Finding Your Voice: From Themes to Scenes

The Power of Knowing What You Want to Say

Hi everyone,

This lesson was a lot of fun—I guess because I’m fully immersed now in the actual writing of the story. I’m also writing this a bit sooner than expected to ensure we hit our two-lesson goal for November. Over the break, I’ve been diving into the thematic structure of Silicon Dreams, and I’m more excited than ever to push forward with the novel.

One thing I’d like to clear up early on: while this class is about using an LLM to produce high-quality writing, it’s increasingly clear to me as I use LLMs that we’re all best served as writers by starting with a solid foundation of knowledge and, more importantly, an internal desire to say something worthwhile, interesting, and meaningful. I can’t stress enough the power of knowing what you want to say.

Prompting the LLM becomes both straightforward and rewarding when you provide it a prompt that reflects your own thoughts and intentions. In other words, it’s you writing this novel (or novella). The LLM is a tool—a powerful one, but one that shines brightest when guided by a clear vision.

Let me explain.

While ruminating on Silicon Dreams over the break, I realized the story is profoundly theme-driven. Telling the story of the birth of the web and the rise of AI isn’t just about recounting events—it’s about uncovering and exploring the tensions and themes that defined the era. For example, the DotCom crash isn’t inherently interesting in the abstract—bubbles burst all the time. What’s compelling is what that crash revealed about how the web would (and wouldn’t) unfold and prosper. That’s the theme, buried beneath the surface-level events, press releases, and exuberant futuristic visions of the time.

But the early 2000s weren’t defined by just one theme—there were many, and as a writer, your job is to decide which resonate most with you.

What theme interests you as a writer? What do you wish to focus on?

Let me tell you what I did.

I was working at an “old-fashioned” AI company in the 2000s. It was my first professional job as an AI engineer, and the experience has stayed with me to this day. The theme I wrestled with back then has since been answered, but at the time, it was a battle playing out in academia, industry, and even in public intellectual spaces like books, talks, and debates. The central question?

What’s the future of AI?

Would symbolic methods—the traditional, rule-based systems—find new power in the burgeoning World Wide Web (the so-called “semantic web”)? Or would the “shallow” statistical machine learning methods prevail?

That was my theme, rooted in my personal experience. Take a look at how I explored it in Silicon Dreams.