The irony of this, is that Microsoft was trying to push CoPilot everywhere, however eventually Apple, Google and JetBrains have their own AI integrations, taking CoPilot out of the loop.
Slowly the AI craziness at Microsoft is taking the similar shape, of going all in at the begining and then losing to the competition, that they also had with Web (IE), mobile (Windows CE/Pocket PC/WP 7/WP 8/UWP), the BUILD sessions that used to be all about UWP with the same vigour as they are all AI nowadays, and then puff, competition took over even if they started later, because Microsoft messed up delivery among everyone trying to meet their KPIs and OKRs.
I also love the C++ security improvements on this release.
About GitHub Copilot in specific: One big negative was how when GPT-4 became available that Microsoft didn't upgrade paying Copilot users to it, they simply branded this "coming soon"/"beta" Copilot X for a while. We simply cancelled the only Copilot subscription we had at work.
I've been getting monthly emails that my free access for GitHub Copilot has been renewed for another month… for years. I've never used it, I thought that all GitHub users got it for free.
Just because you can’t or won’t win the market with your opportunistic investment, doesn’t mean you should let your competitors completely annihilate you by taking that investment for themselves.
Google, Apple, FB or AWS would have been suitors for that licensing deal if MS didn’t bite.
CoPilot isn't anything Microsoft is trying to sell outside of their own products. And with GitHub Copilot there is no "copilot" model to choose, you can choose between Anthropic, OpenAI and Google models.
Sure UWP never caught on, but you know why? Win32, which by the way is also Microsoft, was way to popular and more flexible. Devs weren't going to re-write their apps to UWP in order to support phones.
Microsoft owns 49% of OpenAI so why they should worry? JetBrains just proudly announce that they now use GPT-5 by default.
> going all in at the begining and then losing to the competition
Sure, but there are counter examples too. Microsoft went late to the party of cloud computing. Today Azure is their main money printing machine. At some point Visual Studio seemed to be a legacy app only used for Windows-specific app development. Then they released VSCode and boom! It became the most popular editor by a huge margin[0].
Microsoft mistook a product game for a distribution one. AI quality is heterogenous and advancing enough that people will make an effort to use the one they like best. And while CoPilot is excellently distributed, it’s a crap product, in large part due to the limits Microsoft put on GPT.
Maybe because Microsoft is a shit company and anything they do is sus af. And everyone knows it. And I'm tired of pretending like it's not. I wouldn't trust Microsoft to babysit my mortal enemy's kids.
Maybe if they weren't literally the borg people would open their hearts and wallets to Redmond. They saw that Windows 10 was a privacy nightmare and what did they do? They doubled down in Windows 11. Not that I care but it plays really poorly. Every nerd on the internet spouts off about Recall even though it's not even enabled if you install straight to the latest build.
They bought GitHub and now it's a honeypot. We live in a world where we have to assume GitHub is adversarial.
_NSAKEY???
Fuck you Microsoft.
Makes sense karma catches up to them. Maybe if their mission statement and vision were pure or at least convincing they would win hearts and minds.
Also OpenAI pioneered but now the many competitors seem to have either caught up or surpassed them. They might still retain a significant brand recognition advantage as long as they don't fall too far behind, though.
umm I don't know what you are talking about, I use a Github Copilot 40 USD subscription in VSCode to code using various models, and this is the industry standard now in my region, as most employers are now giving employees the 10 USD subscription.
Almost no one uses copilot unless they are not allowed to use anything else or don’t know any better. MS could have been a leader in this space but MS couldn’t understand why people didn’t like copilot but loved the competition.
3 days ago I saw another Claude praising submission on HN, and finally I signed up for it, to compare it with copilot.
I asked 2 things.
1. Create a boilerplate Zephyr project skeleton, for Pi Pico with st7789 spi display drivers configured. It generated garbage devicetree which didn't even compile. When I pointed it out, it apologized and generated another one that didn't compile. It configured also non-existent drivers, and for some reason it enabled monkey test support (but not test support).
2. I asked it to create 7x10 monochromatic pixelmaps, as C integer arrays, for numeric characters, 0-9. I also gave an example. It generated them, but number eight looked like zero. (There was no cross in ether 0 nor 8, so it wasn't that. Both were just a ring)
What am I doing wrong? Or is this really the state of the art?
It’s good at doing stuff like “host this all in Docker. Make a Postgres database with a Users table. Make a FastAPI CRUD endpoint for Users. Make a React site with a homepage, login page, and user dashboard”.
It’ll successfully produce _something_ like that, because there’s millions of examples of those technologies online. If you do anything remotely niche, you need to hold its hand far more.
The more complicated your requirements are, the closer you are to having “spicy autocomplete”. If you’re just making a crud react app, you can talk in high level natural language.
Did you try claude code and spend actual time going back and forth with it, reviewing it's code and providing suggestions; Instead of just expecting things to work first try with minimal requirements?
I see claude code as pair programming with a junior/mid dev that knows all fields of computer engineering. I still need to nudge it here and there, it will still make noob mistakes that I need to correct and I let it know how to properly do things when it gets them wrong. But coding sessions have been great and productive.
In the end, I use it when working with software that I barely know. Once I'm up and running, I rarely use it.
> Did you try claude code and spend actual time going back and forth with it, reviewing it's code and providing suggestions; Instead of just expecting things to work first try with minimal requirements?
I did, but I always approached LLM for coding this way and I have never been let down. You need to be as specific as possible, be a part of the whole process. I have no issues with it.
I agree, but I think there's an important distinction to be made.
In some cases, it just doesn't have the necessary information because the problem is too niche.
In other cases, it does have all the necessary information but fails to connect the dots, i.e. reasoning fails.
It is the latter issue that is affecting all LLMs to such a degree that I'm really becoming very sceptical of the current generation of LLMs for tasks that require reasoning.
They are still incredibly useful of course, but those reasoning claims are just false. There are no reasoning models.
FWIW, I used Gemini to write an Objective-C app for Apple Rhapsody (!) that would enumerate drivers currently loaded by the operating systems (more or less save level of difficulty as the OP, I'd say?), using the PDF manual of NextStep's DriverKit as context.
It... sort of worked well? I had to have a few back-and-forth because it tried to use Objective-C features that did not exist back then (e.g. ARC), but all in all it was a success.
So yeah, niche things are harder, but on the other hand I didn't have to read 300 pages of stuff just to do this...
I remember writing obj-c naturally by hand. Before swift was even a twinkle in tim cooks eye. One of my favorite languages to program in I had a lot of fun writing ios apps back in the day it seems like
In other words, the vibe coders of this world are just redundant noobs who don't really belong on the marketplace. They've written the same bullshit CRUD app every month for the past couple of years and now they've turned to AI to speed things up
Last week I asked Claude to improve a piece of code that downloads all AWS RDS certificates to just the ones needed for that AWS region. It figured out several ways to determine the correct region, made a nice tradeoff and suggested the most reliable way. It rewrote the logic to download the right set, did some research to figure out the right endpoint in between. It only made one mistake, it fallback mechanism was picking EU, which was not correct. Maybe 1 hour of work. On my own it would have taken me close to a working day to figure it all out.
I don't mean to be treading on feet but I'm noticing this more and more in the debates around AI. Imagine if there are developers out there that could have done this task in 30 mins without AI.
The level of performanace of AI solutions is heavily related to the experience level of the developer and of the problem space being tackled - as this thread points out.
Unfortunately the marketing around AI ignores this and makes every developer not using AI for coding seem like a dinosauer, even though they might well be faster in solving their particular problems.
AI is moving problem solving skills from coding to writing the correct prompts and teaching AI to do the right thing - which, again, is subjective, since the "right thing" for one developer isn't the "right thing" for the another developer. "Right thing" being the correct solution, the understandable solution, the fastest solution, etc depending on the needs of the developer using the AI.
I think the majority of coders out there write the same CRUD app over and over again in different flavors. That's what the majority of businesses seem to pay for.
If a business needs the equivalent of a Toyota Corolla, why be upset about the factory workers making the millionth Toyota Corolla?
> I think the majority of coders out there write the same CRUD app over and over again in different flavors
In my experience, that's not entirely true. Sure, a lot of app are CRUD apps, but they are not the same. The spice lies in the business logic, not in programming the CRUD operations. And then of course, scaling, performance, security, organization, etc etc.
Your first prompt is testing Claude as an encyclopedia: has it somehow baked into its model weights the exactly correct skeleton for a "Zephyr project skeleton, for Pi Pico with st7789 spi display drivers configured"?
Frequent LLM users will not be surprised to see it fail that.
The way to solve this particular problem is to make a correct example available to it. Don't expect it to just know extremely specific facts like that - instead, treat it as a tool that can act on facts presented to it.
For your second example: treat interactions with LLMs as an ongoing conversation, don't expect them to give you exactly what you want first time. Here the thing to do next is a follow-up prompt where you say "number eight looked like zero, fix that".
Trying two things and giving up. It's like opening a REPL for a new language, typing some common commands you're familiar with, getting some syntax errors, then giving up.
You need how to learn to use your tools to get the best out of them!
Start by thinking about what you'd need to tell a new Junior human dev you'd never met before about the task if you could only send a single email to spec it out. There are shortcuts, but that's a good starting place.
In this case, I'd specifically suggest:
1. Write a CLAUDE.md listing the toolchains you want to work with, giving context for your projects, and listing the specific build, test etc. commands you work with on your system (including any helpful scripts/aliases you use). Start simple; you can have claude add to it as you find new things that you need to tell it or that it spends time working out (so that you don't need to do that every time).
2. In your initial command, include a pointer to an example project using similar tech in a directory that claude can read
3. Ask it to come up with a plan and ask for your approval before starting
I guess many find comfort in being able to task an ai with assignments that it cannot complete. Most sr developers I work with take this approach. It's not really a good way of assessing the usefulness of a tool though.
too big of tasks. break them down and then proceed from there. have it build out task lists in a TASKS.md. review those tasks. do you agree? no? work with it to refine. implement one by one. have it add the tests. refactor after awhile as {{model}} doesn't like to do utility functions a lot. right now, about +50k lines in to a project that's vibecoded. i sit back and direct and it plays.
Imagine the CS 100 class where they ask you to make a PB&J. saying for it to make it, there's a lot of steps, but determine known the steps. implement each step. progress.
I'm inclined to agree with this approach because someone not using AI who fails would likely fail for the same reasons. If you can't logically distill a problem into parts you can't obtain a solution.
What an odd thing to ask it. I installed claude code and ran it from my terminal. Just asked it to simply give me a node based rest API with X endpoints with these jobs, and then I told it to write the unreal engine c++ to consume those endpoints. 2500 lines of code later, it worked.
The only way I manage to get any benefits from LLMs is to use them as an interactive rubber duck.
Dump your thoughts in a somewhat arranged manner, tell it about your plan, the current status, the end goal, &c. After that tell it to write 0 code for now but to ask questions and find gaps in your plan. 30% of it will be bullshit but the rest is somewhat useable. Then you can ask for some code but if you care about quality or consistency with you existing code base you probably will have to rewrite half of it, and that's if the code works in the first place
Garbage in garbage out is true for training but it's also true for interactions
> It configured also non-existent drivers, and for some reason it enabled monkey test support (but not test support).
If it doesn't have the underlying base data, it tends to hallucinates. (It's getting a bit difficult to tell when it has underlying data, because some models autonomously search the web). The models are good at transforming data however, so give it access to whatever data it needs.
Also let it work in a feedback loop: tell it to compile and fix the compile errors. You have to monitor it because it will sometimes just silence warnings and use invalid casts.
> What am I doing wrong? Or is this really the state of the art?
It may sound silly, but it's simply not good at 2D
> It may sound silly, but it's simply not good at can2D.
It's not silly at all, it's not very good at layouts either, it can generally make layouts but there is a high chance for subtle errors, element overlaps, text overflows, etc.
Mostly because it's a language model, i.e it doesn't generally see what it makes, you can send screenshots apparently and it will use it's embedded vision model, but I have not tried that.
> What am I doing wrong? Or is this really the state of the art?
You're treating the tool like it was an oracle. The correct way is to treat it as a somewhat autistic junior dev: give it examples and process to follow, tell it to search the web, read the docs, how to execute tests. Especially important is either directly linking or just copy pasting any and all relevant documentation.
The tool has a lossily compressed knowledge database of the public internet and lots of books. You want to fix the relevant lossy parts in the context. The less popular something is, the more context will be needed to fill the gaps.
Try this prompt: Create a detailed step by step plan to implement a boilerplate Zephyr project skeleton for Pi Pico with configured st7789 SPI display drivers
Ask Opus or Gemini 2.5 Pro to write a plan. Then ask the other to critique it and fix mistakes. Then ask Sonnet to implement
I tried this myself and IMO, this might be basic and day-to-day for you, with unambiguous correct paths to follow, but this is pretty niche nevertheless. LLMs thrive when there's a wealth of examples and I struggle to Google what you asked myself, meaning that LLM will perform even worse than my try.
I found that second line works well for image prompts too. Tell one AI to help you with a prompt, and then take it back to the others to generate images.
Sounds like you picked some obscure tasks to test it that would obviously have low representation in the data set? That is not to say it can't be helpful augmenting some lower represented frameworks/tools - just you'll need to equip it with better context (MCPs/Docs/Instruction files)
A key skill in using an LLM agentic tool is being discerning in which tasks to delegate to it and which to take on yourself. Try develop that skill and maybe you will have better luck.
There's a lot of people caricaturing the obvious fact that any model works best in distribution.
The more esoteric your stack, and the more complex the request, the more information it needs to have. The information can be given either through doing research separately (personally, I haven't had good results when asking Claude itself to do research, but I did have success using the web chat UI to create an implementation plan), or being more specific with your prompt.
As an aside, I have more than 10 years of experience, mostly with backend Python, and I'd have no idea what your prompts mean. I could probably figure it out after some google searches, tho. That's also true of Claude.
Here's an example of a prompt that I used recently when working on a new codebase. The code is not great, the math involved is non trivial (it's research-level code that's been productionized in hurry). This literally saved 4 hours of extremely boring work, digging through the code to find various hardcoded filenames, downloading them, scp'ing them, and using them to do what I want. It one-shotted it.
> The X pipeline is defined in @airflow/dags/x.py, and Y in `airflow/dags/y.py` and the relevant task is `compute_X`, and `compute_Y`, respectively. Your task is to:
> 1. Analyze the X and Y DAGs and and how `compute_X` functions are called in that particular context, including it's arguments. If we're missing any files (we're probably missing at least one), generate a .sh file with aws cli or curl commands necessary for downloading any missing data (I don't have access to S3 from this machine, but I do have in a remote host). Use, say, `~/home` as the remote target folder.
> 2. If we needed to download anything from S3, i.e. from the remote host, output rsync/scp commands I can use to copy them to my local folder, keeping the correct/expected directory structure. Note that direct inputs reside under `data/input`, while auxiliary data resides in other folders under `data`. Do not run them, simply output them. You can use for example `scp user@server.org ...`
> 3. Write another snapshot test for X under `tests/snapshot`, and one for Y. Use a pattern as similar as possible to the other tests there. Do not attempt to run the tests yet, since I'll need to download the data first.
> If you need any information from Airflow, such as logs or output values, just ask and I can provide them. Think hard.
Real vibe coding is fake, especially for something niche like what you asked it to do. Imagine a hyperactive eidetic fresh out of high school was literally sitting in the other room. What would you tell her? That’s a good rule of thumb for the level of detail and guidance
Ok. several tips I can give.
1. Setup a sub-agent to do RESEARCH. It is important that it only has read-only and web access tools.
2. Use planning mode and also ask the agent to use the subagent to research best pratices with the tech that you are wanting to do, before it builds a plan.
3. When ever it gets hung up.. tell it to use the sub-agent to research the solution.
That will get you a lot better initial solution. I typically use Sonnet for the sub-agents and Opus for the main agent, but sonnet all around should be fine too for the most part.
You can no longer answer "what is the state of the art” by pointing to a model.
Generating a state-of-the-art response to your request involves a back-and-forth with the agent about your requirements, having a agent generate and carry out a deep research plan to collect documentation, then having the agent generate and carry out a development plan to carry it out.
So while Claude is not the best model in terms of raw IQ, the reason why it's considered the best coding model is because of its ability to execute all these steps in one go which, in aggregate, generates a much better result (and is less likely to lose its mind).
In my experience Claude is quite good at the popular stacks in the JavaScript, Python and PHP world. It struggled quite a bit when I asked it non-trivial questions in C or Rust for example. For smaller languages (e.g., Crystal) it seems to hallucinate a lot. I think since a lot of people work in JS, Python and PHP, that’s where Claude is very valuable and that’s where a lot of the praise feel justified too.
I have had no problems with using Claude on large rust projects. The compiler errors usually point it towards fixing its mistakes (just like they do for me).
I've had similar experiences when working on non-web tech.
There are parts in the codebase I'd love some help such as overly complex C++ templates and it almost never works out. Sometimes I get useful pointers (no pun intended) what the problem actually is but even that seems a bit random. I wonder if it's actually faster or slower than traditional reading & thinking myself.
One of the things you can do is provide a guidance file like CLAUDE.md including not only style preferences but also domain knowledge so it has greater context and knows where to look. Just ask it make one and then update and change as needed.
The thing you are doing wrong is asking it to solve hard problems. Claude Code excels at solving fairly easy, but tedious stuff. Refactors that are brainless but take an hour. It will knock those out of the park. Fire up a git worktree and let it spin on your tedious API changes and stuff while you do the hard stuff. Unfortunately, you'll still need to use your brain for that.
Tbh dawg, those tasks sound intentionally obtuse. It looks like u are doing more esoteric work than the crud react slop us mortals play in on the daily which is where ai shines.
I work almost exclusively with embedded devices, with low level code (mostly C, Rust, Assembly and related frameworks) - and that's where I also ask for help from LLMs.
My coding ranges from "exotic" to "boiler plate" on any given day.
> Create a boilerplate Zephyr project skeleton, for Pi Pico
Yea... Asking Claude to help you with a low documentation build root system is going to go about the same way, I know first hand about how this works.
> I asked it to create 7x10 monochromatic pixelmaps
Wrong tool for the job here. I dont think IDE and Pixelmaps have as large of an intersection as you think they do. Claude thinks in tokens not pixels.
Pick a common language (js, python, rust, golang) pick something easy (web page, command line script, data ingestion) and start there. See what it can do and does well, then start pushing into harder things.
What you're doing wrong is that you're asking it for something more complicated than babby's first webapp in javascript/python.
When people say things like "I told Claude what I wanted and it did it all on the first try!", that's what they mean. Basic web stuff that that is already present in the model's training data in massive volumes, so it has no issue recreating it.
No matter how much AI fanatics try to convince you otherwise, LLMs are not actually capable of software engineering and never will be. They are largely incapable of performing novel tasks that are not already well represented in their weights, like the ones you tried.
What they are not capable of is replacing YOU, the human who is supposed to be part of the whole process (incl. architectural). I do not think that this is a limitation. In fact, I like being part of the process.
So I've used Zephyr. The thing you're doing wrong is expecting LLMs to scaffold you a bunch of files from a relatively niche domain. Zephyr is also a mess of complexity with poor documentation. You should ask it to consult official docs and ask it to use existing tools (west etc) and board defs to do the scaffolding.
I just had AI write me a scraper and download 5TB of invaluable data which I had been eyeing for a long time. All in ten days. At the end of it, I still don’t know anything about python. It’s a bliss for people like me. All dependencies installed themselves. I look forward to using it even more.
One frustration was the code changed so much in ChatGPT so had to be lots of prompts. But I had no idea what the code was anyways. Understood vibe coding. Just used ChatGPT on a whim. Liked the end result.
Interesting to think about how Apple get to make product decisions based on Gatekeeper OCSP analytics now that every app launch phones home. They must know exactly how popular VSCode is.
Facebook got excoriated for doing that with Onavo but I guess it's Good Actually when it's done in the name of protecting my computer from myself lol
> At Apple's World Wide Developer Conference on Monday, Tim Cook mentioned that there are now 34 million registered developers with the company's platform.
I think that means either:
* they have revenues of $3.4b/year just from the $100 annual fees, or
* some decent percentage of people have signed up for a free developer account and then never done anything with it (like me)
If my experience is anything to go by - a good proportion of this will be people accidentally double clicking a .md (or other random text suffix), and cursing whilst they wait for XCode to slowly load enough that they can quit it and open the file in a proper lightweight editor..
I feel like the #1 reason to install Xcode is to get Git working on macOS. Yours is probably #2. I wouldn't bet money on iOS/macOS development sitting at #3.
Compared to stock Claude Code, this version of Claude knows a lot more about SwiftUI and related technologies. The following is output from Claude in Xcode on an empty project. Claude Code gives a generic response when it looked at the same project:
What I Can Help You With
• SwiftUI Development: Layout, state management, animations, etc.
• iOS/macOS App Architecture: MVVM, data flow, navigation
• Apple Frameworks: Core Data, CloudKit, MapKit, etc.
• Testing: Both traditional XCTest and the new Swift Testing framework
• Performance & Best Practices: Swift concurrency, memory management
Example of What We Could Do Right Now
Looking at your current ContentView.swift, I could help you:
• Transform this basic "Hello World" into a recovery tracking interface
• Add navigation, data models, or user interface components
• Implement proper architecture patterns for your Recovery Tracker app
If a bunch of markdown files forced into the context is “knowing”, then yes. They are usually located at /Applications/Xcode-beta.app/Contents/PlugIns/IDEIntelligenceChat.framework/Versions/A/Resources/AdditionalDocumentation
You are free to point Claude Code to that folder, or make a slash command that loads their contents. Or, start CC with -p where the prompt is the content of all those files.
Claude Code integration in Xcode would be very cool indeed, but I might still stick with VSCode for pure coding.
> Claude Code integration in Xcode would be very cool indeed, but I might still stick with VSCode for pure coding.
I'm sticking with VSCode too, but it's a bit silly to suggest that anyone is using XCode because it's their preferred IDE. It's just the one that's necessary for any non-trivial Apple platform development.
Adding a code generator isn't a marketing ploy to get people to switch editors, it's just a small concession to the many hapless souls stuck dealing with Apple on the professional side, or masochistically building mac SwiftUI apps just to remind themselves what pain feels like.
I mean you can stay in VSCode for most activities if you hate Xcode that much (I can relate btw). Plugins like Sweetpad make this possible. My approach now is to develop all logic in small Swift packages and run swift test in VSCode (or Claude Code), so I only absolutely need Xcode for debugging and building releases. Every once in a while I try SwiftUI previews, but those are usually broken anyways.
Isn’t that easy to add with some rules and guidelines documents? I usually ask Claude code to research modern best practices for SwiftUI apps and to summarize the learnings in a rules file that will be part of the SwiftUI project.
Yes and no. Proper Agentic coding tools like Claude Code are a bit more than just a bunch of markdown rulesets.
For example: it uses Haiku as a model to run tools and most likely has automatic translations for when the model signals it wants to search or find something -> either use the built-in search or run find/fd/grep/rg
All that _can_ be done by prompting, but - as always with LLMS - prompts are more like suggestions.
Its not shipping the model in Xcode. You are still sending your data off to a remote provider, hoping that this provider behaves nicely with all this data and that the government will never force the provider to reveal your data.
It seems every IDE now has AI built-in. That's a problem if you're working on highly confidential code. You never know when the AI is going to upload code snippets to the server for analysis.
Not trying to be mean but I would expect comments on HN on these kind of stories to be from people who have used AI in IDEs at this point. There is no AI integration that runs automatically on a codebase.
This is not a realistic concern. If you're working on highly confidential code (in a serious meaning of that phrase), your while environment is already either offline or connecting only through a tightly controlled corporate proxy. There's no accidental leaks to AI from those environments.
The right middle ground is running Little Snitch in alert mode. The initial phase of training the filters and manually approving requests is painful, but it's a lot better than an air gap.
There are ranges of security concerns and high confidentiality.
For most corporate code (that is highly confidential) you still have proper internet access, but you sure as hell can't just send your code to all AI providers just because you want to, just because it's built into your IDE.
They both support it via plugins. Xcode doesn’t enable it by default, you need to enable it and sign into an account. It’s not really all that different.
What commonly gets installed in those cases is actual malware, a RAT (Remote Admin Tool) that lets the attacker later run commands on your laptop (kinda like an OpenSSH server, but also punching a hole through nat and with a server that they can broadcast commands broadly to the entire fleet).
If the attacker wants to use AI to assist in looking for valuables on your machine, they won't install AI on your machine, they'll use the remote shell software to pop a shell session, and ask AI they're running on one of their machines to look around in the shell for anything sensitive.
If an attacker has access to your unlocked computer, it is already game over, and LLM tools is quite far down the list of dangerous software they could install.
Maybe we should ban common RAT software first, like `ssh` and `TeamViewer`.
So called "Secure" shell is how many attackers got in my Bitcoin. They call it secure but then attackers were able to guess my password and now all my apes gone.
Edit: okay I paid the price by being rate-limited so I have to edit to reply.
The real thing is that top level comment was being utterly retarded so I decided to post something equivalently stupid. But in the end, it turns out the rest of you are kinder to idiots. That's probably better for all of you so I shall try to learn that lesson.
They could install anything. Including Claude Code and then run it in background as agent to exfiltrate data. I'm a security professional. This is unacceptable
I think the parent commenter was pointing out that, instead of installing Claude Code, they could just install actual malware. It's like that phrase Raymond Chen always uses: "you're already on the other side of the airtight hatchway."
There is a gulf and many shades between "this code should never be on an internet-connected device" and "it doesn't matter if this code is copied everywhere by absolutely anyone".
> In the OpenAI API, “GPT-5” corresponds to the “minimal” reasoning level, and “GPT-5 (Reasoning)” corresponds to the “low” reasoning level. (159135374)
It's interesting that the highest level of reasoning that GPT-5 in XCode supports is actually the "low" reasoning level. Wonder why.
you can use the API key, and it’ll give you access to all the model.
This is Claude sign in using your account. If you’ve signed up for Claude Pro or Max then you can use it directly. But, they should give access to Opus as well.
Still shocked Apple has not created thier own LLM, they have bought so many AI companies and have a rich talent pool and money so what's stopping them ?
“Boycott” is a pretty strong term. I’m sensing a strong dislike of ai from you which is fine but if you dislike a feature most people like it shouldn’t be surprising to you that you’ll find yourself mostly catered to by more niche editors.
I think it's a pretty good word, let's not forget how LLMs learned about code in the first place... by "stealing" all the snippets they can get their curl hands on.
And by reading the docs, and by autogenerating code samples and testing them against verifiers, and by paying a lot of people to write sample code for sample questions.
Your claim: "by reading the docs, and by autogenerating code samples and testing them against verifiers, and by paying a lot of people to write sample code for sample questions."
Your link: "Grade school math problems from a hardcoded dataset with hardcoded answers" [1]
GSM8K consists of 8.5K high quality grade school math word problems. Each problem takes between 2 and 8 steps to solve, and solutions primarily involve performing a sequence of elementary calculations using basic arithmetic operations (+ − × ÷) to reach the final answer.
1. OpenAI has been doing verifier-guided training since last year.
2. No SOTA model was trained without verified reward training for math and programming.
I supported the first claim with a document describing what OpenAI was doing last year; the extrapolation should have been straightforward, but it's easy for people who aren't tracking AI progress to underestimate the rate at which it occurs. So, here's some support for my second claim:
I couldn't get it to properly syntax highlight and autosuggest even after spending over an hour hunting through all sorts of terrible documentation for kate, clangd, etc. It also completely hides all project files that aren't in source control, and the only way to stop it is to disable the git plugin. What a nightmare. Maybe I'll try VSCodium next.
It can't access most Microsoft online services including Copilot, which happens to disable most of the features I don't want. (I understand this is both by design, as well as because Microsoft forbids unofficial forks from doing so.)
If you're on macOS there's Code Edit as a native solution (fully open source, not VC backed, MIT licensed), but it's currently in active development: https://www.codeedit.app/.
Otherwise there's VSCodium which is what I'm using until I can make the jump to Code Edit.
Okay dann lass die Ablage erst laufen ohne Teig dann kannst du mit Teig machen wenn du übergaben machst zwischen 13:30 und 14:00 Uhr dann bitte schichtführer/in Bescheid sagen bzw. geben tschüss
The so-called "guardrails" used for LLM are very close to expert systems, imo.
Since the landscape of potentially malicious inputs in plain english is practically infinite, without any particular enforced structure for the queries you make of it, means that those "guardrails" are, in effect, an expert system. An ever growing pile of if-then statements. Didn't work then, won't work now.
Neovim already supports LSP servers. The fact that a language server exists for anything, doesn't make neovim (or any other editor) "support" the technology. It doesn't, what it does support is LSP, and it doesn't and couldn't care less what language/slop the LSP is working with.
At the level of "Having to configure something to use it", they're the same, but then that's the same as the hundreds of other config options then. I think we can be slightly more precise than that.
In Neovim the choice of language server and the choice of LLM is up to the user, (possibly even the choice of this API, I believe, having only skimmed the PR) while both of those choices are baked in to XCode, so they're not the same thing.
That's fair enough, but it's the opposite complaint, that XCode's LLM support is more limited because it is proprietary. That's a perfectly valid and reasonable objection, of course.
If enough examples are in-distribution, the model's scroll bar implementation will work just fine. (Eventually, after the human learns what to ask for and how to ask for it.)
That is funny for how much is wrong. Ask the LLMs to vibe code a text editor and you'll get a React app using Supabase. Engineering !== Token prediction
Code is still there, but humans are done dealing with it. We're at a higher level of abstraction now. LLMs are like compilers, operating at a higher level. Nobody programs assembly language any more, much less machine language, even though the machine language is still down there in the end.
They certainly do, and I can't really follow the analogy you are building.
> We're at a higher level of abstraction now.
To me, an abstraction higher than a programming language would be natural language or some DSL that approximates it.
At the moment, I don't think most people using LLMs are reading paragraphs to maintain code. And LLMs aren't producing code in natural language.
That isn't abstraction over language, it is an abstraction over your computer use to make the code in language. If anything, you are abstracting yourself away.
Furthermore, if I am following you, you are basically saying, you have to make a call to a (free or paid) model to explain your code every time you want to alter it.
I don't know how insane that sounds to most people, but to me, it sounds bat-shit.
Of course it is, because that would be an aggressively stupid thing to do. Like boycotting syntax highlighting, spellckecking, VCS integration or a dozen other features that are th whole pint of IDEs.
If you don’t want to use LLM coding assistants – or if you can’t, or it’s not a technology suitable for your work – nobody cares. It’s totally fine. You don’t need to get performatively enraged about it.
I think all autocomplete solution are crappy, no matter how sophisticated the AI. It is surprising how often the obvious choice is wrong, but it often just is. I deactivated it.
Generating some code is fine, but I now prefer the deterministic autocomplete for my types I have available in my current context.
What was your problem with it? I see it running in a terminal more convenient (can point it to read local files outside of a project folder, for example)
The annoying thing is the official Swift extension can sometimes flag errors in perfectly fine code with zero problem in Xcode. So I’m forced to live with persistent “errors” when editing in VS Code/Cursor.
I’m building my first iOS app ever so I know it has much more to do with me not understanding Xcode but getting builds to succeed after making changes with Claude code has been a nightmare. If you or anyone have any tips, guides, prayers, incantations for how to get changes in one to not clobber the other and leave me in xproj symlink hell I would be so grateful.
if i could just get claude to properly remember it can directly edit the xcode project file, that'd be great.
for whatever reason it ignores my directive that it can from the CLAUDE file at least half the time. one time it even decided it needed to generate a fancy python script to do it. bizarre.
Apple really should open it up to any model provider that has an “OpenAI-style API” by letting the user put in a base URL, api key, model id, and a few params like context limit as needed.
Why would you limit users to Sonnet and not allow Opus when they are paying for their own account? I mean sure some people say Sonnet is good for coding but it seems needless to limit it in this way. Or they are just really slow to catch up… oh, right.
"Be ready for AIpple Revolution! We are making programming something different that hasn’t happen before! We are the first to introduce AI assisted agent coding with full integration with Siri, visionOS and so much more. New, holistic approach to creativity and efficiency"
Does anybody know why Anthropic doesn't let you remove your payment info from your account, or how to get support from them?
I bought a Pro subscription, the send button on their dumb chatbot box is disabled for me (on Safari), and I still get "capacity constraints' limits. Filed a chargeback with my bank just because of the audacity of their post-purchase experience. ChatGPT-5 works good enough for coding too.
From my experience with Claude Opus it seems like it tries to be "too smart" and doesn't seem to keep up with the latest APIs. It suggested some code for a iOS/macOS project that was only valid on tvOS, and other gaffs.
They also upgraded the GPT-4.1 (actually a special Apple variant) to GPT-5 by default, with the option to use GPT-5-thinking, using your ChatGPT subscription. I don't know if it's a special Apple variant of GPT-5 but this is a big upgrade and more exciting than Sonnet 3.7.
I also wonder if it will have separate rate limits from ChatGPT (app/web) and Codex CLI (which currently has its own rate limits).
Big disagree. My Macbook has hard crashed into restarting more times than my PC desktop has in years. The other day I literally just closed my M2 Macbook and it for some reason just completely crashed and shut itself off.
If you want me to blame "software design" then I can point to the fact that the TV app lags constantly on my M2 especially when I click the "download episode" button which should not ever be a thing on their native hardware, that there is a WindowsServer process that steadily eats up all the RAM that consistently needs to be quit in order to free up memory, that the headphone balance constantly resets to some random value where the audio comes mostly out of my left ear which has apparently been a problem for decades, that Spotlight is less than useless and STILL shows "Disk Utility" as the top result when I search "dis" instead of showing Discord an app that I open on an almost daily basis, that randomly the fingerprint login simply will not work for seconds at a time...
This is just the stuff I remember. This is a 16GB M2 Macbook Pro. Not a single native Apple made app or process should be lagging, let alone using up enough memory that sometimes apps just completely crash. Again, my windows desktop hasn't had a blue screen or a hard crash in the decade that I've used it and it had a worse configuration with 16GB of RAM and an old AMD CPU.
My hot take is all OSes are kinda bad for daily driving.
Apple has no qualms breaking backwards compatibility for core functions like bluetooth connectivity in MacOS. Windows has backwards compatibility, but increasingly worsening UX, throwing ads and subscriptions in your face before you can even log in, and a bad security/process isolation model. Desktop Linux is a case of "how many hours before I find out a critical part of my workflow is unsupported/bad/broken/unconfigurable/pain-to-configure in this particular distro/desktop environment".
I think they’re pretty amazing considering how hard a problem it is. Also, we forget how bad os’s used to be. They’re absolutely rock solid compared to the past.
I'm not trying to diminish the complexity of a desktop OS of course, but sometimes it's hard not to feel the priorities are all over the place. Don't get me wrong, I'm not nostalgic about Windows XP, I actually remember how many freezes and crashes I used to have back then.
My frustration is more born out of the OS rough edges constantly getting in the way of tasks I actually want to focus on and accomplish, which doesn't play well with my ADHD.
I have been trying to make iOS/macOS apps for years, but god, every time I have a go at it, Apple's documentation regime is still hot garbage. Eons ago I gave up Windows development because of Microsoft's inconsistent and uncertain APIs, but MS had great documentation. Apple is the opposite.
The "best" way to get the "latest" details on Apple's APIs is to suffer through mind-numbingly vapid WWDC videos with their reverse uncanny valley presenters (where humans pretend to be robots) and keep your full attention on them to catch a fleeting glimpse of a single method or property that does what you were looking for. Even 1.5x/2x speed is torture. I tried to get AIs to sift through the transcripts of their videos, and may Skynet forgive me for this cruelty.
Then when you go try to use that API, oops it's been changed in the current beta and there's no further documentation on it except auto-generated headers.
They also removed bookmarks from Xcode's built-in documentation browser years ago, and it doesn't retain a memory of previously open tabs, and often seems to be behind the docs on their websites.
I wish they would just provide open-source sample apps of each type (document-based, single-window etc.) for each of their platforms that fully use the latest APIs. At least that would be easier to ask AIs on, since that is what they seem to be going for now anyway.
I pretty much had the same experience recently when I had to deal with their Screen Time APIs. Had to go through the wwdc videos because the documentation was lack lustre.
It's a nice sentiment. The companies with the integrations are the ones that could take it back, but they don't have the incentive to break legal agreements and share with the world.
Meanwhile the creative output of humanity is distilled into black boxes to benefit those who can scrape it the most and burn the most power, but this impact is distributed amongst everyone, so again there's little incentive among those who could create (likely legal) change.
> Claude in Xcode is now available in the Intelligence settings panel, allowing users to seamlessly add their existing paid Claude account to Xcode and start using Claude Sonnet 4
All that dedicated silicon taking up space on their SoC and yet you still have to input your credit card in order to use their IDE. Come on...
To run a model locally, they would need to release the weights to the public and their competitors. Those are flagship models.
They would also need to shrink them way down to even fit. And even then, generating tokens on an apple neural chip would be waaaaaay slower than an HTTP request to a monster GPU in the sky. Local llms in my experience are either painfully dumb or painfully slow.
I bet Apple are working on it, it’s just not ready yet and they want to see how much people actually use it.
It’s the Apple way to screw the 3rd party and replace with their own thing once the ROI is proven (not a criticism, this is a good approach for any business where the capex is large…)
"Apple Intelligence", at least the part that's available to developers via the Foundation Models framework is a tiny ~3B model [0] with very limited context window. It's mainly good for simple things like tagging/classification and small snippets of text.
Local models and any OpenAI-compatible APIs are available to the Xcode Beta assistant. This is just a dedicated “sign in with x” rather than manual configuration.
Wow they're finally getting it. The AI breakthrough will not come from procedural generation of memojis - but rather enabling developers to use your platform. But with the nearly hostile stance of your 30% take, we will see how far this goes.
The one where you collect cash directly from users, and magically make handling that have zero overhead.
Credit card processing is hard... Go price out stripe + customer service + dealing with charge backs and tell me if you really want to do processing your self.
Well seeing that the most popular apps aside from games don’t have in app purchases and another subset of that has means to do payments subscriptions outside of the App Store, the 30% (actually 15% for small developers) is a boogeymen
The irony of this, is that Microsoft was trying to push CoPilot everywhere, however eventually Apple, Google and JetBrains have their own AI integrations, taking CoPilot out of the loop.
Slowly the AI craziness at Microsoft is taking the similar shape, of going all in at the begining and then losing to the competition, that they also had with Web (IE), mobile (Windows CE/Pocket PC/WP 7/WP 8/UWP), the BUILD sessions that used to be all about UWP with the same vigour as they are all AI nowadays, and then puff, competition took over even if they started later, because Microsoft messed up delivery among everyone trying to meet their KPIs and OKRs.
I also love the C++ security improvements on this release.
About GitHub Copilot in specific: One big negative was how when GPT-4 became available that Microsoft didn't upgrade paying Copilot users to it, they simply branded this "coming soon"/"beta" Copilot X for a while. We simply cancelled the only Copilot subscription we had at work.
Copilot subscription?
I've been getting monthly emails that my free access for GitHub Copilot has been renewed for another month… for years. I've never used it, I thought that all GitHub users got it for free.
Just because you can’t or won’t win the market with your opportunistic investment, doesn’t mean you should let your competitors completely annihilate you by taking that investment for themselves.
Google, Apple, FB or AWS would have been suitors for that licensing deal if MS didn’t bite.
CoPilot isn't anything Microsoft is trying to sell outside of their own products. And with GitHub Copilot there is no "copilot" model to choose, you can choose between Anthropic, OpenAI and Google models.
Sure UWP never caught on, but you know why? Win32, which by the way is also Microsoft, was way to popular and more flexible. Devs weren't going to re-write their apps to UWP in order to support phones.
Microsoft owns 49% of OpenAI so why they should worry? JetBrains just proudly announce that they now use GPT-5 by default.
> going all in at the begining and then losing to the competition
Sure, but there are counter examples too. Microsoft went late to the party of cloud computing. Today Azure is their main money printing machine. At some point Visual Studio seemed to be a legacy app only used for Windows-specific app development. Then they released VSCode and boom! It became the most popular editor by a huge margin[0].
[0]: https://survey.stackoverflow.co/2025/technology#most-popular...
Anecdotally: Azure is the Teams of cloud services - nobody uses it voluntarily or because it's technically the best solution.
They use it because the corporation mandates it.
> Microsoft owns 49% of OpenAI
Power at OpenAI seems orthogonal to ownership, precedent or even frankly their legal documents.
Microsoft mistook a product game for a distribution one. AI quality is heterogenous and advancing enough that people will make an effort to use the one they like best. And while CoPilot is excellently distributed, it’s a crap product, in large part due to the limits Microsoft put on GPT.
Maybe because Microsoft is a shit company and anything they do is sus af. And everyone knows it. And I'm tired of pretending like it's not. I wouldn't trust Microsoft to babysit my mortal enemy's kids.
Maybe if they weren't literally the borg people would open their hearts and wallets to Redmond. They saw that Windows 10 was a privacy nightmare and what did they do? They doubled down in Windows 11. Not that I care but it plays really poorly. Every nerd on the internet spouts off about Recall even though it's not even enabled if you install straight to the latest build.
They bought GitHub and now it's a honeypot. We live in a world where we have to assume GitHub is adversarial.
_NSAKEY???
Fuck you Microsoft.
Makes sense karma catches up to them. Maybe if their mission statement and vision were pure or at least convincing they would win hearts and minds.
Also OpenAI pioneered but now the many competitors seem to have either caught up or surpassed them. They might still retain a significant brand recognition advantage as long as they don't fall too far behind, though.
"Taking Copilot out of the loop" if you ignore the massive ecosystems of Github, Visual Studio, and Visual Studio Code.
Different CoPilot product. Typical Microsoft naming confusion.
umm I don't know what you are talking about, I use a Github Copilot 40 USD subscription in VSCode to code using various models, and this is the industry standard now in my region, as most employers are now giving employees the 10 USD subscription.
Almost no one uses copilot unless they are not allowed to use anything else or don’t know any better. MS could have been a leader in this space but MS couldn’t understand why people didn’t like copilot but loved the competition.
3 days ago I saw another Claude praising submission on HN, and finally I signed up for it, to compare it with copilot.
I asked 2 things.
1. Create a boilerplate Zephyr project skeleton, for Pi Pico with st7789 spi display drivers configured. It generated garbage devicetree which didn't even compile. When I pointed it out, it apologized and generated another one that didn't compile. It configured also non-existent drivers, and for some reason it enabled monkey test support (but not test support).
2. I asked it to create 7x10 monochromatic pixelmaps, as C integer arrays, for numeric characters, 0-9. I also gave an example. It generated them, but number eight looked like zero. (There was no cross in ether 0 nor 8, so it wasn't that. Both were just a ring)
What am I doing wrong? Or is this really the state of the art?
It’s good at doing stuff like “host this all in Docker. Make a Postgres database with a Users table. Make a FastAPI CRUD endpoint for Users. Make a React site with a homepage, login page, and user dashboard”.
It’ll successfully produce _something_ like that, because there’s millions of examples of those technologies online. If you do anything remotely niche, you need to hold its hand far more.
The more complicated your requirements are, the closer you are to having “spicy autocomplete”. If you’re just making a crud react app, you can talk in high level natural language.
Did you try claude code and spend actual time going back and forth with it, reviewing it's code and providing suggestions; Instead of just expecting things to work first try with minimal requirements?
I see claude code as pair programming with a junior/mid dev that knows all fields of computer engineering. I still need to nudge it here and there, it will still make noob mistakes that I need to correct and I let it know how to properly do things when it gets them wrong. But coding sessions have been great and productive.
In the end, I use it when working with software that I barely know. Once I'm up and running, I rarely use it.
> Did you try claude code and spend actual time going back and forth with it, reviewing it's code and providing suggestions; Instead of just expecting things to work first try with minimal requirements?
I did, but I always approached LLM for coding this way and I have never been let down. You need to be as specific as possible, be a part of the whole process. I have no issues with it.
I agree, but I think there's an important distinction to be made.
In some cases, it just doesn't have the necessary information because the problem is too niche.
In other cases, it does have all the necessary information but fails to connect the dots, i.e. reasoning fails.
It is the latter issue that is affecting all LLMs to such a degree that I'm really becoming very sceptical of the current generation of LLMs for tasks that require reasoning.
They are still incredibly useful of course, but those reasoning claims are just false. There are no reasoning models.
FWIW, I used Gemini to write an Objective-C app for Apple Rhapsody (!) that would enumerate drivers currently loaded by the operating systems (more or less save level of difficulty as the OP, I'd say?), using the PDF manual of NextStep's DriverKit as context.
It... sort of worked well? I had to have a few back-and-forth because it tried to use Objective-C features that did not exist back then (e.g. ARC), but all in all it was a success.
So yeah, niche things are harder, but on the other hand I didn't have to read 300 pages of stuff just to do this...
I remember writing obj-c naturally by hand. Before swift was even a twinkle in tim cooks eye. One of my favorite languages to program in I had a lot of fun writing ios apps back in the day it seems like
In other words, the vibe coders of this world are just redundant noobs who don't really belong on the marketplace. They've written the same bullshit CRUD app every month for the past couple of years and now they've turned to AI to speed things up
Last week I asked Claude to improve a piece of code that downloads all AWS RDS certificates to just the ones needed for that AWS region. It figured out several ways to determine the correct region, made a nice tradeoff and suggested the most reliable way. It rewrote the logic to download the right set, did some research to figure out the right endpoint in between. It only made one mistake, it fallback mechanism was picking EU, which was not correct. Maybe 1 hour of work. On my own it would have taken me close to a working day to figure it all out.
This is just a thought experiment.
I don't mean to be treading on feet but I'm noticing this more and more in the debates around AI. Imagine if there are developers out there that could have done this task in 30 mins without AI.
The level of performanace of AI solutions is heavily related to the experience level of the developer and of the problem space being tackled - as this thread points out.
Unfortunately the marketing around AI ignores this and makes every developer not using AI for coding seem like a dinosauer, even though they might well be faster in solving their particular problems.
AI is moving problem solving skills from coding to writing the correct prompts and teaching AI to do the right thing - which, again, is subjective, since the "right thing" for one developer isn't the "right thing" for the another developer. "Right thing" being the correct solution, the understandable solution, the fastest solution, etc depending on the needs of the developer using the AI.
I think the majority of coders out there write the same CRUD app over and over again in different flavors. That's what the majority of businesses seem to pay for.
If a business needs the equivalent of a Toyota Corolla, why be upset about the factory workers making the millionth Toyota Corolla?
> I think the majority of coders out there write the same CRUD app over and over again in different flavors
In my experience, that's not entirely true. Sure, a lot of app are CRUD apps, but they are not the same. The spice lies in the business logic, not in programming the CRUD operations. And then of course, scaling, performance, security, organization, etc etc.
"What am I doing wrong?"
Your first prompt is testing Claude as an encyclopedia: has it somehow baked into its model weights the exactly correct skeleton for a "Zephyr project skeleton, for Pi Pico with st7789 spi display drivers configured"?
Frequent LLM users will not be surprised to see it fail that.
The way to solve this particular problem is to make a correct example available to it. Don't expect it to just know extremely specific facts like that - instead, treat it as a tool that can act on facts presented to it.
For your second example: treat interactions with LLMs as an ongoing conversation, don't expect them to give you exactly what you want first time. Here the thing to do next is a follow-up prompt where you say "number eight looked like zero, fix that".
> What am I doing wrong
Trying two things and giving up. It's like opening a REPL for a new language, typing some common commands you're familiar with, getting some syntax errors, then giving up.
You need how to learn to use your tools to get the best out of them!
Start by thinking about what you'd need to tell a new Junior human dev you'd never met before about the task if you could only send a single email to spec it out. There are shortcuts, but that's a good starting place.
In this case, I'd specifically suggest:
1. Write a CLAUDE.md listing the toolchains you want to work with, giving context for your projects, and listing the specific build, test etc. commands you work with on your system (including any helpful scripts/aliases you use). Start simple; you can have claude add to it as you find new things that you need to tell it or that it spends time working out (so that you don't need to do that every time).
2. In your initial command, include a pointer to an example project using similar tech in a directory that claude can read
3. Ask it to come up with a plan and ask for your approval before starting
I guess many find comfort in being able to task an ai with assignments that it cannot complete. Most sr developers I work with take this approach. It's not really a good way of assessing the usefulness of a tool though.
He asked what he was doing wrong?
too big of tasks. break them down and then proceed from there. have it build out task lists in a TASKS.md. review those tasks. do you agree? no? work with it to refine. implement one by one. have it add the tests. refactor after awhile as {{model}} doesn't like to do utility functions a lot. right now, about +50k lines in to a project that's vibecoded. i sit back and direct and it plays.
Imagine the CS 100 class where they ask you to make a PB&J. saying for it to make it, there's a lot of steps, but determine known the steps. implement each step. progress.
Too big and requiring too much niche specific knowledge, you somehow have to inject that knowledge and allow it to iterate.
This is the way.
I run interviews at my company. We allow/encourage AI.
The number one failure method is people throwing all of the requirements in upfront. They get one good pass then fail.
I'm inclined to agree with this approach because someone not using AI who fails would likely fail for the same reasons. If you can't logically distill a problem into parts you can't obtain a solution.
What an odd thing to ask it. I installed claude code and ran it from my terminal. Just asked it to simply give me a node based rest API with X endpoints with these jobs, and then I told it to write the unreal engine c++ to consume those endpoints. 2500 lines of code later, it worked.
The only way I manage to get any benefits from LLMs is to use them as an interactive rubber duck.
Dump your thoughts in a somewhat arranged manner, tell it about your plan, the current status, the end goal, &c. After that tell it to write 0 code for now but to ask questions and find gaps in your plan. 30% of it will be bullshit but the rest is somewhat useable. Then you can ask for some code but if you care about quality or consistency with you existing code base you probably will have to rewrite half of it, and that's if the code works in the first place
Garbage in garbage out is true for training but it's also true for interactions
> It configured also non-existent drivers, and for some reason it enabled monkey test support (but not test support).
If it doesn't have the underlying base data, it tends to hallucinates. (It's getting a bit difficult to tell when it has underlying data, because some models autonomously search the web). The models are good at transforming data however, so give it access to whatever data it needs.
Also let it work in a feedback loop: tell it to compile and fix the compile errors. You have to monitor it because it will sometimes just silence warnings and use invalid casts.
> What am I doing wrong? Or is this really the state of the art?
It may sound silly, but it's simply not good at 2D
> It may sound silly, but it's simply not good at can2D.
It's not silly at all, it's not very good at layouts either, it can generally make layouts but there is a high chance for subtle errors, element overlaps, text overflows, etc.
Mostly because it's a language model, i.e it doesn't generally see what it makes, you can send screenshots apparently and it will use it's embedded vision model, but I have not tried that.
Think of Claude as a typical software developer.
If you just selected a random developer do you think they're going to have any idea why your talking about?
The issue is LLMs will never say, sorry, IDK how to do this. Like a stressed out intern they just make up stuff and hope it passes review.
> What am I doing wrong? Or is this really the state of the art?
You're treating the tool like it was an oracle. The correct way is to treat it as a somewhat autistic junior dev: give it examples and process to follow, tell it to search the web, read the docs, how to execute tests. Especially important is either directly linking or just copy pasting any and all relevant documentation.
The tool has a lossily compressed knowledge database of the public internet and lots of books. You want to fix the relevant lossy parts in the context. The less popular something is, the more context will be needed to fill the gaps.
> What am I doing wrong?
Providing a woefully inadequate descriptions to others (Claude & us) and still expecting useful responses?
Try this prompt: Create a detailed step by step plan to implement a boilerplate Zephyr project skeleton for Pi Pico with configured st7789 SPI display drivers
Ask Opus or Gemini 2.5 Pro to write a plan. Then ask the other to critique it and fix mistakes. Then ask Sonnet to implement
I tried this myself and IMO, this might be basic and day-to-day for you, with unambiguous correct paths to follow, but this is pretty niche nevertheless. LLMs thrive when there's a wealth of examples and I struggle to Google what you asked myself, meaning that LLM will perform even worse than my try.
I found that second line works well for image prompts too. Tell one AI to help you with a prompt, and then take it back to the others to generate images.
Sounds like you picked some obscure tasks to test it that would obviously have low representation in the data set? That is not to say it can't be helpful augmenting some lower represented frameworks/tools - just you'll need to equip it with better context (MCPs/Docs/Instruction files)
A key skill in using an LLM agentic tool is being discerning in which tasks to delegate to it and which to take on yourself. Try develop that skill and maybe you will have better luck.
There's a lot of people caricaturing the obvious fact that any model works best in distribution.
The more esoteric your stack, and the more complex the request, the more information it needs to have. The information can be given either through doing research separately (personally, I haven't had good results when asking Claude itself to do research, but I did have success using the web chat UI to create an implementation plan), or being more specific with your prompt.
As an aside, I have more than 10 years of experience, mostly with backend Python, and I'd have no idea what your prompts mean. I could probably figure it out after some google searches, tho. That's also true of Claude.
Here's an example of a prompt that I used recently when working on a new codebase. The code is not great, the math involved is non trivial (it's research-level code that's been productionized in hurry). This literally saved 4 hours of extremely boring work, digging through the code to find various hardcoded filenames, downloading them, scp'ing them, and using them to do what I want. It one-shotted it.
> The X pipeline is defined in @airflow/dags/x.py, and Y in `airflow/dags/y.py` and the relevant task is `compute_X`, and `compute_Y`, respectively. Your task is to:
> 1. Analyze the X and Y DAGs and and how `compute_X` functions are called in that particular context, including it's arguments. If we're missing any files (we're probably missing at least one), generate a .sh file with aws cli or curl commands necessary for downloading any missing data (I don't have access to S3 from this machine, but I do have in a remote host). Use, say, `~/home` as the remote target folder.
> 2. If we needed to download anything from S3, i.e. from the remote host, output rsync/scp commands I can use to copy them to my local folder, keeping the correct/expected directory structure. Note that direct inputs reside under `data/input`, while auxiliary data resides in other folders under `data`. Do not run them, simply output them. You can use for example `scp user@server.org ...`
> 3. Write another snapshot test for X under `tests/snapshot`, and one for Y. Use a pattern as similar as possible to the other tests there. Do not attempt to run the tests yet, since I'll need to download the data first.
> If you need any information from Airflow, such as logs or output values, just ask and I can provide them. Think hard.
Real vibe coding is fake, especially for something niche like what you asked it to do. Imagine a hyperactive eidetic fresh out of high school was literally sitting in the other room. What would you tell her? That’s a good rule of thumb for the level of detail and guidance
Ok. several tips I can give. 1. Setup a sub-agent to do RESEARCH. It is important that it only has read-only and web access tools. 2. Use planning mode and also ask the agent to use the subagent to research best pratices with the tech that you are wanting to do, before it builds a plan. 3. When ever it gets hung up.. tell it to use the sub-agent to research the solution.
That will get you a lot better initial solution. I typically use Sonnet for the sub-agents and Opus for the main agent, but sonnet all around should be fine too for the most part.
You can no longer answer "what is the state of the art” by pointing to a model.
Generating a state-of-the-art response to your request involves a back-and-forth with the agent about your requirements, having a agent generate and carry out a deep research plan to collect documentation, then having the agent generate and carry out a development plan to carry it out.
So while Claude is not the best model in terms of raw IQ, the reason why it's considered the best coding model is because of its ability to execute all these steps in one go which, in aggregate, generates a much better result (and is less likely to lose its mind).
> So while Claude is not the best model in terms of raw IQ
Which one is, and by what metric? I always end up back at Claude after trying other models because it is so much better at real world applications.
In my experience Claude is quite good at the popular stacks in the JavaScript, Python and PHP world. It struggled quite a bit when I asked it non-trivial questions in C or Rust for example. For smaller languages (e.g., Crystal) it seems to hallucinate a lot. I think since a lot of people work in JS, Python and PHP, that’s where Claude is very valuable and that’s where a lot of the praise feel justified too.
Feed it Crystal documentation and example code. That is what I did with more obscure programming languages and it worked out well in the end.
I have had no problems with using Claude on large rust projects. The compiler errors usually point it towards fixing its mistakes (just like they do for me).
You didn't specify any architecture design. Your prompts are about 10% of what would be needed to one shot this. This is what you do wrong.
I've had similar experiences when working on non-web tech.
There are parts in the codebase I'd love some help such as overly complex C++ templates and it almost never works out. Sometimes I get useful pointers (no pun intended) what the problem actually is but even that seems a bit random. I wonder if it's actually faster or slower than traditional reading & thinking myself.
One of the things you can do is provide a guidance file like CLAUDE.md including not only style preferences but also domain knowledge so it has greater context and knows where to look. Just ask it make one and then update and change as needed.
I find it useful to ask it to build a design document first and push to add details where i see it lacking.
After a few iteration i then ask it to implement the design doc to mostly-better results.
I managed to get most AIs to generate C# code when I ask for Java stuff, so it is always a kind of template generator that still isn't quite there.
The thing you are doing wrong is asking it to solve hard problems. Claude Code excels at solving fairly easy, but tedious stuff. Refactors that are brainless but take an hour. It will knock those out of the park. Fire up a git worktree and let it spin on your tedious API changes and stuff while you do the hard stuff. Unfortunately, you'll still need to use your brain for that.
Tbh dawg, those tasks sound intentionally obtuse. It looks like u are doing more esoteric work than the crud react slop us mortals play in on the daily which is where ai shines.
I work almost exclusively with embedded devices, with low level code (mostly C, Rust, Assembly and related frameworks) - and that's where I also ask for help from LLMs.
Did you intentionally pick your career to make the AI look bad?
It works fine in those domains. I speak from experience. You need CI tools the agent can access, and lots of tests.
If you ask more than a single function, its more trouble than worth
> What am I doing wrong?
My coding ranges from "exotic" to "boiler plate" on any given day.
> Create a boilerplate Zephyr project skeleton, for Pi Pico
Yea... Asking Claude to help you with a low documentation build root system is going to go about the same way, I know first hand about how this works.
> I asked it to create 7x10 monochromatic pixelmaps
Wrong tool for the job here. I dont think IDE and Pixelmaps have as large of an intersection as you think they do. Claude thinks in tokens not pixels.
Pick a common language (js, python, rust, golang) pick something easy (web page, command line script, data ingestion) and start there. See what it can do and does well, then start pushing into harder things.
What you're doing wrong is that you're asking it for something more complicated than babby's first webapp in javascript/python.
When people say things like "I told Claude what I wanted and it did it all on the first try!", that's what they mean. Basic web stuff that that is already present in the model's training data in massive volumes, so it has no issue recreating it.
No matter how much AI fanatics try to convince you otherwise, LLMs are not actually capable of software engineering and never will be. They are largely incapable of performing novel tasks that are not already well represented in their weights, like the ones you tried.
What they are not capable of is replacing YOU, the human who is supposed to be part of the whole process (incl. architectural). I do not think that this is a limitation. In fact, I like being part of the process.
So I've used Zephyr. The thing you're doing wrong is expecting LLMs to scaffold you a bunch of files from a relatively niche domain. Zephyr is also a mess of complexity with poor documentation. You should ask it to consult official docs and ask it to use existing tools (west etc) and board defs to do the scaffolding.
I just had AI write me a scraper and download 5TB of invaluable data which I had been eyeing for a long time. All in ten days. At the end of it, I still don’t know anything about python. It’s a bliss for people like me. All dependencies installed themselves. I look forward to using it even more.
One frustration was the code changed so much in ChatGPT so had to be lots of prompts. But I had no idea what the code was anyways. Understood vibe coding. Just used ChatGPT on a whim. Liked the end result.
Write some hooks dawg
Interesting to think about how Apple get to make product decisions based on Gatekeeper OCSP analytics now that every app launch phones home. They must know exactly how popular VSCode is.
Facebook got excoriated for doing that with Onavo but I guess it's Good Actually when it's done in the name of protecting my computer from myself lol
Apple doesn't need telemetry to send emails about their favorite coding AI to the 2 Xcode users
Off by about 33,999,998 users, but still a decent dunk.
https://appleinsider.com/articles/22/06/06/apple-now-has-ove...
34 million developers? That number doesn't even pass a basic sniff test. Are there 34 million people that have Xcode installed? That I can believe.
> At Apple's World Wide Developer Conference on Monday, Tim Cook mentioned that there are now 34 million registered developers with the company's platform.
I think that means either:
If my experience is anything to go by - a good proportion of this will be people accidentally double clicking a .md (or other random text suffix), and cursing whilst they wait for XCode to slowly load enough that they can quit it and open the file in a proper lightweight editor..
I feel like the #1 reason to install Xcode is to get Git working on macOS. Yours is probably #2. I wouldn't bet money on iOS/macOS development sitting at #3.
This won't make a dent. It still doesn't support any agentic operation.
The real news is when Codex CLI / Claude Code get integrated, or Apple introduces a competitor offering to them.
Until then this is a toy and should not be used for any serious work while these far better tools exist.
I just installed it—definitely not a toy.
Compared to stock Claude Code, this version of Claude knows a lot more about SwiftUI and related technologies. The following is output from Claude in Xcode on an empty project. Claude Code gives a generic response when it looked at the same project:
If a bunch of markdown files forced into the context is “knowing”, then yes. They are usually located at /Applications/Xcode-beta.app/Contents/PlugIns/IDEIntelligenceChat.framework/Versions/A/Resources/AdditionalDocumentation
You are free to point Claude Code to that folder, or make a slash command that loads their contents. Or, start CC with -p where the prompt is the content of all those files.
Claude Code integration in Xcode would be very cool indeed, but I might still stick with VSCode for pure coding.
> Claude Code integration in Xcode would be very cool indeed, but I might still stick with VSCode for pure coding.
I'm sticking with VSCode too, but it's a bit silly to suggest that anyone is using XCode because it's their preferred IDE. It's just the one that's necessary for any non-trivial Apple platform development.
Adding a code generator isn't a marketing ploy to get people to switch editors, it's just a small concession to the many hapless souls stuck dealing with Apple on the professional side, or masochistically building mac SwiftUI apps just to remind themselves what pain feels like.
I mean you can stay in VSCode for most activities if you hate Xcode that much (I can relate btw). Plugins like Sweetpad make this possible. My approach now is to develop all logic in small Swift packages and run swift test in VSCode (or Claude Code), so I only absolutely need Xcode for debugging and building releases. Every once in a while I try SwiftUI previews, but those are usually broken anyways.
Isn’t that easy to add with some rules and guidelines documents? I usually ask Claude code to research modern best practices for SwiftUI apps and to summarize the learnings in a rules file that will be part of the SwiftUI project.
Yes and no. Proper Agentic coding tools like Claude Code are a bit more than just a bunch of markdown rulesets.
For example: it uses Haiku as a model to run tools and most likely has automatic translations for when the model signals it wants to search or find something -> either use the built-in search or run find/fd/grep/rg
All that _can_ be done by prompting, but - as always with LLMS - prompts are more like suggestions.
I'm as crazy about AI as the next dev, but that has to be the weakest example of AI capability that I have ever seen.
Its not shipping the model in Xcode. You are still sending your data off to a remote provider, hoping that this provider behaves nicely with all this data and that the government will never force the provider to reveal your data.
They are already forcing OpenAI to keep all logs. Go figure.
And people talk to GPT about very private things, using it as a shrink. What can go wrong.
It seems every IDE now has AI built-in. That's a problem if you're working on highly confidential code. You never know when the AI is going to upload code snippets to the server for analysis.
Not trying to be mean but I would expect comments on HN on these kind of stories to be from people who have used AI in IDEs at this point. There is no AI integration that runs automatically on a codebase.
This could change on a daily basis, and it's a valid concern anyway.
There is automatic code indexing from Cursor.
Autocomple is also automatically triggered when you place your cursor inside the code.
Yes, Cursor, “The AI Code Editor.”
Cursor is an AI IDE and not what they are describing.
> There is no AI integration that runs automatically on a codebase.
Don't be naive.
Gitkraken does
This is not a realistic concern. If you're working on highly confidential code (in a serious meaning of that phrase), your while environment is already either offline or connecting only through a tightly controlled corporate proxy. There's no accidental leaks to AI from those environments.
thanks for giving the security department more reasons to think that way.
I spent the last 6 months trying to convince them not to block all outbound traffic by default.
The right middle ground is running Little Snitch in alert mode. The initial phase of training the filters and manually approving requests is painful, but it's a lot better than an air gap.
that’s what I do, but since it’s in my control the security teams don’t like it. ;)
There are ranges of security concerns and high confidentiality.
For most corporate code (that is highly confidential) you still have proper internet access, but you sure as hell can't just send your code to all AI providers just because you want to, just because it's built into your IDE.
> "add their existing paid Claude account to Xcode and start using Claude Sonnet 4"
Wont work by default if I'm reading this correctly
Neovim and Emacs don’t have it built in. Use open source tools.
They both support it via plugins. Xcode doesn’t enable it by default, you need to enable it and sign into an account. It’s not really all that different.
[flagged]
What commonly gets installed in those cases is actual malware, a RAT (Remote Admin Tool) that lets the attacker later run commands on your laptop (kinda like an OpenSSH server, but also punching a hole through nat and with a server that they can broadcast commands broadly to the entire fleet).
If the attacker wants to use AI to assist in looking for valuables on your machine, they won't install AI on your machine, they'll use the remote shell software to pop a shell session, and ask AI they're running on one of their machines to look around in the shell for anything sensitive.
If an attacker has access to your unlocked computer, it is already game over, and LLM tools is quite far down the list of dangerous software they could install.
Maybe we should ban common RAT software first, like `ssh` and `TeamViewer`.
> They won’t install AI on your machine
Actually they’ll just the AI you already have on your machine[0]
In this attack, the malware would use Claude Code (with your credentials) to scan your own machine.
Much easier than running the inference themselves!
[0]https://semgrep.dev/blog/2025/security-alert-nx-compromised-...
So called "Secure" shell is how many attackers got in my Bitcoin. They call it secure but then attackers were able to guess my password and now all my apes gone.
Edit: okay I paid the price by being rate-limited so I have to edit to reply.
The real thing is that top level comment was being utterly retarded so I decided to post something equivalently stupid. But in the end, it turns out the rest of you are kinder to idiots. That's probably better for all of you so I shall try to learn that lesson.
You know, I should have realized this was a troll account with the previous comment.
I guess that's on me for being oblivious enough that it took this obvious of a comment for me to be sure you're intentionally trolling. Nice work.
If you're worried about someone accessing your unlocked computer to install LLMs, you might need to rethink your security model.
They could install anything. Including Claude Code and then run it in background as agent to exfiltrate data. I'm a security professional. This is unacceptable
I think the parent commenter was pointing out that, instead of installing Claude Code, they could just install actual malware. It's like that phrase Raymond Chen always uses: "you're already on the other side of the airtight hatchway."
Yes but Claude Code could install malware when I'm not paying attention. And when I remove with MalwareBytes it will return because LLMs are not AGI.
Yes. I am so worried as well. This is why I installed an AI to double-check if the password I entered is correct when logging in. Fight fire with fire
You should install LuLu if you’re that concerned. There are far more nefarious ways of “getting your data”.
https://objective-see.org/products/lulu.html
On IDEA the organisation who controls the license can disable the build in (remote) AI. (Not the local auto complete one)
But I guess the user could still get a 3rd party plugin.
Well that depends on whether you give it access or not, apple’s track record with privacy gives me some hope
Most of the big corporations will have a special contract with the AI labs with 0 retention policies.
I do not think this will be an issue for big companies.
No. It’s always something you have to turn on or log into.
Also, there are plenty of editors and IDEs that don’t.
Let’s stop pretending like you’re being forced into this. You aren’t.
Sublime Text doesn't by default.
If it's that confidential you should be on an airgapped network.
There's simply no way to properly secure network connected developer systems.
People working on highly confidential code will NOT have access to the public internet.
There is a gulf and many shades between "this code should never be on an internet-connected device" and "it doesn't matter if this code is copied everywhere by absolutely anyone".
To me "highly confidential" would mean "isolated from the internet" or else it isn't going to be "highly confidential" for very long.
Have you seen a lot of code from Klarna, Storytel, Spotify (companies I've worked at)?
None of these companies are isolated from the internet.
many enterprises store their code on GitHub, owned by Microsoft, operator of Copilot
you can use Claude via bedrock and benefit from AWS trust
Gemini? Google owns your e-mail. Maybe you're even one of those weirdos who doesn't use Google for e-mail - I bet your recipient does.
so... they have your code, your secrets, etc.
You do know: when it's enabled.
> In the OpenAI API, “GPT-5” corresponds to the “minimal” reasoning level, and “GPT-5 (Reasoning)” corresponds to the “low” reasoning level. (159135374)
It's interesting that the highest level of reasoning that GPT-5 in XCode supports is actually the "low" reasoning level. Wonder why.
Yeah I don't get why they don't support Opus given that you're bringing your own API key.
you can use the API key, and it’ll give you access to all the model.
This is Claude sign in using your account. If you’ve signed up for Claude Pro or Max then you can use it directly. But, they should give access to Opus as well.
They should document it that way.
It's available now. Here’s short but more complete context than the submitted title or the Xcode release note: https://sixcolors.com/link/2025/08/apples-new-xcode-beta-add...
I haven't opened up xcode in years (thank god) I suppose this was inevitable. They have to keep up
Still shocked Apple has not created thier own LLM, they have bought so many AI companies and have a rich talent pool and money so what's stopping them ?
But they won't fix the infinite number of bugs Xcode has, its slowness and subpar ux
It's getting harder to find IDEs that properly boycott LLMs.
In a similar vein I can barely find an OS that refuses to connect to the internet
Wouldn't the more correct analogy be a text editor without "Klippy?"
too many of them these days: https://kakoune.org/
They don’t think it be like it is, but it do.
https://templeos.org/
I hate that most browsers are willing to render React SPAs.
[dead]
Really?
“Boycott” is a pretty strong term. I’m sensing a strong dislike of ai from you which is fine but if you dislike a feature most people like it shouldn’t be surprising to you that you’ll find yourself mostly catered to by more niche editors.
I think it's a pretty good word, let's not forget how LLMs learned about code in the first place... by "stealing" all the snippets they can get their curl hands on.
And by reading the docs, and by autogenerating code samples and testing them against verifiers, and by paying a lot of people to write sample code for sample questions.
Yeah, none of that happened with LLMs
https://openai.com/index/prover-verifier-games-improve-legib... OpenAI has been doing verifier-guided training since last year. No SOTA model was trained without verified reward training for math and programming.
Your claim: "by reading the docs, and by autogenerating code samples and testing them against verifiers, and by paying a lot of people to write sample code for sample questions."
Your link: "Grade school math problems from a hardcoded dataset with hardcoded answers" [1]
It really is the same thing.
[1] https://openai.com/index/solving-math-word-problems/
--- start quote ---
GSM8K consists of 8.5K high quality grade school math word problems. Each problem takes between 2 and 8 steps to solve, and solutions primarily involve performing a sequence of elementary calculations using basic arithmetic operations (+ − × ÷) to reach the final answer.
--- end quote ---
My two claims:
1. OpenAI has been doing verifier-guided training since last year.
2. No SOTA model was trained without verified reward training for math and programming.
I supported the first claim with a document describing what OpenAI was doing last year; the extrapolation should have been straightforward, but it's easy for people who aren't tracking AI progress to underestimate the rate at which it occurs. So, here's some support for my second claim:
https://arxiv.org/abs/2507.06920 https://arxiv.org/abs/2506.11425 https://arxiv.org/abs/2502.06807
Just disable the feature/plugin in your IDE of choice. Android Studio/IntelliJ: https://i.imgur.com/RvRMvvK.png
Just don't use the features.
https://kate-editor.org/
I couldn't get it to properly syntax highlight and autosuggest even after spending over an hour hunting through all sorts of terrible documentation for kate, clangd, etc. It also completely hides all project files that aren't in source control, and the only way to stop it is to disable the git plugin. What a nightmare. Maybe I'll try VSCodium next.
I thought vscodium was just vscode but open source. Won't any issues in vscode also be present in vscodium?
It can't access most Microsoft online services including Copilot, which happens to disable most of the features I don't want. (I understand this is both by design, as well as because Microsoft forbids unofficial forks from doing so.)
However, MS do everything they can to make plugins not work in VSCodium. And the plugin marketplaces are separate now.
Many of the popular features in VS Code are provided by plugins that are not open source and thus not provided with VSCodium.
Kate is brilliant.
If you're on macOS there's Code Edit as a native solution (fully open source, not VC backed, MIT licensed), but it's currently in active development: https://www.codeedit.app/.
Otherwise there's VSCodium which is what I'm using until I can make the jump to Code Edit.
Okay dann lass die Ablage erst laufen ohne Teig dann kannst du mit Teig machen wenn du übergaben machst zwischen 13:30 und 14:00 Uhr dann bitte schichtführer/in Bescheid sagen bzw. geben tschüss
How about Sublime Text (not really an IDE, just text editor)
Neovim, emacs?
Amusing that Emacs that came out of the MIT AI lab, and heavily uses Lisp, a language that used to be en vogue for AI research.
Amusing is one word for it. Expert systems were all the rage until they weren't. We'll see how LLMs do by comparison.
The so-called "guardrails" used for LLM are very close to expert systems, imo.
Since the landscape of potentially malicious inputs in plain english is practically infinite, without any particular enforced structure for the queries you make of it, means that those "guardrails" are, in effect, an expert system. An ever growing pile of if-then statements. Didn't work then, won't work now.
neovim will support llms natively (though a language server) https://github.com/neovim/neovim/pull/33972
Neovim already supports LSP servers. The fact that a language server exists for anything, doesn't make neovim (or any other editor) "support" the technology. It doesn't, what it does support is LSP, and it doesn't and couldn't care less what language/slop the LSP is working with.
That’s not really native support for LLMs? It’s supporting some LSP feature for completions.
You have to enable it and install a language server, that's not the same as an LLM being baked in.
It’s not baked in, in that sense. You still have to enable it in XCode and link it to a Claude account. It’s basically the same.
At the level of "Having to configure something to use it", they're the same, but then that's the same as the hundreds of other config options then. I think we can be slightly more precise than that.
In Neovim the choice of language server and the choice of LLM is up to the user, (possibly even the choice of this API, I believe, having only skimmed the PR) while both of those choices are baked in to XCode, so they're not the same thing.
That's fair enough, but it's the opposite complaint, that XCode's LLM support is more limited because it is proprietary. That's a perfectly valid and reasonable objection, of course.
LSP != LLM
Gosh, it's almost like a proper IDE has synonymous features with LLMs
Ironically, you could probably vibe code your own.
Good luck getting just scroll bar right with vibe coding. You'll be surprised how much engineering is done to get that part work smoothly.
If enough examples are in-distribution, the model's scroll bar implementation will work just fine. (Eventually, after the human learns what to ask for and how to ask for it.)
Why wouldn't it?
Most programs today regularly have bugs with scrolling. Thus, an LLM will produce for you... A buggy piece of code.
LLMs are not Xerox machines. They can, in fact, produce better code than is in their training set.
That is funny for how much is wrong. Ask the LLMs to vibe code a text editor and you'll get a React app using Supabase. Engineering !== Token prediction
Do you really think so? Have you ever explored the source of something like:
https://github.com/JetBrains/intellij-community
Doesn't have to. The LLM will do it! We're done with code, aren't we?
Code is still there, but humans are done dealing with it. We're at a higher level of abstraction now. LLMs are like compilers, operating at a higher level. Nobody programs assembly language any more, much less machine language, even though the machine language is still down there in the end.
> Nobody programs assembly language
They certainly do, and I can't really follow the analogy you are building.
> We're at a higher level of abstraction now.
To me, an abstraction higher than a programming language would be natural language or some DSL that approximates it.
At the moment, I don't think most people using LLMs are reading paragraphs to maintain code. And LLMs aren't producing code in natural language.
That isn't abstraction over language, it is an abstraction over your computer use to make the code in language. If anything, you are abstracting yourself away.
Furthermore, if I am following you, you are basically saying, you have to make a call to a (free or paid) model to explain your code every time you want to alter it.
I don't know how insane that sounds to most people, but to me, it sounds bat-shit.
even nvim is getting native support for llms
It doesn't matter how they feel about LLMs, ignoring their battle hardened plugin system and going native would be bad architecture.
It’s just native support for ghost text. It’s not llm specific
You have to opt in and set up a language server
Is it? Link?
https://github.com/neovim/neovim/pull/33972
Of course it is, because that would be an aggressively stupid thing to do. Like boycotting syntax highlighting, spellckecking, VCS integration or a dozen other features that are th whole pint of IDEs.
If you don’t want to use LLM coding assistants – or if you can’t, or it’s not a technology suitable for your work – nobody cares. It’s totally fine. You don’t need to get performatively enraged about it.
Anything’s better than the current Xcode autocomplete.
My pet peeve is it will try to autocomplete any string you start typing with just random crap it thinks you might want in a string.
I think all autocomplete solution are crappy, no matter how sophisticated the AI. It is surprising how often the obvious choice is wrong, but it often just is. I deactivated it.
Generating some code is fine, but I now prefer the deterministic autocomplete for my types I have available in my current context.
This is great. I've been using Xcode with a separate terminal to run Claude Code, which has been a painful setup.
What was your problem with it? I see it running in a terminal more convenient (can point it to read local files outside of a project folder, for example)
Agreed. Claude Code is an amazing experience with Jetbrains IDEs, but for some reason Xcode just hates having claude directly edit the files.
How do you use it with Jetbrains? Junie? Or just as a separate CLI session?
They might be referring to the plugin https://plugins.jetbrains.com/plugin/27310-claude-code-beta-
I use VS Code with Claude Code, then I just use Xcode to build and launch
The annoying thing is the official Swift extension can sometimes flag errors in perfectly fine code with zero problem in Xcode. So I’m forced to live with persistent “errors” when editing in VS Code/Cursor.
I’m building my first iOS app ever so I know it has much more to do with me not understanding Xcode but getting builds to succeed after making changes with Claude code has been a nightmare. If you or anyone have any tips, guides, prayers, incantations for how to get changes in one to not clobber the other and leave me in xproj symlink hell I would be so grateful.
Same, only it's Zed for me and Claude Code in a terminal
You can use VSCode and XCode will automatically update when the files change.
if i could just get claude to properly remember it can directly edit the xcode project file, that'd be great.
for whatever reason it ignores my directive that it can from the CLAUDE file at least half the time. one time it even decided it needed to generate a fancy python script to do it. bizarre.
How so? I don't use xcode, but I much prefer having an agent in its own "app" so to speak.
Likely so it can auto suggest, directly edit code, integrate properly etc
The “Cursor for Xcode” startups just got Sherlocked…
Were there really such startups? It's so obviously a bad idea..
There's Alex (https://www.alexcodes.app), YC-backed.
Apple really should open it up to any model provider that has an “OpenAI-style API” by letting the user put in a base URL, api key, model id, and a few params like context limit as needed.
Xcode (26) already has that.
Weren’t the AI API’s converging? Why not let the users use whatever LLM they like.
Why would you limit users to Sonnet and not allow Opus when they are paying for their own account? I mean sure some people say Sonnet is good for coding but it seems needless to limit it in this way. Or they are just really slow to catch up… oh, right.
Another decade, another claim Apple’s behind and struggling to catch up
"Be ready for AIpple Revolution! We are making programming something different that hasn’t happen before! We are the first to introduce AI assisted agent coding with full integration with Siri, visionOS and so much more. New, holistic approach to creativity and efficiency"
>> Coding intelligence provides inconsistent results when modifying files that contain thousands of lines.
Under the known issues
This is true of all LLM agents. It’s a context window problem.
Sonnet only?
Does anybody know why Anthropic doesn't let you remove your payment info from your account, or how to get support from them?
I bought a Pro subscription, the send button on their dumb chatbot box is disabled for me (on Safari), and I still get "capacity constraints' limits. Filed a chargeback with my bank just because of the audacity of their post-purchase experience. ChatGPT-5 works good enough for coding too.
From my experience with Claude Opus it seems like it tries to be "too smart" and doesn't seem to keep up with the latest APIs. It suggested some code for a iOS/macOS project that was only valid on tvOS, and other gaffs.
The Pro plan ($20/mo?) is not and never was unlimited.
Does it have agent mode? Copilot for XCode has it and provides both GPT and Claude models, free or paid
They also upgraded the GPT-4.1 (actually a special Apple variant) to GPT-5 by default, with the option to use GPT-5-thinking, using your ChatGPT subscription. I don't know if it's a special Apple variant of GPT-5 but this is a big upgrade and more exciting than Sonnet 3.7.
I also wonder if it will have separate rate limits from ChatGPT (app/web) and Codex CLI (which currently has its own rate limits).
[dead]
[flagged]
I’m sensing just a touch of anger that is making your comment very non-objective.
> Apple can’t even ship software that works.
Hey they're doing better than anyone else.
Big disagree. My Macbook has hard crashed into restarting more times than my PC desktop has in years. The other day I literally just closed my M2 Macbook and it for some reason just completely crashed and shut itself off.
FWIW I abuse my M2 daily and only reboot once a year basically, and only because of some update.
Something could be wrong with your specific Macbook
I haven’t seen a crash in at least a decade of daily use.
Back in the older Intel days I did sometimes, perhaps due to the GPU switching and flakey Nvidia drivers. But even then it certainly wasn’t common.
Something is wrong with your laptop. That is absolutely not normal, and not something you should blame on the software design.
If you want me to blame "software design" then I can point to the fact that the TV app lags constantly on my M2 especially when I click the "download episode" button which should not ever be a thing on their native hardware, that there is a WindowsServer process that steadily eats up all the RAM that consistently needs to be quit in order to free up memory, that the headphone balance constantly resets to some random value where the audio comes mostly out of my left ear which has apparently been a problem for decades, that Spotlight is less than useless and STILL shows "Disk Utility" as the top result when I search "dis" instead of showing Discord an app that I open on an almost daily basis, that randomly the fingerprint login simply will not work for seconds at a time...
This is just the stuff I remember. This is a 16GB M2 Macbook Pro. Not a single native Apple made app or process should be lagging, let alone using up enough memory that sometimes apps just completely crash. Again, my windows desktop hasn't had a blue screen or a hard crash in the decade that I've used it and it had a worse configuration with 16GB of RAM and an old AMD CPU.
My hot take is all OSes are kinda bad for daily driving.
Apple has no qualms breaking backwards compatibility for core functions like bluetooth connectivity in MacOS. Windows has backwards compatibility, but increasingly worsening UX, throwing ads and subscriptions in your face before you can even log in, and a bad security/process isolation model. Desktop Linux is a case of "how many hours before I find out a critical part of my workflow is unsupported/bad/broken/unconfigurable/pain-to-configure in this particular distro/desktop environment".
I think they’re pretty amazing considering how hard a problem it is. Also, we forget how bad os’s used to be. They’re absolutely rock solid compared to the past.
I'm not trying to diminish the complexity of a desktop OS of course, but sometimes it's hard not to feel the priorities are all over the place. Don't get me wrong, I'm not nostalgic about Windows XP, I actually remember how many freezes and crashes I used to have back then.
My frustration is more born out of the OS rough edges constantly getting in the way of tasks I actually want to focus on and accomplish, which doesn't play well with my ADHD.
Man, agree to disagree. What you're describing is somehow still miles better than any experience I've ever had with a windows or linux latptop.
Exactly
I have been trying to make iOS/macOS apps for years, but god, every time I have a go at it, Apple's documentation regime is still hot garbage. Eons ago I gave up Windows development because of Microsoft's inconsistent and uncertain APIs, but MS had great documentation. Apple is the opposite.
The "best" way to get the "latest" details on Apple's APIs is to suffer through mind-numbingly vapid WWDC videos with their reverse uncanny valley presenters (where humans pretend to be robots) and keep your full attention on them to catch a fleeting glimpse of a single method or property that does what you were looking for. Even 1.5x/2x speed is torture. I tried to get AIs to sift through the transcripts of their videos, and may Skynet forgive me for this cruelty.
Then when you go try to use that API, oops it's been changed in the current beta and there's no further documentation on it except auto-generated headers.
They also removed bookmarks from Xcode's built-in documentation browser years ago, and it doesn't retain a memory of previously open tabs, and often seems to be behind the docs on their websites.
I wish they would just provide open-source sample apps of each type (document-based, single-window etc.) for each of their platforms that fully use the latest APIs. At least that would be easier to ask AIs on, since that is what they seem to be going for now anyway.
I pretty much had the same experience recently when I had to deal with their Screen Time APIs. Had to go through the wwdc videos because the documentation was lack lustre.
Can't each of these companies with IDE integrations slurp up the network traffic and distill Anthropic's models?
If you can listen to billions of tokens a day, you can basically capture all the magic.
Terms of service specifically prohibits this.
How much of the training set comes from websites with "no automated scraping" in their terms?
The companies stole that data from the world, so I don't see why we couldn't take it back.
It's a nice sentiment. The companies with the integrations are the ones that could take it back, but they don't have the incentive to break legal agreements and share with the world.
Meanwhile the creative output of humanity is distilled into black boxes to benefit those who can scrape it the most and burn the most power, but this impact is distributed amongst everyone, so again there's little incentive among those who could create (likely legal) change.
That is not how training works…
Apple.com advertising a Mac Mini:
> Built for Apple Intelligence.
> 16-core Neural Engine
These Xcode release notes:
> Claude in Xcode is now available in the Intelligence settings panel, allowing users to seamlessly add their existing paid Claude account to Xcode and start using Claude Sonnet 4
All that dedicated silicon taking up space on their SoC and yet you still have to input your credit card in order to use their IDE. Come on...
To run a model locally, they would need to release the weights to the public and their competitors. Those are flagship models.
They would also need to shrink them way down to even fit. And even then, generating tokens on an apple neural chip would be waaaaaay slower than an HTTP request to a monster GPU in the sky. Local llms in my experience are either painfully dumb or painfully slow.
Hence the "come on".
I bet Apple are working on it, it’s just not ready yet and they want to see how much people actually use it.
It’s the Apple way to screw the 3rd party and replace with their own thing once the ROI is proven (not a criticism, this is a good approach for any business where the capex is large…)
"Apple Intelligence", at least the part that's available to developers via the Foundation Models framework is a tiny ~3B model [0] with very limited context window. It's mainly good for simple things like tagging/classification and small snippets of text.
[0] https://github.com/fguzman82/apple-foundation-model-analysis
Yes, but the Foundation Model framework can seamlessly use Apple's much larger models via Private Cloud Compute or switch to ChatGPT.
When macOS 26 is officially announced on September 9, I expect Apple to announce support for Anthropic and Google models.
Local models and any OpenAI-compatible APIs are available to the Xcode Beta assistant. This is just a dedicated “sign in with x” rather than manual configuration.
Trust me, you wouldn’t want to use a model for agentic code editing that could fit on a Mac mini at this stage.
A 128GB Mac Mini M5 would be sweet.
Wow they're finally getting it. The AI breakthrough will not come from procedural generation of memojis - but rather enabling developers to use your platform. But with the nearly hostile stance of your 30% take, we will see how far this goes.
What’s the ‘30% take’?
30% of all AppStore sales go right to Apple
15% if you’re part of the small business program.
What program do I have to join for 0%?
Being in the EU and releasing in an alternative marketplace.
The one where you create your own mobile operating system.
The one where you collect cash directly from users, and magically make handling that have zero overhead.
Credit card processing is hard... Go price out stripe + customer service + dealing with charge backs and tell me if you really want to do processing your self.
Well seeing that the most popular apps aside from games don’t have in app purchases and another subset of that has means to do payments subscriptions outside of the App Store, the 30% (actually 15% for small developers) is a boogeymen