Claude: Elevated errors across many models [resolved](status.claude.com) |
Claude: Elevated errors across many models [resolved](status.claude.com) |
Claude Code is especially buggy in windows terminal. The rendering is quite slow, choppy and lines frequently get garbled.
In contrast, using antigravity cli is the exact opposite: fast, smooth and very responsive.
This is an aside, but I'm really struck by how many people on HN use Windows (based on repeated mentions I've seen in comments). I've worked for a pretty wide range of companies over the last decade and only one, maybe two companies even had any people that worked on Windows machines. I haven't worked at a company where devs used Windows in 15 years (and even that company eventually switched to linux).
As I've gotten deeper into LLMs/AI roles even Macs have seemed to start having equal share compared to devs running full Linux setups.
Is this just a sign of that a larger and larger portion of HN users are working for large corporations? I honestly can't even remember that last time I saw a serious developer pull out a Windows laptop.
For a long time, high end graphics and games was mainly done on Windows and Visual Studio. I'm from that world, and only made the Linux transition last year November after Microsoft forced everyone's hand.
Windows really has a fantastic support for C++ and rendering programmers imho, the tooling is world class and Visual Studio has no match as an IDE. Even if somehow my tools worked on Mac or Linux I'd still pick windows out of sheer convenience of using it for work.
But as things stand - all major console toolchains are windows only. If you're making a game for PlayStation, Xbox or Switch, you have to be on windows.
A lot of devs like gaming. Gaming is more simple on windows. Gaming PCs are usually high spec. High spec is good for most coding.
That's why I use windows quite often. My laptop is Linux, but when I'm running heavy models I'll still remote into my main Windows PC, which I also use for gaming.
Though in terms of workplaces - sure, I reckon you're on the money. Big corps often still force windows onto their Devs.
I think that your range of companies was much less wide then you think.
> I honestly can't even remember that last time I saw a serious developer pull out a Windows laptop.
What is unserious developer?
That said, I bet you will be really hard pressed to find a single Windows machine at Anthropic.
The risk for them is that someone matches their products while also having non-janky products and reliable services.
Distributed systems infrastructure, especially, is much less forgiving of vibe coding than application code. Coding agents are not even close to being good enough to design and build large-scale systems the way expert humans can.
There is nothing wrong with using agents to help write infrastructure code, but these systems have a way of punishing anyone who builds things they do not fully understand.
I'd love to see either Anthropic or OpenAI really step up their infrastructure game.
It's worth noting that OpenAI's official uptime numbers are significantly better than Anthropic's:
99.9x% for API/Codex, versus <99.5% for Claude API/Claude Code.
I'd obviously like OpenAI's numbers to be higher too, but this is one reason it really annoys me when the head of Claude Code goes on podcasts implying that software engineering as a whole, not just the act of writing out code, is basically solved.
One wonders why hasn't months of presumably near-unlimited internal Mythic solved the issues unrelated to hardware shortages yet.
Claude Code is sluggish, buggy, slow. Typical big enterprise garbage. The only good thing at Anthropic are the models.
The same post on Claude Code: "Even though the System Prompt and tool descriptions are clearly more verbose, most of the extra tokens encode product features and rational design choices: a memory system, scheduled tasks, sub-agents, plan mode, worktree support. Whether those features are worth paying for depends on your needs. Calling the prompt 'bloated' without looking at the whole picture feels wrong to me."
Our org has been attempting to trial Fast Mode on and off but enabling it in Claude Code just says something like "not available with your cloud provider"
It turns out that for all of the matrices on the documentation site, there is a secret "other" set of infrastructure for those who bill Enterprise via AWS Marketplace where certain features like Fast Mode are incompatible when using "aws routing"
Anthropic Support straight up mentioned that Claude Code just doesn't handle this case correctly and erroneously gives you the impression that it's supported against your billing method, and that it's all effectively undocumented.
I suppose this means there are like 5 different AWS methods of use:
- Bedrock (Legacy)
- Bedrock
- Claude for AWS
- Claude via Marketplace
- Anthropic's own "primary infrastructure"
and then roll in other cloud provider variations
On that note, I recall Datadog having coupled billing and infrastructure per cloud (ie; billing via GCP requires using the GCP infra) and was wondering if commenters had any insights into if there is some special requirement/complexity around marketplace billing for cloud providers or if it's just some weird design choice?
Claude harnesses have plenty of bugs but I prefer capability over interface shininess any day. (though if I were running the show I'd have a sizable team set aside to do exclusively boring stability and polish work)
I prefer it over opencode, which is my other option I use with my Codex sub
Or possibly as a result of.
It takes hours to manhunt these errors and also you cannot risk again to tell the local llm to fix it.
But hey coding is a solved problem.
Good to see it's not just me...
Can you post some images of lines getting garbled. That sounds like a genuine bug Anthropic might want to look into. I haven't seen that ever.
Except, starting this morning, one very long running session decided to start spawning subagents for each task. I'm not sure what caused this emergent behavior, but it seemed to be working fine, so I was eager to see where it went.
Except, as soon as a subagent hits a 500 error, the main agent seemingly doesn't know what to do. It kind of panics—"now the tree/install state is unknown!"—and ultimately does a git checkout "to verify and restore a known-good state before anything else".
I've paused the job for now since it's a sort of background experiment.
Crazy thing is ... its true.
Can't wait for debugging to be solved. Hell, I might even subscribe for 'mostly'.
Yes I know you can run offline models, but it's hard to pass up on a little bit of snark.
Don't jinx it. They might use that name for their next model.
Neat. Didn't have a single request go through for 2 hours. Guess they need to improve their metrics before the IPO...
That sounds like a you issue.. it's wonderful on the terminal. It's their GUI which needs work (they have been improving, but still not a fan).
I've been using it on multiple computers for months and it's generally rock solid and lovely.
That said, I feel icky, like I just made a Facebook account in 2026 :(
I still use Codex, but mostly when I need to check Opus 4.8's work. Pretty sure I will stop doing that soon, because during the short time Fable was available, Codex was not able to find any important issues with the code Fable wrote.
Both were trivial to set up with codex.
Haven’t tried Cowork, interesting. Isn’t it just the same agent minus the git worktree based UI?
Frankly, neither Claude nor Codex are as good as hype entails.
This is all wrong.
Codex ever since ~5.2 has been better at long tasks in large codebases.
Can you move it to background connection?
This implies games are not still mainly done on Windows, which they absolutely are!
But I do prefer a Linux or Mac for development, just because it's so much less hassle to set up.
Windows developers who don't set up a good terminal environment... I honestly don't know how they manage.
Did you even use pi.dev?
I used them both and I can tell you there isn’t a single feature what pi.dev misses from Claude Code.
I see a lot of people speaking about a 10x productivity improvement and so on. When I work on hobby projects, I do see that. Just last weekend, I set up a hobby project that I've been thinking about for a while. I'm pretty sure it would have taken me at least a week to implement manually, but instead, it took me three hours.
But some of the systems I have to work on during the day are so big and complicated that you can spend multiple days on a small feature or even just tracking down a bug, even with the support of Opus.
Expecting even a 2x productivity improvement on some those systems is wildly unrealistic. I'm seeing a lot of people get stressed out because the productivity gains from simple application building trickle into the expectations for these complex systems.
That said, if things keep improving at this rate it might just be a matter of time.
You just have to pay API prices.
Even the HN dataset on HuggingFace shows that most engagement on HN is now during non-US hours [0] and drops off as Europe goes to sleep.
[0] - https://huggingface.co/datasets/open-index/hacker-news
Well, the EU is putting the kibosh on that PDQ. The precious stays at home.
Still requires lots of work to shape the output correctly, but it's still crazy to me that I literally don't write code any more. And in that way, coding is solved for me.
Coding is probably solved, at least to a large extent, but that doesn't mean engineering is solved too.
This is like someone saying that the wright brothers solved sustained/powered human flight, and another person saying "well if that's the case, why do planes still crash? obviously flight isn't solved.". Well, there are always improvements, but planes can fly and llm's can code.
They been claiming more than just “coding is solved” for a while now.
https://www.businessinsider.com/anthropic-claude-code-founde...
Sounds more like a vague prediction of what the future might be like rather than an iron-clad claim.
Personally, Claude Opus (and in the few interactions I had with it, Fable) has been the far the superior experience. GPT-5.5 seems dumber and more certain about presenting me bullshit. Opus has better humor, and is less pretentious in its presentation. But this may all boil down to how the models react to my prompting.
What is without a doubt is that I wish they both were more intelligent – or maybe it is their wisdom I find lacking!