The 50/50 Club

Recently, Inga asked a simple question: “Where have we actually been?”

That stopped me. We have been doing this monthly lunch for years. We split the meal. We split the bill. We split the soup, the salad, the sandwich. Ice cream is mandatory and never shared. But we had never written any of it down.

Then I remembered something: Google Maps keeps a detailed record of your location history, stored on your phone, not just in the cloud, if you have that feature turned on. I did.

This is not a project I would normally spend time on. But it was a perfect excuse to practice something useful: pulling structured data from a messy archive, filtering it into a pattern, and combining it with open source mapping tools. A small learning lab for these new AI coding agents.

From Noise to Pattern

The exported file contained tens of thousands of location records. Every errand. Every stop. Everything.

The first step was defining the rules of the 50/50 Club:

• After 2018
• Last Tuesday or Wednesday of each month
• Between 11:00 a.m. and 2:00 p.m.
• Not near home

That reduced the dataset from a life log to a recognizable rhythm.

Making It Human

Another friction point appeared immediately. Many locations were labeled with internal place codes instead of names. Technically accurate. Visually useless.

Those codes were resolved into readable place names. Then the list was narrowed again to food-related categories only. Restaurants, cafes, bakeries, markets. No hardware stores. No offices.

Now it reflected what it was meant to show. Lunch.

From List to Map

A list answered what. It did not answer where.

So the project became a minimal, touch-friendly map centered on Sonoma County. No extra layers. No clutter. Each point shows:

• The place name
• The date
• A direct link to the Google Maps listing

Two real-world corrections came late. A couple of San Francisco lunches were excluded by overly strict geography rules and were added back intentionally. Some map links did not open the precise listing and were reformatted to resolve correctly.

The result is live at: 5050.pages.dev

What This Represents

This is not a review site. Not a growth project. Not a product.

It is a memory device.

The interesting part is not the map. It is the shift from raw personal data to something readable. A recurring lunch turned into a visible pattern. A private archive became a small public artifact.

Without AI coding tools, I would not have bothered. With them, this became a one-weekend experiment in filtering, sorting, and presenting personal data clearly.