Apple turns to Anthropic to speed up coding & fix buggy tools
Apple is bringing in Anthropic's Claude AI to help its engineers code faster, marking a quiet but telling shift in its approach to artificial intelligence.
Xcode
Apple has long resisted leaning on outside help for core technologies, preferring to build its tools in-house. But that wall is starting to crack. The company is now partnering with Anthropic to integrate its Claude Sonnet model into an upgraded version of Xcode, Apple's software development platform.
New tool could reshape developer workflows
Internally, the project is being rolled out to help Apple engineers write, edit, and test code more efficiently. The collaboration signals a strategic shift, reports Bloomberg on Friday.
Apple is no longer treating AI as a self-contained project siloed within its walls. Instead, it's beginning to view outside partnerships as a necessary part of keeping up with rapid advancements in generative AI.
Apple previously developed its own AI coding assistant, Swift Assist, but internal complaints about hallucinations and slow performance stalled its release. Swift Assist already works alongside third-party tools like GitHub Copilot and ChatGPT, both of which have been integrated into Xcode to offer additional AI-driven support for developers.
However, Swift Assist still hasn't been fully released to the public after its announcement at WWDC 2024.
Apple is laying the groundwork for a more open and flexible AI strategy
Claude's integration adds natural language capabilities that let engineers request code snippets, debug interfaces, or streamline test cycles. While the tool is still limited to internal use, a wider release to developers is on the table if the internal rollout goes well.
A strategic shift, not just a feature drop
The shift also aligns with broader internal changes. Apple has recently restructured its AI leadership, moving Siri development under software chief Craig Federighi and narrowing the role of AI chief John Giannandrea to focus more on foundational research.
CEO Tim Cook addressed the new direction on Apple's latest earnings call, telling analysts the company is excited about its AI roadmap. He emphasized that Apple will continue to build some of its own models but said partnerships will play a role.
That's a different tone from just a couple of years ago, when Apple seemed caught off guard by the AI boom. It launched Apple Intelligence with on-device features like custom emoji and writing tools, but those efforts were often seen as lagging behind what rivals like OpenAI and Google had already delivered.
Now, with Anthropic in the mix and WWDC approaching on June 9, Apple is laying the groundwork for a more open and flexible AI strategy. The company that once insisted on going it alone is beginning to see the value in collaboration.
Read on AppleInsider
Comments
It shows that Apple is really behind in AI.
RIP Apple.
Yeah, "yawn". Apple's going to be stealing other people's work with everyone else and you meet it with a "yawn".
Apple not only survived that switch, but thrived. And eventually developed their own CPU that now completely beats Intel.
People don’t realize AI is a long game and right now, even if Apple
os behind, they will skip certain processes that are either not sustainable, fads or certain avenues which are dead ends. They will benefit from the research of the big AI companies, just as others and Apple competitors often benefit from Apple’s research.
The Wall Street mafia, tech pundits (🙄) and others kept vomiting that Apple will miss the 5G train and is behind. Now, Apple has it and MWave, which is STILL largely limited to certain areas, and they’re making their own 5G chips!
In the near future It will be fun to see the actual comparisons when you turn off the internet connection and run Robbie the AI agent and who can answer or perform that test will be similar to unplugging a MacBook Pro from the wall and doing the same with a Microsoft PC laptop, and seeing which one actually still performs at full speed and also seeing which one has longer battery life under normal usage, Apples mission hardware, software and privacy is a much harder/longer way to go than their so-called competition.
Cars aren't taking life saving decisions by sending them to the cloud and back, for example. The sensing, interpretation and result are done on-device (the car).
Interpreting responses based on communication with traffic lights is done through edge processing combined with on device processing. There is no need for that to be done in the cloud.
Non-time/privacy critical processing can be done in the cloud.
The above also applies to handsets. Things are done where they are best done depending on multiple factors.
As with most things today (in the digital age) losing your internet connection will render most devices virtually useless and that includes iPhones with even with on-device processing.
Bill Gates might disagree but the more memory you have available, the better. Anything with a limited amount of memory will hit a wall sooner or later.
On device processing is a good goal/option but everyone knows this and is working on it. Cloud and edge processing are also key, though.
Just for clarity, I use the term 'edge' to refer to the ICT concept of near-to-device processing without needing a central server.
It can also mean on-device processing but I prefer to say 'on-device' in those cases.
So all the talk about Apple being "behind" in LLM based AI is mostly blather since the LLM part is largely inconsequential versus the database being used and whether it's legal.
We need to accept that. There is enough evidence of that now.
They didn't have anything when Gen AI came to market. They deliberately avoided any mention of 'AI' at the WWDC when people were expecting to hear about it and deliberately chose to refer to ML instead. The following year (and now unable to avoid the term due to all the progress made by Gen AI) we got 'Apple Intelligence' but zero of that shipped on the 2024 AI iPhone. It was to come out over the release cycle of iOS.
Almost all previous phones were not elegible for Apple Intelligence (another HUGE sign that AI of this kind just wasn't on the front burner until it was far too late). Then we got news of the delay to the enhanced Siri and then news of the AI executive shakeup. Then we got rumours of Apple trying to create its own AI training hardware but together with reports saying it was buying Nvidia hardware. If they truly weren't behind they would have had at least a plan to have everything in place long before now.
Nothing in all of this points to Apple being even remotely catching up with the almost daily upgrades to swatches of LLMs and Tiny LLMs.
No doubt it is desperately doing all it can but that doesn't mean it's not behind.
LLMs are anything but inconsequential as without them the data they feed off is just that - data.
Yes. There are ongoing debates and even challenges as to the ethical and legal aspects and they may prove consequential down the line but that has nothing to do with what is (and has been) available today. The current state of play.
This year's WWDC is going to be interesting for many reasons and AI is going to be one of them.
I'm not a fan of Gruber but he did ruffle some feathers and point out some uncomfortable Apple Intelligence truths.
I use Perplexity Pro which has proven to be ultra reliable for my needs. If Siri ever reaches that level it would be a true milestone.
Putting only six gigs of memory in the Pro iPhones because they could came back to bite Apple when the AI rush came up (being vertical didn’t help them there), however Google and Qualcomm problem is their SOC being under powered and 5-6 generations (performance wise) back even though they both use eight cores to Apples, six cores SOC processor.
Google‘s problem is that their SOC is six generations behind Apple and their hardware can’t operate on the edge which is why Google and Samsung have to phone home which Google probably doesn’t mind doing so since their business is collecting info probably a win-win situation for them.
https://browser.geekbench.com/mobile-benchmarks
Apple wants to use AI for coding - this has nothing directly to do with user facing AI they are developing for iOS etc. So of course they would look to outside AI tools to embed in Xcode. So trying to conflate that decision with some sort of change of AI focus is a stretch.
I never thought we would get to the point of preferring multi-platform frameworks and tools to doing native development for iOS and Android respectively, but in our case most projects avoid working directly with Xcode
It was actually impossible, but Steve Jobs and a team of superstar engineers made it work.
And now... I know that I could just drop DeepSeek into one of these massive Apple data centers, tune it, and run it, and be better than Claude at whatever they're doing with claude. It's shocking because it means Apple engineering is so bad these days, they can't even take an open source model and run it on their own hardware. It's crazy.
Apple has a massive management problem.
Some pointy haired boss picked up the phone and called Claude to help out, no doubt sending a few 100 million over there in the process - and that's the best Apple can do?
That's OK for a non-tech company but for a tech company it's simply terrible.
Apple never got around to fix the XCode mess.
XCode is so bad there used to be a blog called "XCode said" which was just making fun of absurd XCode errors and behaviors.
Look at your project settings and you see a thousand avon b7 said:
Apple being behind in AI is not really the big problem - check out xAI and how fast they came from 0 to competing with the best models for the crown!
It took them less than a year!
Apple doesn't have anyone like Elon who has the vision, attracts the people, and executes - Apple is run by a bunch of pensioners whose only reason for still being there is they made a promise to their friend Steve Jobs on his deathbed.
It worked for a while but Apple desperately needs to find its vision - and dump the vision Pro, one of the most laughable, most expensive, C-suite nonsense ideas I've ever seen. The amount of research that has gone into this totally useless thing is crazy - talent that could be used to make a car, or figure out AI, things that actually matter. They have no plan.
The same goes with any supposed LLM by Apple; they are still as bad with AI as ever. You cannot replace decades of research and evolution and make people believe you 're going to provide such great, groundbreaking, revolutionary AI models with nothing short of the greatest AI experience ever available, a leapfrog from those stupid competitors and the many years of advance they have.
*sight* when I read how impossible it is for Apple to be dealing in shady and despicable behaviors in Africa... even if it's not a first, genocides aren't an issue either. Moreover, this is already clearly ignoring how loathsome this company is and the many other unacceptable ways this company resorts to threatening providers. Smaller ones are sometimes acquired (there is not a single part that is not of external origin), but if possible, they prefer to poach employees. With crucial other partners, they never hesitate to get a lower price by pressuring them. Those companies doing research, spending billions in manufacturing their products, and building the telephones with close to no Apple parts... Even TSMC, at some point, was producing the monstrous M SoCs, a big one-piece thingy... with yields of 55%, which is very low; it means half the wafers are trashed—which is ludicrous, but Apple pays as if the yields were 70%. Great partner.
And as always, great company, so sincere, spontaneous, and friendly.