Sky Preview
Introducing Sky for Mac.
[…]
Sky floats over whatever you’re doing so you can:
- Ask questions from anywhere on your Mac
- Take action in your apps (send a message, schedule an event, etc)
- Use your own custom tools by adding prompts, scripts, shortcuts, or MCPs
For the past two weeks, I’ve been able to use Sky, the new app from the people behind Shortcuts who left Apple two years ago. As soon as I saw a demo, I felt the same way I did about Editorial, Workflow, and Shortcuts: I knew Sky was going to fundamentally change how I think about my macOS workflow and the role of automation in my everyday tasks.
Only this time, because of AI and LLMs, Sky is more intuitive than all those apps and requires a different approach, as I will explain in this exclusive preview story ahead of a full review of the app later this year.
[…]
Sky is an AI-powered assistant that can perform actions and answer questions for any window and any app open on your Mac. On the surface, it may look like any other launcher or LLM with a desktop app: you press a hotkey, and a tiny floating UI comes up.
[…]
What sets Sky apart from anything I’ve tried or seen on macOS to date is that it uses LLMs to understand which windows are open on your Mac, what’s inside them, and what actions you can perform based on those apps’ contents.
Pressing ⌘⌘ to grab your current context is a delightfully natural interaction.
Sky saves the current window or file as well as its metadata, so you can ask AI about it right away.
[…]
To go further, you can add Custom Tools – which can include custom instructions, MCPs, AppleScripts, & shell scripts – and yes, Shortcuts!
You can extend Sky’s capabilities however you want – and designing them is easy with prompting built right into the editor interface.
This feels like the so-far-unfulfilled promise of Apple Intelligence — but more. The ways I want to automate iOS are limited. But the kinds of things I want help with on my Mac are boundless.
As with Grammarly, it’s amazing that they seem to be doing more than what Apple promised, yet without requiring the apps to rearchitect everything around intents.
The Apple Intelligence team meeting after seeing Sky, after very publicly failing to ship their own version of this stuff.
The real question is why couldn’t the founders of Shortcuts build this stuff at Apple, and what were the systemic failures that pushed them out to go it alone.
This week on @appstories, I share my first impressions of Sky and we share our wishes for Shortcuts and Apple Intelligence.
Previously:
- The Automation Gap
- Model Context Protocol (MCP) Tools for Mac
- Grammarly Raises $1 Billion
- Apple Delays “More Personalized Siri” Apple Intelligence Features
- Privacy of Windows Copilot+ Recall
Update (2025-06-04): Rui Carmo (Hacker News):
I mean, people have free will and all, and can choose to work wherever they want, but this makes my earlier rant about their having neglected automation feel like the first clue to a corporate culture murder scene.
Not having made it possible for them to thrive feels like vanilla corporate politics, but having brilliant people leave Apple and ship something that is, even in preview, much better than anything that Apple Intelligence promised (including the made up bits they paraded as marketing material) is just gross mismanagement (now you know why I held back on this draft).
[…]
I can see Apple balking at doing something like Sky (if they ever even considered it) because it not only has to share bits of your screen with an LLM, but also because it would have to open up the Mac to third-party automation in a way that it has never done before, and that would be a huge departure from their current approach.
[…]
But the privacy angle is interesting, because Apple was in a perfect position to do something exactly like Sky and ensure that it was done in a way that respected user privacy. Even though local models are still not quite there yet (remember that RAM requirements are still very high as far as running truly useful models are concerned), they do have the confidential computing tech to run inference in a privacy-preserving way–which might be the only bit of Apple Intelligence that actually works at this point.
3 Comments RSS · Twitter · Mastodon
Colour me confused about the hype surrounding AI and MCP as applied to automation. The point of automation, as I see it, is to chain together actions in a linear, predictable way, which is antithetical to the way LLMs work.
I have written many scripts, macros, services, and shortcuts to do things I could very well do by hand — and am willing to bet that, in a few cases, I will never recoup the time it took me to write them in the first place. The point of these scripts, however, is to ensure 100% consistency no matter how tired, distracted, or busy I might be.
Automation is what allows me to rename, file, tag, and generally process hundreds of files across decades with the certainty that there will be no typos in the file names, no mistakes in the dates, and no confusion as to what the correct folder is to store the files.
Asking an LLM to “sort the files in my Downloads folder” is a very impressive trick, yes, and may be useful in some contexts, but that is hardly what automation is about.
Whether one likes or dislikes LLMs or believes them to have creative potential, their application to automation seems to be a solution in search of a problem. Even asking LLMs to write code is a different matter, because one can, at least, check out the code they output and, if satisfied, run the exact same code over and over again. But direct interaction with an LLM? That’s not automation in my book.
@Tarsier Yes the biggest automation thing is that you want things done *in a specific way*, the journey is as important as the destination.
WRT the 2025-06-04 update; company founders leaving Apple the nanosecond their acquisition contract times-out is pretty much the norm. Get your stopwatches on the crew from Photomator.
