If iCloud Becomes iClaw at Next Month's WWDC?

With less than a month until Apple's Worldwide Developers Conference (WWDC), Bloomberg has once again released new information about Apple's new system:

Apple is preparing a small adjustment to the interface of macOS 27, further refining the "Liquid Glass" design language's visual presentation.

However, the problem is that for today's Macs, besides the UI that still needs polishing, the most urgent updates are obviously far more than that.

Liquid Glass, Another Year of Patching

Compared to iPhone and iPad, Macs have more performance and battery life. In fact, the interface of macOS 26 is visually closer to the "full-powered" Liquid Glass demonstrated at last year's WWDC.

However, Liquid Glass was originally designed specifically for OLED screens, while all current Mac products use LCD screens. As a result, the effect of semi-transparency, shadows, and glass texture is not as good on LCDs.

Therefore, some high transparency effects and shadows in macOS 26 can reduce the readability of lists and text — this has been a long-standing issue with "Liquid Glass."

Like iOS 27, macOS 27 will also refine Liquid Glass, making it closer to what Apple initially envisioned: balancing transparency and readability, while further optimizing energy consumption.

However, the issues with macOS 26's UI go beyond just Liquid Glass itself. Misaligned rounded corners, numerous distracting small icons, and application icons that have lost much of their recognizability after redesign have affected both readability and aesthetics.

Source: Daring Fireball

Apple commentator John Gruber made a sharp comment on macOS Liquid Glass: as a "content-first" design language, Liquid Glass makes the system UI hide behind media. It may work on iPhone, but as a desktop platform emphasizing productivity rather than content consumption, Mac contains many windows and groups, so it is more complex. Therefore, application interfaces need to ensure clear structure, distinct functional areas, and strong visual recognition.

After Stephen Lemay took over as design director, this veteran who has worked at Apple for nearly 30 years has raised expectations — Lemay is known for his high internal reputation and stable performance, and he might be the person within Apple who understands the company's system interface best.

Under his guidance, how macOS 27 and iOS 27 will turn around the mixed opinions on Liquid Glass and return to a direction that balances practicality and aesthetics is indeed worth looking forward to.

But for macOS, "correcting the chaos" on the interface is necessary, yet it is no longer the most important update.

For Apple, future system updates have two main lines: one is to optimize system stability, and the other is to prepare for Apple Intelligence.

The Best AI Carrier Needs an AIOS

According to Bloomberg's report, Apple plans to create an "Extensions" feature for "Apple Intelligence," allowing users to replace third-party AI models such as Google Gemini and Claude.

Siri will not only be integrated into applications like email, messages, and photo albums, but will also transform into a chatbot, becoming a standalone app. More AI features will cover text and image generation and editing tasks.

However, these updates are, to be honest, more about single-point AI functions rather than system-level orchestration capabilities, and they do not fully leverage the advantages of Mac hardware.

The lobster craze earlier this year brought the Mac mini, a product that had already gained popularity the previous year, back into the spotlight. This time, it became so popular that Apple itself ran out of stock, with the "entry-level" version completely sold out on the official website.

Mac and Windows have their own strengths and weaknesses in many aspects, but when it comes to AI, the claim that Mac is the "best AI container" is almost undisputed.

This issue has been discussed in detail in an article by iFANR. In short, because whether it's the UNIX system underneath or the hardware architecture integrating memory, Mac is very suitable for running AI agents and large models. And due to the characteristics of the ARM architecture, it runs with low power consumption and is quiet, making it ideal for AI to run continuously.

This seems more like a case of "planting willows without intending to grow them." Apple didn't initially build its Mac around AI, but it accidentally completed all the technical preparations for AI. Strictly speaking, this is a "compatibility advantage."

From this perspective, even if macOS does nothing, the Mac itself is already a good AI platform. Apple could follow the App Store logic, letting users deploy the third-party AI agents they want, and continue to play the role of "collecting tolls."

This is indeed Apple's long-term approach: when mobile internet emerged, Apple didn't need to build its own search engine or e-commerce platform. In the AI era, people's needs vary widely; some need an agent that can edit, while others need an AI for research. It must rely on third parties to meet these needs.

At Apple's May earnings call this year, Apple specifically mentioned Perplexity's intelligent agent product, Personal Computer, and considered it a great utilization of the Mac platform's capabilities.

If you think someone else is doing it well, why not make your own "iClaw"?

Although the diversity of third-party AI is good, it doesn't conflict with Apple making its own. And there are many things that only first-party can do well and make people feel safe.

Even the strongest third-party apps find it difficult to naturally obtain system-level context. Apple cannot open the most basic permissions, and only the system itself knows everything about file locations, window states, and local personal data. Often, the experience of AI apps gets stuck at these permission boundaries.

In fact, Apple has had this idea before. The long-awaited AI Siri actually had a similar concept, which could read users' text and application windows, and perform retrieval and processing across applications.

Compared to iPhones and smartphones, the mainstream usage scenarios for AI apps are actually on the desktop. That's why Mac has become the hottest AI hardware this year, but Apple hasn't given Mac enough native AI capabilities at the system level.

The Windows camp is much more aggressive in this regard, with system-level AI features like "Recall" and "Copilot." OEM manufacturers like Lenovo and Honor have even prepared pre-installed lobster apps, reducing barriers and saving tokens due to deep integration with the local environment.

Microsoft itself has also been unable to sit still. According to reports, it is transforming Copilot, which used to only support question-and-answer interactions, into a 24/7 online digital avatar, achieving lobster-like capabilities.

Compared to OpenAI, Anthropic, or Google, I honestly prefer to entrust these sensitive data to Apple, which is more committed to privacy protection.

Going deeper, macOS lacks not AI apps, but the "infrastructure" of the AI era. Mac has already prepared fertile ground for AI, but macOS hasn't yet become a true "AI system."

Apple can not only develop its own AI agent capabilities but also needs to reorganize models, permissions, context, automation, and cross-application tasks, making the system a native intermediary for AI workflows and a central hub controlling all AI.

Like the "personal knowledge base" required for agent operation, we can currently build it using folders, but it's not convenient enough.

Apple can take on this part itself. Users can build and generate a "knowledge base" file using Mac's built-in tools, which can be tied to an Apple ID and synchronized via iCloud. This way, regardless of which intelligent agent service is used, it can quickly access the user's knowledge base without starting from scratch, while ensuring that the content is protected by Apple's privacy policy.

Moreover, these configuration modules can be integrated into Apple's subscription system, with iClaw and Tokens becoming value-added services provided by Apple in the AI era.

iClaw diagram, AI-generated

In fact, Apple has already started this process. In macOS 26.1, Apple integrated the "Model Context Protocol," a universal open standard for different AI systems. Agents can access users' personal data through this protocol; Apple's basic model framework allows macOS developers to call the system's built-in basic models, with zero network latency, zero API costs, and data staying on the device.

As the progenitor of computer graphics systems, macOS has been a desktop system built around "applications" for decades.

In the next decade, applications and graphical interfaces will still be the mainstream of human-computer interaction, so it is certainly important for macOS 27 to improve its interface style.

However, in the next fifty years or even further, AI will inevitably become the main theme. macOS will inevitably be further transformed into an AIOS centered around "tasks."