Dear XYZ,

in this piece “AI” refers to LLMs and the “agents” built on top of such language models

I feel like opinions around AI usage is polarizing into two encampments. On one end, you have the AI maximalists who believe that AI is the inevitable future and are trying to use it in any way that they can. On the opposite end, some would oppose any use of it.

I think both sides each have their own rationale for justifying their positions, but that’s not what I’m interested in writing about today.

As someone who makes things, I’m interested in the tools that we use and how it shapes the things that we make. I often recall the quote misattributed to Marshall McLuhan that goes something along the lines of “we shape our tools and thereafter they shape us”.

I’ve observed that for myself, my comfort around using AI differs quite a lot depending on the task. For instance, I actively use AI for programming and making software, but I can never imagine using AI to write articles like these.

Then again, even for writing software, there are times when I’d rather read and review every line of code even when it’s generated by an AI tool. There’re so many different ways to use the tool that also impacts one’s experience with it.

I think much of it boils down to how much you care about the thing that you’re doing or making, and how that artifact relates to you. btsv is made using AI assistance, but I will go pretty far to say that it wasn’t vibe-coded. I know the code base well, and had a strong hand in shaping all the major design decisions. I’m willing to own both the merits and mistakes in the project, instead of shrugging my shoulders. Because it is a project that I care about, I will make sure that I “paint the back of the fence”. I also want to understand how it works so that I can imagine how to grow it over time.

In physical tools, the tool itself provides a lot of constraints over how something is made. For instance, a Japanese hand saw is going to cut a lot slower than a bandsaw, but in a way it forces you to be a lot more deliberate and careful. The hand saw also offers a lot more control compared to a band saw.

But I’m not sure if using this analogy is applicable for AI. AI is such an amorphous tool, I really wonder if its effects as production tooling will be obvious. It can simultaneously help you learn a lot or nothing at all, depending on your choice. (I’m assuming that you have the volition, but that’s not always true in organizational settings – but that’s a different conversation.)

AI tools can do a lot nowadays, but they can’t desire. And even if they someday were to desire, that doesn’t in any way change how we should desire. Just like how my cat’s desires are parallel to mine and can coexist simultaneously.

I guess our relationship with this powerful tool really boils down to the intentions that we bring to it. It’s like a genie; it reveals the character of its master. If you want to use it to do difficult things, you can. Conversely, you can also use it to avoid doing anything.

Individually, you should be aware of what using AI is doing to you as a person. Is it making you better than you were before? And perhaps more profoundly, is it congruent with how you want to live?

xoxo,
K