I think the biggest revelation in this episode was the tool that Charlie used to generate the RSS feed from the local newspaper (Cursor). It’s a fork of Microsoft’s Visual Studio Code with a programming window and an AI Chat feature, and you develop the code together. I’ve seen it used to generate something that would get you a solid pass for a programming course in about 15 minutes. Imagine pointing that at any automation problem, like interacting with APIs from Keyboard Maestro, and then tying that to a Stream Deck button. It’s just as applicable to an Alfred workflow, or a TextExpander JavaScript snippet, an Omni-automation or a Boop script. Much of programming is doing something that is “good enough” to get the job done with the least effort. You could get a lot further with the first iteration of a programming idea with AI tools. I’m sorry @RosemaryOrchard missed the episode, but I’d love to hear some follow up on Cursor.
Another possibility is to use Ollama and Continue:
Open source, free, and runs entirely locally on your device. I’ve not yet tried it, but it is on my list for when I have a free evening or weekend.
I have installed Ollama and it runs just fine on my Mac Studio, albeit with 64 GB of RAM.
AI seems promising as a way to improve the “Help” menu for applications. In programming, it’s proving to have some utility in drafting a starting point to a problem, or in refactoring existing code into a different format. Even when it is incorrect (trust me, AI-generated code very often will not run correctly!), it gives you enough of a hint to get started. It’s basically a quicker way than combing through Stack Overflow and coding documentation.
Here’s an example of the AI Assistant that is now available in PyCharm by JetBrains: https://blog.jetbrains.com/ai/2024/08/jetbrains-ai-assistant-2024-2/