I've just used this extensively to build 200 Shortcuts for my event-based automation app on macOS [0], because some actions you simply can't do without Shortcuts: changing Focus Mode, toggling Accessibility functions like Color Filters, accessing the Private Cloud Compute model etc.
I also wrote about how Claude was able to basically learn the language from scratch and write those fully compilable Shortcuts for me [1] because it was mind boggling to me that an LLM can do that. Curiously, this is becoming more and more normal in my mind.
When you say Claude learned it. That's in the current context window it is able to do that, right? Or is there a more permanent way to make it learn something?
Cool to hear Claude was able to learn it. I was planning on leveraging it in a future version of this project I was hacking on that lets you execute shortcut actions as tools (without creating actual shortcuts): https://tarq.net/posts/action-relay-shortcut-actions-mcp/
Well, that’s a domain that has caught my attention so I’ll give this more weight (ltg). I recall novel Mac apps that weren’t quite right for me but seemed thoughtful.
AppleScript was just a little weird but I could get my head around it. Shortcuts just doesn’t make sense. Even the simplest things are hard to do and the scripts are totally unmaintainable. I don’t know why Apple is doing this.
I felt really smart after I made a fancy Shortcut that did complex playlist generation based on rules and whatnots.
Of course, adding music to a playlist broke a couple of updates down the line and, as far as I'm aware, still doesn't work properly several years later.
(I moved to Marvis Pro[0] because it has reasonably complex smart playlists that just about mimic what I was doing with my generator except they're transient and not saved as mine were. Win some, lose some.)
[0] No link, sorry, because it'd either be iOS App Store or ad-laden bloat sites "reviewing" it. https://www.reddit.com/r/MarvisApp/ might be worth a read though.
This is what prevents me from putting any effort in to shortcuts. I have zero confidence in Apple respecting my time by committing to compatibility or even just not breaking things. The constant feature churn from new PM’s tearing down Chesterton’s fence to make way for their own career ambitions makes Apple user hostile.
Apple doesn’t respect its users with constant changes to iTunes and then Music. My wife is still mad that after redoing her 12,900 songs, and correcting Meta Data, Apple Music said your wrong and changed at least 10% of her music in some way.
I take it this only supports Apple’s built-in actions, and doesn’t plug into the broader AppIntents system? AppIntents includes a packaging concept, would be cool to see if this could use third-party AppIntents in a similar way to how scientific Python uses C modules for performance critical sections.
There are some things that are only available in Shortcuts because Apple gave the app entitlements to communicate with parts of the system that an AppleScript or other apps can't. Things like setting/getting the Focus Mode, changing some system settings like Airdrop Receiving, Color Filters, Background Sounds etc.
Also some apps export Shortcut actions that can run in-app code: for example my Lunar app has an action that can help fixing arrangement when monitors flip around [1]
It's much easier to implement a struct for a Shortcut, than exporting AppleScript sdef files or creating IPC command-line tools, so a lot of apps take this route for code that needs access to the memory of the running app.
Thank you for the kind words! Love to hear from people that were helped by my work!
That was also my pain point with Lunar, working on a small balcony in a small apartment where the light from the window was constantly changing and the monitor always being way too bright or way too dim.
I broke one of those LG monitor joystick OSD buttons before I got to building Lunar.
Apple designed Shortcuts to look like a shiny toy demo, and it gets awkward fast once you leave the happy path. AppleScript is old, weird, and still wired into a lot of Mac apps, but the moment it trips over something basic like text encoding you can lose an hour digging through forum posts from 2004.
With Hammerspoon, you get Lua and direct macOS APIs, so you can push much further if you don't mind writing more glue code. If you care about serious Mac automation, you'll probably mix them and curse each one for a different reason.
> If you care about serious Mac automation, you'll probably mix them and curse each one for a different reason.
I do. I completely left all of the automation options. I just use Swift and the private macOS APIs. Though, I will admit that there are still some things that Shortcuts can do that I have not found a to hack around in Swift. The difference is likely due to App Intents, which is big lame.
From the repo, it signs natively on macOS and falls back to a cloud signing server (shortcut-signing-server). That fallback matters -- without macOS you would have to reverse-engineer Apple signing format yourself, and it changes across iOS versions. The hosted signing server is really what makes the whole cross-platform toolchain viable.
Looks quite cool and I'd like to give a try. What is the main use case for compiling code to shortcuts? I ask because I'm working on a tool[0] that in a way does the opposite.
What you're doing is visual programming. On its own there isn't anything wrong with it. However, specifically with Shortcuts it's not very pleasant for anything complex.
I had a full garden automation running on shortcuts, but it was extreme hard to maintain and improve due to "editor" being so bare bones.
Thanks for sharing. By the editor being bare bones do you mean some missing feature might change your mind about using it, or do you find the text-based editor much more comfortable?
I was talking specifically about Apple Shortcuts. Once you have anything complex, you're going to need code reuse and its a sad story with shortcuts as provided by Apple.
Love this approach of compiling a readable language to a platform-specific format. Same pattern works well for i18n. Write in one language, compile/translate to the target formats automatically.
The developer experience of writing real code instead of clicking through a GUI (Shortcuts app in this case) is always the right bet for power users
I’m interested to understand how this is different than Jelly; they seem to be similar. Same for Scriptable. I’ve been looking at this to hand over to Claude to build Shortcuts, something which has a terrible development experience.
You can definitely have Claude work with cherri files.
Jelly was a confusing experience for me, with JellyCuts becoming closed source and focusing on advertising, then Open-Jellycore branching out but not actually keeping up with the latest shortcut actions.
Cherri has almost every action you can find in the Shortcuts app, easy to use, and easy to create Shortcuts that can accept input and output so that they can be automated or scripted further.
You’ll have challenges with this too but you can get something by working with the three top labs’ models. Tried on Arena.ai and sent any errors back (in a personal effort to further iOS accessibility, but I digress).
I built a small app to follow my infant son's feedings and diaper changes. Simply used the shortcuts get content of url to call the API rest endpoints. This is much better !
The biggest problem IMO is the loooooooong animation delay built into every UI interaction. It's so hard to get into a flow state or even maintain a train of thought when you have to wait for everything on the screen to stop sliding, wiggling, wobbling, jiggling, refracting, distorting, and flashing.
I’m guessing you might know already, but just in case you don’t, “reduce motion” in the accessibility settings menu is essential to productivity and sanity.
There needs to be a happy medium. I don't like reduce motion because it amputates many of the spatial arrangement metaphors inherent in the iOS UI.
My biggest problem with authoring shortcuts is that the editor goes out of its way to obliterate context while you're working. Full-screen editors to change a setting lay on top everything else you're working on. Placeholders for variables, which themselves have no actual names. It's a mess.
It does seem like the logical automation platform to prepare repeatable tasks that the end user might want to do.
The permissions and secure app integration models are all there, and it’s reasonably stable.
It was always puzzling why there was never an exportable scripting language, just shareable links. I think I ended up sharing screen shots with Claude last time I wanted to troubleshoot something.
While it's not in quite the same product category, a name change might be in order; this is uncomfortably close to CHERI (cf. https://cheri-alliance.org/).
I also wrote about how Claude was able to basically learn the language from scratch and write those fully compilable Shortcuts for me [1] because it was mind boggling to me that an LLM can do that. Curiously, this is becoming more and more normal in my mind.
[0] https://lowtechguys.com/crank
[1] https://alinpanaitiu.com/blog/how-good-is-claude-really/#che...
reply