The Generalist’s AI Advantage: Unlocking New Capabilities in the Age of LLMs
Sometimes being a generalist sucks. I know a bit about a lot of things, but I don’t always have the depth of skills required to turn the picture in my head into reality.
But in one area of my life, that is starting to change thanks to the new breed of AI tools (like ChatGPT). That area is to do with writing code. It enables me to do things like creating productivity hacks on my Mac and writing complex automations in low-code tools like Airtable, which were previously out of my reach.
A great part of being a generalist is having a sense of what’s possible. I have worked in a wide variety of roles and domains, including alongside software developers for years. I’ve also watched more than my fair share of YouTube instructional videos. All that random experience usually gives me a decent inkling of when there is some way to create a solution to whatever challenge is at hand.
All too often I would embark on a project to develop an automation or a plugin of some kind. I’d be in the flow, imagining the possibilities, getting excited until…. I would encounter the need to write some custom code in order to progress. Damn! After a few clumsy attempts and then getting confused after going down various rabbitholes on Stack Overflow my enthusiasm would have waned. Another half completed project on the scrapheap.
Sure, I could have pushed through and learned the fundamentals of coding in JavaScript and honed that specialist skill. But my generalist brain would always move on too quickly to the next shiny thing, and my project would remain half done.
What’s changed is that I can now work with an LLM (Claude Sonnet 3.5 is my current daily driver) to explain what I want to do and quickly get the code snippets I need to bring my ideas to life.
Before I share some examples, a caveat. None of these are groundbreaking or elaborate, and that is part of the point. They are just little quality of life improvements that I have been able to achieve with fairly minimal effort. Each one of these creations gave me a dopamine hit of satisfaction in the moment when I first cracked the nut, and then an ongoing sense of achievement every time I use them.
Example 1: JavaScript shortcut to clip a webpage into Roam
Do you know that you can use “bookmarks” to run JavaScript functions in your web browser to perform certain tasks? There are a few I use every day.
One example removes all of the junk characters at the end of a URL that are there for purposes such as tracking and are annoying when you want to copy and paste a clean link.
Another is a bookmarklet that copies the highlighted text and formats in a way that is ready to go into my note capture software — Roam Research. I didn’t actually write this function originally, but I was able to tweak it to my exact liking by iterating it through the LLM and asking it to make changes to the code until it was working exactly as I wanted.
Example 2: Using AppleScript to do a mail merge on a Mac
It’s easy to forget that we spend all day working on a computer. Our MacBooks have become so slick and “appliance-like” that we don’t see or interact with the underlying commands that make it do its magic. The ability to control these machines sits there just below the surface if we are willing to peel back a layer and interact with them.
On the Mac, we can use AppleScript to write instructions that can control applications and parts of the macOS operating system. AppleScript uses a natural language-like syntax, making it somewhat easy to read and write. But there is a learning curve, and I never managed to bust through it — until now, thanks to AI.
One cool use case is to ask an LLM to create a script that creates, say, 20 emails to individual recipients, with the salutation correctly updated in each one. You just need to give ChatGPT or Claude the email text, the list of recipients, and instructions of what you need. It will then return the script, which you can then run in the Script Editor application on your Mac, and within seconds you will have 20 draft emails in Apple Mail ready to send. There are other ways to do this (e.g., using Sheets and Pages), but this method is fast and direct.
In addition to the web browser and AppleScript examples I’ve shared, there are many other ways these methods can be applied. Here are some of the other applications where I apply the similar technique:
- OmniFocus — creating automations in my task manager
- Alfred — creating custom workflows in my launcher
- Airtable — doing transformations in automations
- Home Assistant — writing YAML to control my smart home
- Keyboard Maestro — creating clever and complex macros
These types of applications are all inviting us to be more than passive consumers. They are platforms we can use to solve our own unique challenges. I believe that generalists are in prime position to take advantage of this opportunity, so that we can be more efficient and effective in doing our work. What once might have taken hours to do, or would rely on paying a specialist to do, a generalist can now achieve in minutes.
There are two things you need to be able to create solutions like this for yourself — a mindset and a skill.
The key mindset is curiosity. You should be hungry to find new and better ways to improve your workflows, and be willing to try some ideas.
The key skill required to enable this is knowing how to work with the LLM to iterate on the code until it is working as you want it to, as it often takes a few attempts to get it perfect. This usually involves just copying and pasting any error messages you receive, and then the LLM will know what to do to fix it. To make improvements, you just need to briefly describe how you want it to work differently. But either way, what I have found works more for me is to not labour over reading the code and trying to work out how to improve it. Instead, I just focus on instructing the LLM to make it better and testing it until I’m happy.
While the examples I have shared are more to do with personal productivity, this way of thinking and working opens up a world of possibility in terms of product development and even how future companies are formed. Every is an example of a trail blazer in this regard. Even though they started out as a collective of writers with a newsletter business, they have been able to use this way of thinking and working to incubate software products that they now sell to customers — all without needing to hire armies of coding specialists. I believe that this is a harbinger of the future of work. It belongs to curious. Generalists with a keen sense for interesting problems to solve are now unshackled thanks to freely available LLMs. Who knows what they might create?
