Give your wildest idea fifteen minutes

We had our annual madewithlove retreat this week, and I was asked to give a talk. A talk needs a slide deck, these days I start a Codex session to get to work.

I remembered an article in Lenny’s Newsletter by Anshu Chimala, which encouraged being more ambitious with your prompts. You know how people always say you should tell a story instead of giving a presentation? What if the deck enacted that story? I wanted a pixel-art office where the characters were doing the work I was talking about.

Fifteen minutes to try the idea

My first reaction was that this was too ambitious. I had a talk to prepare, and building a moving office sounded like a distraction from the actual task. I wasn't looking for another side project.

But I'd seen enough posts on Reddit of people building 3D games in a few hours to wonder whether I was underestimating what LLMs could do. I decided to give it a quick prompt and see how far I got. After fifteen minutes of prompting, I had an HTML deck built with Three.js, complete with a pixel-art office and moving characters.

Fifteen minutes gave me something concrete enough to judge. Did the office help me tell the story? Did the moving characters earn their place? Those questions still needed answering, but I'd nearly dismissed the idea before giving myself anything to evaluate.

A colleague mentioned that their kids watch these absurd cat videos, so I tried generating some with ComfyUI on my local GPU machine. The workflow ran. If you think about what effort was needed in the past to create this, it's impressive. But the result was, of course, slop. Shit in, shit out still applies. I'd asked for cat videos without bringing much to the table, and I got exactly that.

I'd nearly ruled out moving the office because I assumed trying it would turn into another project. Once I had something in front of me, I could decide whether it deserved more of my time.

LLMs can help you tackle wild ideas. Give it fifteen minutes.