I expect that those of you who consider running a script table stakes, especially in a Power Users forum, may roll your eyes. I do not blame you. Nevertheless, at the risk of embarrassing myself, I am delighted that I was able to use Claude to create a script that saved me at least an hour or two of tedious work. And, it is one I can reuse.
I had a very long timeline of critical dates that I wanted to get into Reminders. The only way I knew to do this was to copy each task laboriously, paste it into a Reminders list, and then complete the multiple clicks required to add the due dates to each task. That is the type of work I hate, and it is not a good use of my time. So I decided to ask Claude whether there was a way to recreate the list and have it import with the due dates into Reminders.
The short version is that Claude created a script, something I could never have done. The result was a very large list of tasks for this particular project imported into Reminders with all dates entered and sorted by due date. Now that is what I call a good use of AI.
That sounds great. As an Apple Reminders (and Claude) user, I never thought of using it in that way. I know you can connect to Reminders in iOS, did you do this on desktop? Did it employ the terminal to execute the script or Shortcuts?
Glad to hear you succeeded with AI-augmented scripting! Hope that makes future scripting endeavors less daunting and you will try again. Thanks also for sharing it here. Hopefully this encourages/empowers other non-dev MPUs to create their own scripts.
Not yet, but I’m open to it. As long as I don’t have to create them and I can get them from reputable sources, I’m open to learning new ways to use a tool I’ve never touched before.
You and I and everyone else on this forum could write scripts—and enjoy doing so—had we world enough and time, much as we could sew our own clothes, make our own furniture, or keep a set of books with a pencil and a thirteen-column accounting pad.
If I begin a prompt with something along the lines of “I want to do this thing on my computer” Claude Cowork / Code will routinely propose that it write a script for me to use myself as needed (applescript, bash, python, whatever depending on the use case), whether I ask for one or not.
Gemini will also write scripts for me, but amusingly enough, often follows up with the observation that it could teach me how to fish, too. (Once I responded “sure, why not” and it asked me if I wanted to learn the right way … I wonder what it would have done if I’d said, “Nah, teach me the quick and dirty way.”)
Anyway, it’s liberating to have a little library of purpose-built scripts at my disposal that I can use whenever I need to.
Out of curiosity, I asked Claude what was going on under the hood when I used Cowork (or the non-terminal version of Claude Code in the desktop app as the case may be), and got this answer:
Yes, there is a terminal (or something functionally equivalent) running in the background. Here’s what’s actually happening:
When Cowork starts up, it spins up a lightweight Linux virtual machine on your computer — on macOS via Apple’s Virtualization Framework, and similarly on Windows. When you see “Starting Workspace,” Cowork isn’t loading a UI; it’s provisioning a compute environment, booting a lightweight Linux root filesystem, and creating a sandboxed jail that mounts only the folders you’ve granted access to.
Inside that VM, Claude runs commands in a sandboxed terminal environment. You never see it directly, but it’s there — writing scripts, executing them, checking output, and iterating. When it writes a script as part of solving your task, it runs that script immediately, checks the output, and iterates. You never see the terminal. You never touch a dependency.
So the short answer: yes, it’s essentially a terminal running inside a hidden Linux VM. The chat interface is just a layer on top of the same underlying agentic architecture that powers Claude Code, which is itself an explicitly terminal-based tool. The VM is what makes Cowork meaningfully different from regular Claude chat — it’s not just generating text, it’s actually executing code and modifying files.
One thing worth noting: every Cowork session starts from a blank slate for security — it doesn’t remember previous sessions’ temporary files, so it rebuilds the environment each time.
Next steps we could explore: looking at what the VM can and can’t access, how the sandboxing compares to Docker containers, or how sub-agents are coordinated within that environment.
I should note that Claude searched the web for this answer. Here are its sources if you want to check its work or learn more:
No eye rolling required. We each have our strengths and priorities.
I recently used Claude for the first time, to create a very simple Wordpress plugin. I could look up plugin basics, but had no idea what hooks I needed to use nor how. Claude did the work and then I inspected it. It made sense, so I installed it. It works great.
Indeed. I have neither the time nor the inclination to learn how to script. Even if I did, given my lack of knowledge in this area, it would probably take me ten years to recapture the time devoted to creating scripts in the first place.
I am also beginning to see the productivity gains at work.
Yesterday, our HR director was asked for information on our leave policies. Those policies are spread across different sections of employee policy manuals. Rather than copy and paste the information into another document, she uploaded the PDF and had Claude extract all the relevant policies into a Word document that she then sent to the person requesting the information.
Our CFO was preparing a presentation for employees on compensation and other related matters. Instead of taking the time to create slides, or asking my EA to help, he had Claude create them. They turned out excellent.
I wanted to highlight a Youtube address with Popclip and send it to Reminders into a Watch Later list. Claude provided me with an Apple script and a step by step to make a Shortcut. Now I can click into the browser address bar, click on a Youtube icon in Popclip and send it to Reminders.
Those of us of a certain age have been here before. I’m not downplaying the impact that AI in general and LLMs in particular will have on our lives, both at home and at work, but rather harkening back to the other tech revolutions I’ve lived through.
When I first started my corporate career, working with computers meant working with mainframes. Hardly anyone had a PC on their desk. For knowledge workers, looking something up meant hauling down a book or rummaging through a published report. Writing something meant longhand on a legal pad, which you then handed to a secretary to type into a document created on a shared minicomputer buried deep in a sub-basement. Revising was literal cutting and pasting and handing it back to the same secretary for re-processing. If you were away from your desk and the phone rang, a secretary took a message for you and wrote it on a little pink slip bearing the title “While You Were Away.” All math that wasn’t done on a mainframe was done with an HP12C and an analysis pad. Barely five years later, I had a PC with a 10MG (yes, MG) hard drive on my desk, voice mail, a shared printer and a fax machine on my floor, and no secretaries for anyone but the department head. Biggest of all: I was doing my financial modeling in Lotus 1-2-3. (Lotus 1-2-3 is long gone, but I still have my trusty HP12C.) In not many more years, I was on a LAN sending email messages to the far-flung outposts in our corporate empire, building a departmental website, and downloading EDGAR filings and government data sets.
I wasn’t sure that LLM chatbots would be as big a game-changer as the computer spreadsheet or the internet, but I’m pretty sure agentic tools like Claude Code/Cowork will have a meaningful impact in the workplace.
Gemini will also write scripts for me, but amusingly enough, often follows up with the observation that it could teach me how to fish, too. (Once I responded “sure, why not” and it asked me if I wanted to learn the right way … I wonder what it would have done if I’d said, “Nah, teach me the quick and dirty way.”)
That is hilarious. Gemini must have a better sense of humor than Claude.
Anyway, it’s liberating to have a little library of purpose-built scripts at my disposal that I can use whenever I need to.
It may be how I’ve dialed in my preferences, but Gemini appears to have no sense of humor at all: it’s all business all the time. Claude at least deigns to laugh at my jokes from time to time.
Nevertheless, at the risk of embarrassing myself, I am delighted that I was able to use Claude to create a script that saved me at least an hour or two of tedious work. And, it is one I can reuse.
(Once I responded “sure, why not” and it asked me if I wanted to learn the right way … I wonder what it would have done if I’d said, “Nah, teach me the quick and dirty way.”)
