🔗 iA: Turning the tables on AI

Saturday, 1 March, 2025 — links generative-ai

iA, developers of iA Writer and iA Presenter:

Tech companies big and small sell AI as something that thinks for us. It does replace thought with statistics—but it is not intelligent. No one knows what the future will bring. But is a future without thought a better future?

Now, with a tool that might help us think… How about using AI not to think less but more?

There are two primary areas I am investigating using generative AI tooling with, writing and software engineering. This is true for both work and at home. I’m still trying to wrap my head around what the evolving possibilities are, and I’m still building a habit of reaching for the tools in order to learn how best they’ll work for me.

The questions and examples in this article are a critical pivot in generative AI’s use from, “let it think for you”, to, “here’s how to leverage generative AI into higher quality writing”. That’s a bit different in approach than how I think generative AI is commonly used.

I am way more comfortable using generative AI as a reviewing tool, for help in unsticking things, or spotting problems vs. “doing the work for me.” I can easily see using similar approaches to edit and improve my general writing and my software development.

On the software engineering side at work, my engineering organization is strongly encouraging the use of tools like GitHub Copilot and ChatGPT and is providing licenses for their use. The key suggestions are working with GitHub Copilot in an IDE and ChatGPT elsewhere for generating code, troubleshooting code, explaining code, and brainstorming.

The institutional sense is these tools are roughly equivalent to an exuberant, but inexperienced, intern. Accordingly, the guidance is to delegate work to the tools, but thoroughly validate the results. I think this recommendation is a correct approach. Another, and where I’m finding more natural alignment with is using these tools for review and editorial feedback. The iA suggestions here will help inform how I approach prompt writing and what I’m looking to get out of code generation, to the point where I may build a shell first, then run snippets by and seek improvements.

Similarly, Writing-wise, I’ve made some use of the Apple Intelligence writing tools and had a niggling sense there was a better way of handling their use than I have been so far. This post has just such an example of comparing Apple Intelligence writing tools output against the original and then reviewing a change-by-change comparison. I like that and have an easy way to do that in BBEdit. As with software engineering, it feels more congruent to my way of working.