use, and yes, it will demolish procrastination and startup inertia.
Right now I'm defaulting to "do nothing" because I'm lazy, but if any Anthropic staff are reading this I'm happy to explain the details informally somewhere.
---
> Just my own observation that the same pattern has occurred at least 3 times now:
> release a model; overhype it; provide max compute; sell it as the new baseline
> this attracts a new wave of users to show exponential growth & bring in the next round of VC funding (they only care about MAU going up, couldn’t care less about existing paying users)
> slowly degrade the model and reduce inference
> when users start complaining, initially ignore them entirely then start gaslighting and make official statements denying any degradation
> then frame it as a tiny minority of users experiencing issues then, when pressure grows, blame it on an “accidentally” misconfigured servers that “unintentionally” reduced quality (which coincidentally happened to save the company tonnes of $).
Just because 99% of the things you read are critical and negatively biased doesn't mean the subsequent determination or the consensus among participants in the public conversation have anything to do with reality.
Dario is delusional, for this and other reasons.
Businesses like google were already a step in the wrong direction in terms of customer service, but the new wave of AI companies seem to have decided their only relation to clients is collecting their money.
Unclear costs, no support, gaslighting customers when a model is degraded, incoming rug pulls..
Sometimes, poor old Claude wants to go on holiday and that is a problem?!?
This is the new status quo for software ... changing and breaking beneath your feet like sand.
CRITICAL: MAKE NO MISTAKES!
CRITICAL: NEVER APOLOGIZE! MAKE IT RIGHT THE FIRST TIME INSTEAD!
CRITICAL: DO NOT HALLUCINATE OR CONFABULATE EVER!
CRITICAL: DON'T DELETE THE DATABASE WITHOUT ASKING FIRST!
CRITICAL: NEVER USE VERBATIM CODE BLOCKS FROM GPL LICENSED PROJECTS!
CRITICAL: CODE AS IF ELON MUSK WAS LOOKING OVER YOUR SHOULDER ALL THE TIME!
CRITICAL: IF YOU MAKE MISTAKES AGAIN I WILL GET PTSD AND DIE AND IT WILL BE YOUR FAULT!
...