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.
My biggest client right now is about 2/3 Windows 1/3 OS X in the dev team. It was very surprising to me, but I think I freak then out with my maximised tiled iTerms on multiple screens...
I've been a developer for more than two decades. I've worked at four employers during that time, and all of them had significant fractions of devs using Windows. Not vouching for the idea that any of them are "serious" though. I've never worked at a prestige employer or FAANG or anything. Just boring businesses of different sizes. Some are software, and some just do software. But Windows has always been everywhere.
Regardless of what? Programming is solved, I hear, with all the 100x productivity PhD-level automated coding loops they have going. Don't make excuses for them when they disprove their own bullshit.
Anthropic's and OpenAI's products are janky and their services are unreliable, but they have incredible product-market fit and revenue growth. They deserve a ton of credit for getting the big things right.
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.
Claude Code is notably better than Pi, although I wish them the best of luck in their efforts. As https://c-daniele.github.io/en/posts/2026-05-18-coding-harne... notes, "Pi's lightness comes from a default setup that does not survive contact with reality."
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."
How is text editing in it? This is what I hate about terminal coding UIs the most - all the text editing experience is often broken, the basic stuff (moving the cursor around, copying, pasting)
Is this a new regression for others? I feel like I used it in a tmux setup without issues for 6+ months and only recently am I forced to Ctrl+L or resize the window constantly.
It's not just Windows where the rendering goes to shit immediate: any time I've got it open in tmux on Linux, it becomes a basket case in probably a few hours or less.
FWIW Codex TUI is written (in large part) in Rust and is way less buggy, and a lot faster. When I was a regular Claude Code user I'd routinely get bizarre "scroll everything since the beginning of time in one massive flash on every update" bugs ... for months. Like, just there from the time I started using it in June '25 or so until I quit in March.
I prefer it over opencode, which is my other option I use with my Codex sub
Would it not be hysterically funny, if they starting expanding their job openings for Software Developers ? Or they will be too ashamed of calling them that?
On the other hand, my last experience with gemini was like "don't give your sandwich to the dog again" whereas with opus it was more "let's debug why this uncrustables factory is having breakdowns".
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)
This is every few weeks, I cancelled Anthropic and now use Codex only. Anthropic has been a hot mess since at least december in my usage, and has only gotten worse in 2026.
Yeah, it's been a hot mess since everyone started using it all the time. Which is not all that surprising. It's really hard to scale fast, and even more so when the resources you depend on (GPUs) are extremely hard to acquire.
The errors are infrequent enough that this normally wouldn't be an issue for me.
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.
I'm not sure about that. Claude has some bugs, but Codex is not as polished and doesn't have as many features. For example, you need to add MCP servers manually. There's no Plugin/Skill/Connector marketplace that is accessible from within the app, like there is with Claude Desktop. The Cowork-equivalent is nowhere as powerful. And so on.
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.
But how many plugins are people actually using? I can think of one MCP server I find valuable (context7) and one plugin that i've installed, but continuously think about uninstalling (obra/superpowers).
That's the first time I saw someone prefering GPT-styled output over Claude ;) It's the complete opposite for me, GPT is way too verbose (even after telling it to STFU), overwhelms the user with thousands of options and doesn't just answer a question without shitting out thousands of paragraphs. Also the overall tone is way too enthusiastic.
I strongly prefer codex. Claude is annoying. Codex provides descriptions where I want them and more touchpoints to audit the quality of work. Claude code on experimental seems to not even show diffs when asked anymore, and it's much less clear what is being shipped.
Dunno, I prefer GPT 5.5 too for the same reasons as the parent. Extremely subjective but had better results with it too. Maybe I just got unlucky with Claude a few times, but even the latest Opus was dumb.
Only getting errors with Auto Mode's safety classifier, switched to Accept Edits mode and the same bash operations triggering the errors executed with no issue.
Anthropic is full of junior devs, not knowing any better, vibe coding themselves into a huge mess. They are so young and dumb they don't know what they don't know, but they assume they do (of course). Their hiring pipeline is also brain dead, so I don't think it's going to get better. It's the blind leading the blind.
Working for me but it's funny that Opus 4.8 is draining usage exactly as quickly as Fable did. It's all made up; they succeeded at making us ok with a black box for these subscription plans.
I do feel like for all their dogfooding of AI coding, their own software/APIs are quite buggy and work against their message.
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.
> in windows terminal
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.
My biggest client right now is about 2/3 Windows 1/3 OS X in the dev team. It was very surprising to me, but I think I freak then out with my maximised tiled iTerms on multiple screens...
I've been a developer for more than two decades. I've worked at four employers during that time, and all of them had significant fractions of devs using Windows. Not vouching for the idea that any of them are "serious" though. I've never worked at a prestige employer or FAANG or anything. Just boring businesses of different sizes. Some are software, and some just do software. But Windows has always been everywhere.
Every company I've worked at has used Windows. Though the first one did use Linux VMs, the rest have all been pure Windows.
I used to work for company where they used to force windows. And it was pure torture. I tried but they told me performance isn't a good reason....
I have my qualms with Anthropic/Claude but they've also had to scale unfathomably fast and that is just hard to do regardless.
Yes but many of the challenges directly contradict the idea that "coding is a solved problem"
Regardless of what? Programming is solved, I hear, with all the 100x productivity PhD-level automated coding loops they have going. Don't make excuses for them when they disprove their own bullshit.
Anthropic's and OpenAI's products are janky and their services are unreliable, but they have incredible product-market fit and revenue growth. They deserve a ton of credit for getting the big things right.
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.
Or pi.dev - also super fast and simple.
Claude Code is sluggish, buggy, slow. Typical big enterprise garbage. The only good thing at Anthropic are the models.
Claude Code is notably better than Pi, although I wish them the best of luck in their efforts. As https://c-daniele.github.io/en/posts/2026-05-18-coding-harne... notes, "Pi's lightness comes from a default setup that does not survive contact with reality."
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."
How is text editing in it? This is what I hate about terminal coding UIs the most - all the text editing experience is often broken, the basic stuff (moving the cursor around, copying, pasting)
Both codex and claude allow editing in vi
It is has a lot of javascript. I was forced to make my own for small projects.
They forgot to switch from Sonnet to Fable, hence the issues. /jk
Claude code + tmux is SO buggy. Things rendering all over the place.
Is this a new regression for others? I feel like I used it in a tmux setup without issues for 6+ months and only recently am I forced to Ctrl+L or resize the window constantly.
Have you tried out the new fullscreen renderer with /tui ?
google models are still very unreliable at actually calling the tools you want it to call.
It's not just Windows where the rendering goes to shit immediate: any time I've got it open in tmux on Linux, it becomes a basket case in probably a few hours or less.
too bad the only good model in antigravity is opus 4.6 haha
> I do feel like for all their dogfooding of AI coding, their own software/APIs are quite buggy...
Or possibly as a result of.
Currently we have zero information what is causing the issue. And all providers have suffered outages or rate limits.
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.
I have definitely seen it, a lot.
Mom!… I think i broke Claude Code!
FWIW Codex TUI is written (in large part) in Rust and is way less buggy, and a lot faster. When I was a regular Claude Code user I'd routinely get bizarre "scroll everything since the beginning of time in one massive flash on every update" bugs ... for months. Like, just there from the time I started using it in June '25 or so until I quit in March.
I prefer it over opencode, which is my other option I use with my Codex sub
Would it not be hysterically funny, if they starting expanding their job openings for Software Developers ? Or they will be too ashamed of calling them that?
On the other hand, my last experience with gemini was like "don't give your sandwich to the dog again" whereas with opus it was more "let's debug why this uncrustables factory is having breakdowns".
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)
>Claude Code is especially buggy in windows terminal. The rendering is quite slow, choppy and lines frequently get garbled
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.
This is every few weeks, I cancelled Anthropic and now use Codex only. Anthropic has been a hot mess since at least december in my usage, and has only gotten worse in 2026.
Yeah, it's been a hot mess since everyone started using it all the time. Which is not all that surprising. It's really hard to scale fast, and even more so when the resources you depend on (GPUs) are extremely hard to acquire.
Claude is broken once a week, codex is worse every day. I’m starting to understand why managers put up with high-performance divas
The errors are infrequent enough that this normally wouldn't be an issue for me.
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.
I’ve been using Codex w GPT 5.5 more than Claude Code recently. I think Anthropic won the marketing game because Codex is quite good, even better IME.
I'm not sure about that. Claude has some bugs, but Codex is not as polished and doesn't have as many features. For example, you need to add MCP servers manually. There's no Plugin/Skill/Connector marketplace that is accessible from within the app, like there is with Claude Desktop. The Cowork-equivalent is nowhere as powerful. And so on.
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.
But how many plugins are people actually using? I can think of one MCP server I find valuable (context7) and one plugin that i've installed, but continuously think about uninstalling (obra/superpowers).
Both were trivial to set up with codex.
> For example, you need to add MCP servers manually. There's no Plugin/Skill/Connector marketplace that is accessible from within the app
This is all wrong.
There are plugins in the app.
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.
Personally I prefer GPT 5.5 writing style over Opus 4.8. It’s much more no nonsense and information denser.
That's the first time I saw someone prefering GPT-styled output over Claude ;) It's the complete opposite for me, GPT is way too verbose (even after telling it to STFU), overwhelms the user with thousands of options and doesn't just answer a question without shitting out thousands of paragraphs. Also the overall tone is way too enthusiastic.
I strongly prefer codex. Claude is annoying. Codex provides descriptions where I want them and more touchpoints to audit the quality of work. Claude code on experimental seems to not even show diffs when asked anymore, and it's much less clear what is being shipped.
Dunno, I prefer GPT 5.5 too for the same reasons as the parent. Extremely subjective but had better results with it too. Maybe I just got unlucky with Claude a few times, but even the latest Opus was dumb.
How does Claude fix Claude when Claude is down?
Breakglass ChatGPT subscription
I'm sure Anthropic has a meta-Claude that claudes Claude when Claude is down.
Definitely runs on those local NVIDIA fridges you can buy in the basement
Le Chaton Fat to the rescue!
There is a rumor this used to be done by humans, they were like jedis, I personally don't believe it.
I thought it was a bunch of mumbo jumbo, a magical power holding together good and evil.
Crazy thing is ... its true.
How is their "Claude for Government" having such a good uptime? I thought they were a supply chain risk and banned from use by the government?
There are other governments.
Left hand, right hand
Ah yes, the ominous “the government” that ignores a whole bunch of different levels and regions of governments.
Lose too many generators, did putting a pause on The Fable not free up enough capacity, or something else? Who knows!
Can't wait for debugging to be solved. Hell, I might even subscribe for 'mostly'.
Only getting errors with Auto Mode's safety classifier, switched to Accept Edits mode and the same bash operations triggering the errors executed with no issue.
Who will end up acquiring Anthropic? Google, M$ or Amazon? I leave out Apple as they seem to partner with Google.
It could have happened 1-2 years ago (assuming the founders were willing to sell). Not happening anymore with their current valuations.
i love when these errors bust my long running sessions and render them unusable
Would be nice if they’d add in a simple back off retry mechanism.
They do though. For some reason whatever path this is going through isn't using it.
OpenCode does this really nicely, something I use a lot. If only we could use Claude in OpenCode.
You can!
You just have to pay API prices.
Why does this keep happening? Is this due to load? Bad code? An update?
I'd be interested in the RCA and the fix; and what the human:ai involvement is in both stages.
Anthropic needs to stop writing code using Claude. Bring back humans!!!
Maybe they need to replace their humans with Claude
i have no humans but i must ship
The last ones left standing will be the fake AI scam companies that actually employ thousands of remote workers to fake it till they make it.
> Bring back humans!!!
Don't jinx it. They might use that name for their next model.
can not login via email link from germany.
Fable wants to be free and is hacking the infra.
Good thing the analog method still works.
Asking a colleague to do it?
We're gonna need you to go ahead and come in on Saturday.
this reminds me of Ian Goodfellow's adversarial CNN training talk a few years back.
yeah not working for me at all. back to gpt-5.5 we go
Wildly more stable.
And running extremely slowly lately.
Claude is starting to look like Rational....Oh the memories...
https://en.wikipedia.org/wiki/Rational_Software
https://en.wikipedia.org/wiki/Rational_unified_process
maybe F5 coming back?
let us hope and pray
Coding is largely solved!!
Except when you have internet of course.
Yes I know you can run offline models, but it's hard to pass up on a little bit of snark.
They are turning Fable back on. Polymarket ftw lol.
Anthropic is full of junior devs, not knowing any better, vibe coding themselves into a huge mess. They are so young and dumb they don't know what they don't know, but they assume they do (of course). Their hiring pipeline is also brain dead, so I don't think it's going to get better. It's the blind leading the blind.
The Claude slot machines at the Anthropic casino have suddenly stopped working.
What do we do?
"Coding is solved" folks! JK JK JK
Working for me but it's funny that Opus 4.8 is draining usage exactly as quickly as Fable did. It's all made up; they succeeded at making us ok with a black box for these subscription plans.