The Automation Gap

I really tried to get into Apple's automation solutions - and built some significant edifices of mass file processing for publishing workflows with Automator and Folder Actions - half a dozen stages processing hundreds of files at a time, moving from folder to folder doing different things at each stage.

Automator was a great app to work with; automation made by publishing and file workflow people - super empowering for someone who knew NOTHING about programming, because it was all about plumbing - pipe files rom here to there, every function and task named in plain english, with icons & thumbnails to illustrate effects, drag & drop of units of functionality. Every task had droplists for its modifier options; you didn't need to KNOW anything off by heart, because everything that it could do was revealed by the UI. It's a universal hierarchical menubar UI mindset.

Then Shortcuts came on the scene. Shortcuts is automation made by *programming* people. Everything about it presumes an understanding of the concepts of programming. Nothing is revealed by the UI. One cannot learn the full scope of the application's capabilities, just by looking at the bits of the UI. It requires rote knowledge of unrevealed, and undiscoverable knowledge. It's a commandline UI mindset.

Realising that relying on an in-house Apple automation solution would leave me forever a victim of Apple's whims, that the politics of who is ascendent within the organisation and what technologies and external purchases they champion will have real serious effects on my ability to get things done, I made a decision to abandon wherever possible Apple solutions for doing things on the Mac, in favour of third party alternatives. There is no first party advantage in using an Apple-made solution, when Apple solutions are subject to corporate politics, and short attention spans.

So now, all my automation is done with Hazel and Keyboard Maestro. I know both these companies have a vested interest in concentrating on keeping these products good. They're not just shovelware to get me to buy a new computer when the old one has been made slow by inefficent software *cough* Catslyst & Swift/UI *cough*. They have to be good in isolation, for their own sake, or their developers go out of business. That's the sort of Cortez-burning-his-ships motivation a product needs to be good, *and stay good*.

I don't use Apple's Podcasts app to subscribe to Podcasts. I use Doughnut to download shows from RSS feeds, and then Hazel looks at each folder Doughnut creates, and copies (with APFS zero extra disk space copying) the files to corresponding folders in my master podcast archive where they're effectively airgapped from Doughnut, and then if I want to listen to a show on my iPhone, I drag it from Finder to my iPhone in iMazing.

Cutting Apple out of he loop is the only way to make your system safe from Apple's corporate politics. That corporate politics is why *Apple* automation on the Mac is a dumpsterfire. That's also why Smart Folders is moribund, despite having the potential to have been one of the best, most transformative technologies Apple ever added to Finder - someone got politicked out of the way before a proper deeply capable UI (think Aperture's Smart Albums creator UI) could be made a core part of Finder, so we have this half-hearted implementation that's only a shadow of the potential or it, because someone had a fantasy of talking to their computer and getting Siri to do all the things Smart Folders and saved searches could do.

We literally live in the dumbest macOS timeline these days.

If the day comes when Keyboard Maestro ceases development or is otherwise constricted in its functionality by whatever MacOS security theater Apple dreams up next, that's the day I just give up and try Linux. I still run Automation and AppleScript scripts, and think highly of both programs, but it's extremely clear no one is left at Apple who understands the value of robust automation and that this company no longer has a prayer when it comes to creating the kind of software it used to make. At least we have cute hair jokes though, tee hee.

Shortcuts might be my least favorite piece of Apple software of all time. It achieves the trifecta of (1) being needlessly hard to use, (2) bafflingly less powerful, even on a Mac, than all of its decades-old forebears and (3) unreliable and prone to random timeouts and performance issues without any insight into why. Keyboard Maestro provides the kitchen sink, and AppleScript throws you into an empty window, and both are still somehow so much easier to grasp than Shortcuts, which does its clumsy interface no favor by frequently organizing and naming its actions in ways that defy all logic. I'd love to do more automation on iOS, but I find Shortcuts such a nauseating program to use that I just don't bother most of the time.

I hate Shortcuts with a passion.

I have a simple shortcut on my phone, which I run via an icon on the main screen, that gets the info of the track currently playing in Music, concatenates the artist with the song title, then opens a web view with a YouTube search for the resulting string.

I've had it for years, and for a good while it worked flawlessly. Then at some point, after a major os update, it started randomly failing with cryptic error messages (no description, just negative numbers as error codes), but not frequently enough to become a nuisance.

Enter iOS 18 (up to and including 18.5) and now it fails to run two out of three attempts! Nothing changes (except the launch time, of course), yet it fails the first two times and runs successfully the third time, every. fricking. time!

The error message says "The operation couldn't be completed. (MPRequestErrorDomain error 1.)"

I can't believe such incompetence! If my code would fail this often, my employer would throw my ass on the street so fast that my head would spin...

> And why in 2025 is drag and drop of actions so hopelessly janky?

Because the Shortcuts app is written in SwiftUI. That's really all there is to it. And SwiftUI just isn't good enough.
---

In typical Apple fashion the idea seemed to be to replace Automator with a less-capable more "secure" Shortcuts app.

Eventually they'll just get rid of Automator and all you'll have is Shortcuts (well at least that seems to be the plan). But for now you can use both...enable Shortcuts with Applescripts, shell script etc.. (not sure what the point of all this overlapping behavior is). But like Objective-C the idea of just replacing the old with the new is complicated, because the old is still the foundation for a lot of the new and it seems that they bit off more than they can chew.

I think if/when Automator goes away there will be lots of functionality that doesn't make it into Shortcuts...and that functionality will be lost (perhaps forever..) unless there is a change in management.

Recall, everyone, that Shortcuts started out on iOS — and it was using app intents back then and was like Automator-lite-for-iPhone.

It was impressive at the time… a long time ago… and I think will be more useful as AI programming aides come online. But only if Apple starts adding hooks to more apps.

Also, re: creating programs:

I personally would love to see an update to the UI so that I can type in-line what I want and the specific modules appear as an auto-complete. Just like Automator and Applescript, it’s quite hard to figure out what can be done, and going to a side panel and then drag and dropping seems like the slowest-way of doing it.

Also, multi-select.

(I bet it’s also a huge security-risk vector (just like AI will be), hence why Apple may not tout it too much or is held back by privacy/security issues)

Shortcuts is a double whammy of shit technologies; it’s a Catalyst app that uses SwiftUI as its UI framework. 🤦‍♂️

I wish Apple would abandon all these other technologies and refocus on AppleScript. I hate AppleScript, and writing it has always felt like trying to paint Starry Night with a stick and a bucket of mud, but now we have LLMs, which make obscure programming languages accessible to everybody. This feels like a much better, more solid approach. And bring that to iOS as well. There's no need to invent all this new stuff.

It would even make Apple's AI strategy look better, because they'd have at least one genuinely helpful thing to present.

Amusingly enough, Gruber is wrong on the `title` thing.

Yes, there are multiple title tags in the markup, but 39 out of 40 aren't in the HTML per se, but inside SVG!

So what this actually is isn't just a weakness of using Shortcuts to parse HTML, but also on top a classic case of developer assumptions gone wrong: a title tag inside text/html isn't necessarily an HTML title tag.

Applescript: Cursed language. Used only in desperation. Everyone mind-wipes themselves of AS immediately after their task is done.

Automator: Decent, but not as general-purpose by itself. It was okay if you can think of your task as a transformation pipeline. It's better if you write your own actions.

Shortcuts: I don't get it. Like Automator, but iOS-first, and more restrictive? Kludge together pages of control flow blocks instead of typing a few lines of loops and statements? I know 0 people who use it and I don't.

--

Automation remains limited for one reason: If you require significant dev effort to support it, you blew it! Apple failed to learn this. Each new generation of Apple's automation requires more developer effort:

Applescript: Requires scripting definitions and other annoying work.
Automator: Requires AMBundleActions and other annoying work.
Shortcuts: Requires App Intents (infecting your code base with Swift) and other annoying work. App Intents really blew it considering how late in the game it is and how many apps are Electron.

Each new generation of Apple's automation UX also restricts users more than the last. Applescript sucks balls, but a user language (with tooling for the user to produce it) is the right idea. That user language is just not Applescript, not Javascript, and certainly not Swift (or related).

@Plume Yes, I think refocusing on AppleScript would be good. Or perhaps if they really wanted to switch to JavaScript and take it seriously.

@Sören The point of Gruber’s post is that he knows he was wrong to assume one title tag but that this was hard to debug with Shortcuts. HTML vs. SVG is an irrelevant, though perhaps interesting, detail.

@Hammer Yes, each generation has required more developer work and been more restrictive.

I still find myself using AppleScript most of the time. I hardly ever touch Automator or Shortcuts, because if I'm going to automate something, I want to do it in code, even if it's a cursed language. (And I take AppleScript particularly personally after learning to program in HyperCard, since it's basically a degenerate version of HyperTalk!)

It's interesting hearing everyone dumping on Shortcuts. I was curious that I might have been missing something by not using it, but now I won't bother.

One of the things that I hate about modern macOS is how often AppleScript fails now because of some aspect of security theater that Apple added recently. And it gets worse with each release. It was bad enough that 10.14 required granting individual permissions for each application A that wants to automate application B, without providing any sort of good interface for managing it, or granting things like blanket permission!

I pine for the days of Mac OS X where nearly every app supported AppleScript and there were no restrictions on what it could do. Even though AppleScript sucks as a language.

@ Michael I realize that wasn’t his main point, but my read of Gruber’s post was that lots of web sites incorrectly (or needlessly) have multiple title tags, when, surprisingly, the usage is correct and makes sense

On the eve of WWDC, here’s a video of an agent / framework that someone wrote that does stuff across apps on the iPhone. It’s not shippable in this state (requires dev tools and access) but I found it pretty interesting as a concrete example.

It looks like it uses the accessibility hierarchy to read the screen and interact with the apps. So not MCP, but more like a human assistant fiddling with your phone for you.

https://www.youtube.com/watch?v=4rnv6dN-2Lg

(written up more here: https://9to5mac.com/2025/06/01/heres-how-an-iphone-ai-agent-could-change-how-we-complete-tasks/)

A couple thoughts:
0. Apple’s surely thought of this already. Why isn’t it doing this? I can guess plenty of reasons.
1. Could be/would be a privacy nightmare — this agent would have access to everything?
2. I like that it shows it doing things rather than saying “I did it” and you have the go back and confirm
3. ChatGPT used to hallucinate quite a bit a few years ago when I tried pulling data out of a page… perhaps it’s better now? For me / grandma/grandpa, it would have to be basically 100% accurate. “Hey Siri, transfer $1000 to tommy from my bank account”
4. I would find something like this handy if I could be truly multi-modal and use voice as an additional finger, so to speak. E.g. I enter a text chat and say “Put my flight info here” and it pulls that info from an email app and pastes it… I’d still do things manually but I’d use voice to speed some manual sub-tasks. Or perhaps, press and hold on a thing and tell phone what to do with it.

If the rumors are to be believed, Shortcuts + some AI program writer… that sounds interesting. I personally think the iPhone is going to make ‘programming’ more available to the masses via something basically like this.

Oops, to clarify: I think Shortcuts + some AI programmer agent to write scripts will be the a big part of what Apple does.

*Not* the ‘AI can see and control your screen and does things while you watch’ thing. At least not in the near future. Too many risks.