An example of how AI is changing the way I work (and live) as a product designer
Snapshots from the AI-generated app I created. You draw the shape of your driveway and it outputs the optimum route, square footage, estimated time to walk the distance (all at a static shovel width of 2 ft). There’s a lot would improve but I went from idea to research to app in less than an hour.
I’ve been challenging myself to use AI in everyday use cases so that I can get closer to the technology as a product design leader. To that end, I want to share my attempt at using the latest generation of AI tooling to build an experience that helps people shovel snow more efficiently — an often mundane and tedious task.
Some additional context about me:
I used to be a developer in a past life but have since lost touch with most skills associated with being able to “build apps”I work in tech and so I’m probably at the leading edge of most of the available tooling compared to most peopleI have taken a app-forward approach to solving this problem and it’s possible there are alternative approaches using AI to do the same (like deeper research)
TL;DR; I ultimately succeeded in the task and developed a small web app where I can draw the shape of my driveway and calculate the optimum shoveling route. I went from idea to research to app in less than an hour (it took me longer to write this article).
Here is my approach and the tools I used to solve this problem.
Research
First things first, I needed to determine if the answer already existed. Was there research that told me the most efficient route to shoveling a driveway? To find out I went to Gemini Deep Research (GDR).
Using GDR, I ran a meta-analysis across 38 different websites and papers, triggering the research with the following prompt:
“can you please do research on the topic of what the most efficient way to shovel a driveway? My driveway is a long rectangle and I shovel it by hand with a 2 foot wide shovel. I’d like to understand how I can shovel my driveway the most efficient way, what pattern I should use to shovel the fastest and clear the area of my drive way the quickest. If there isn’t enough research on snow shoveling, you can use other research that might be applicable in the areas of mathematics, geometry, and physics.”
Before generating the report, GDR develops a research plan, showing its proposed steps to conduct the research. Here’s an example of its response to my prompt:
The Gemini Deep Research plan describes the steps it will take to conduct research and lets you correct its plan of action before executing the minutes-long process.
This looked good so I click “Start Research,” sat back, and sipped my coffee. In just a few minutes an exhaustive report was spit out with references clearly cited (complete with links to websites, PDFs, and images).
In my case, GDR concluded that there’s little evidence of an optimal route to shovel a driveway like mine. It also gave some great advice around different patterns to try, the physics of snow shoveling, and how to shovel in a healthy way to prevent injury or over exertion.
I cannot yet easily share GDR reports directly from Google. So, I processed my report with Gemini NotebookML (another wonderful tool for conducting first-party research). From NotebookML, I converted my findings into podcast. Feel free to give it a listen here if you’d like.
An excerpt of the Gemini Deep Research report on shoveling snow.
Answering the question (and building an AI-generated app)
Given that the Internet’s top physicists and forum writers didn’t have an answer for me, I took matters into my own hands.
There are several different surfaces I need to shovel so I thought it’d be neat-o for me to be able to draw the shape of those areas and have an app tell me the optimum route to clear them.
What is optimum? I defined optimum as “covering a given area in the least amount of walking time necessary to saturate the space.”
For creating my app I turned to v0 by Vercel, as I believe it to be at the bleeding edge of generative AI app development. With v0, you can turn text prompts into actual functioning apps with code and deployment options.
An example refinement to the app. 1 of just 8 prompts it took me to create the snow shoveling estimator app.
My experience was magic. It took me 8 text prompts to generate what I believe to be a pretty good MVP. Did I build the next Facebook? Absolutely not. But I went from idea, to research, to functioning app in about an hour.
Checkout my snow shoveling route estimator here: https://v0.dev/chat/driveway-snow-route-planner-UP6yTAh9Xnp?b=b_yQkw8ddGh2c it’s not perfect but it’s pretty damn cool. At this link you can see my entire chat transcript and the versions I went through to create the app.After your prompt you can see v0 get to work, modifying the code (in this case React) and eventually pausing to generate the runtime preview.In the app development view, the preview renders to the right of your transcript.
Not an AI-generated conclusion
If this were my full time job, the potential seems limitless (and it’s still early days for this technology). I’ve been using AI on a weekly-basis now and it’s changing the way I approach problem solving.
I used ChatGPT’s multi-modal inputs (voice and video) to help work on an old snow blower, I researched a family trip to Yellowstone (and sent the podcast to my wife), and I’m using AI almost ever day in my work as a product design leader.
This is one of the most exciting times to be in technology and the pace at which these innovations are happening is truly stunning.
Have you been experimenting with AI? What have you been able to accomplish?
Using AI to optimize my snow shoveling was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.
