ColeslawNearMe Update #3

After a few weeks hiatus, today I finally had some time to get back in the saddle and make progress on my AI-generated coleslaw maps website.

For those who are not familiar, about 2 months ago I had the dumb idea to build a website that finds the best coleslaw restaurants in your area ranked from best to worst. And I’ve built the entire thing without knowing writing any code, all through the help of AI.

I’ve made a series of short videos (here’s parts 1, 2, 3, and 4) and blogs (blog 1 and 2) documenting the journey from idea to implementation, following how I’ve managed to build it from scratch without any prior knowledge. The entire thing has probably been about 20 hours of dedicated effort (excluding time spent making the content).

Here’s a quick recap:

1. Used v0.dev to design the original interface for me

2. Used Claude.ai to orchestrate building it into a functional website

3. Used Cursor.com to write and implement the code for me

4. And used Replit.com to actually deploy it to the web

And after a few weeks up front, I finally had a serviceable website that used the Google Places API to identify restaurants with coleslaw on the menu, rank them (roughly) and display them in a nice little map for my pilot city of Chicago:

That’s where I got stuck (which you can read about in blog post #2).

Essentially, I didn’t really think far enough ahead upfront to how I’d get data for the rest of the coleslaw restaurants throughout the US. I just kind of took what Claude told me to do and ran with it, without extrapolating beyond a few hundred restaurants.

The problem? There are 100,000+ restaurants in the United States, and my method for identifying coleslaw was to essentially call Google’s API a few times for each restaurant to 1) identify every restaurant, 2) check it for coleslaw, and 3) come up with a “rating” and other stats.

Ultimately all of those API calls cost money (even if just a few cents) — but when you extrapolate out to thousands of restaurants, the costs adds up. I realized it would cost me $5000+ just to get coleslaw for a few states using my current method.

SO… I was stuck. But I am working around it. Here is my new plan:

1. Start with the biggest cities (a progressive list of the top 50)

2. Add known coleslaw restaurants (using franchise + menu info)

3. Rely on user submissions to handle the ratings

4. Then expand outward to get entire states, one-by-one

This method seems like the best way to keep a reasonable amount of people happy and still have fun with it (roughly 20% of the country’s restaurants are located within the top 50 cities!). And anyways, cities are really where restaurant “decision fatigue” is most prevalent, and where a ranking system for restaurants is probably most impactful anyways — eventually maybe I’ll expand out to other foods and sides like mac ‘n cheese.

With all that outlined, I finally sat back down to make some updates on the app, and here’s how it looks now:

Key Updates:

1. Updated the layout to be one large side-bar vs. two small side-bars on either side of the map. This feels more efficient for showing the stats and scrolling through restaurants

2. Added restaurant data for two more cities: New York City and Los Angeles, and updated the infrastructure to handle adding any number of cities.

3. Updated the scoring and summary stats to be more dynamic based on which restaurants are actively shown in the map area

4. Updated the marker system to use fun coleslaw ranking badges (vs. color coded pins) — a very “small” change that feels like it has added a ton of quality to the user experience

5. Most importantly, finally published it onto an official domain: slawnearme.com

This probably took me two hours to do, and was accomplished almost entirely using Cursor’s chat feature. I’d basically ask it to make improvements one at a time (never ask it to do two things at once or you get stuck in debugging hell), and iterating until I got it right.

For UX adjustments I mocked up basic wireframes in powerpoint (yes, my favorite low-friction way to create mock-ups these days), and started it off with a nice prompt to get the AI up to speed on my next steps. You can see my starter prompt here:

Overall I am pretty happy with the user interface at this point, and will probably keep it this way for awhile. Now that I’ve started adding unranked restaurants city-by-city I’ll be able to make steadier progress without spending so much on cloud costs. The next big steps are:

1. Add a user-submission functionality so people can rank and submit restaurants

2. Continue adding data for my next group of cities, the goal is to get up to 5 cities well-covered in the next few weeks

3. Fine-tune the scoring system (right now I have a lot of 4- and 5-star restaurants, not many 1’s through 3’s — maybe that’s because all coleslaw is good coleslaw???)

4. I’d like to do a little bit of SEO work to see if I can’t get it to rank for search terms like “coleslaw near me” (also as a fun way to practice my SEO skills)

So long story short, we are back in action, and I’m feeling better. A few weeks ago I thought maybe this project would be dead (as you can tell from the blog post #2 title…) but the lesson is that sometimes all you have to do is sit down with the goal of only making a little progress, and a path will present itself.

Cheers to coleslaw and fun with AI. Let me know what questions you have about how I built it and I’d be happy to dive deeper :-)

Peace,
Ramsey