Author: Taylor Robbins Date: September 9th 2025

I recently made the decision to leave my job in the game industry and work on my side project full time. The side project I ended up settling on is a tool in a field of expertise (Geographic Information Systems or GIS) that I have not personally had a lot of experience with, and although the goal of the project is to make something that I might use and therefore I have some intuition about what kinds of tools might not exist that I want to make, I don't actually know much about the existing endeavors in the space, or the breadth of the target audience I am designing for. I suspect that one of the following may be true:

1. My goals have been considered before and either the software already exists but is not well known, or it has been implemented in a sub-optimal way negating the value proposition of the workflow. For example maybe the tool is only available in a web browser, or only available for dekstop, or only works on a small set of data before slowing down, etc.
2. The tool I want is particularly hard to implement, meaning that even though the value of the tool would be high, it's not possible to create and previous attempts have never reached a state of release that I would be able to find.
3. The tool I want is indeed “novel” in the sense that no one in the space has tried to make this type of tool. This is uncommon in the sense that a person that has a lot of experience in a field has likely come across a lot of the “low hanging fruit” kind of ideas, so my ignorant view on the space is less likely to actually prove novel. However, if we talk about the specific form that I imagine the tool existing, the “cross pollination” between two areas of work does often result innovative ideas. For example, my experience making games leads me to think a lot about “fun” interactions and good user experience (UX) patterns as a primary concern and I don't often consider absolute robustness of the tool to be the primary concern (which might not be true for someone coming from an academic background).

With those possibilities in mind it can often feel a little daunting to enter a new space and try to make progress. I am very likely to make incorrect assumptions or follow threads that end up being dead ends. Or I may do a bunch of work on a tool that I think is cool only to realize too late that some tool already exists and is much more capable than my tool.

For this reason I am approaching this project with a much different style than I normally would a video game. I don't have a fully formed idea of what the end product should be. I know I am ignorant of many facets of the problem and most of the history in the space. But I won't make progress on either problem unless I push forward and start trying to make things in the space. While I go I will do research into the space (reading books, finding existing software and testing it's functionality, finding and talking with people in the field, etc.) and then I will take that knowledge I am obtaining and try to apply it to the software. I suspect I will ping-pong back and forth between software development and research quite a lot. This sort of hybrid approach should allow me to make progress rather quickly.

This page should serve a little bit like a blog of my experience, but it will be centered around this core idea that should be applicable to more than just my experience. Each section will cover a particular concrete example of my experience moving from Video Game Development to Geospatial Software, but hopefully the examples help highlight the kind of thing you might experience if you decide to do a similar major transition into a new space. In particular, if you are considering moving into a new field and have been scared of the process, I hope that this page helps encourage you to take the plunge.

The is obviously a lot of existing software in and related to maps, so let's go through a few things that I know exist:

Companies like Google and Apple have a massive investment in collecting information about our planet and providing access to that data in a huge variety of ways. These endeavors represent the sort of primary thing that a computer might help someone do in-relation to geospatial data. Almost everyone has the need for directions and information gathering. The primary goal here is to collect information from various sources (satellites, cars rigged with cameras, etc.) and sort/store/distribute this data to the user in the most convenient way possible. One primary distinguishing factor of this software kind of software is that it deals with largely “timeless information.” The roads marked on google maps are likely to stay the same for a long while, and the rate at which they change can be overcome by the rate at which new data is collected (new satellite or streetview images, and internal work done to catalog and transform those inputs into which roads are shown on the map). In some sense all data is “timeless” if it's paired with the time in which the data was collected. You can look at a Google streetview image from years ago and still find the information useful, as long as you know when the data was collected. But most often these tools try to present the data in a timeless manner, trying to promise that the data they are presenting you is up to date as often as possible. Some kinds of data that people might want related to maps is not as “timeless.” It changes rapidly, or is labor intensive to collect or process into a format that is useful. Not all timeless information is like this. For example traffic data is very time sensitive, but the method by which a company collects the traffic data is rapid enough that the data can be created and consumed before it becomes irrelevant. One other distinction is that the data is “impersonal”, meaning it is useful to many people. A roads existence or name is useful for anyone who may use that road, and although the interest in that road is not equal among everyone, it is useful for enough people that it is worth collecting as part of the dataset and providing access to users.

Another kind of geospatial software that exists is personal fitness aids. Many of these are mobile apps that help you track and plan your exercise. For example, you can record your bike ride or run using a smart phone or smart watch to track your GPS location throughout the run, using one of these apps. That location data often gets uploaded to a server somewhere, and the way in which you use the data is determined by the design of the particular app. This kind of location data is distinct from the global maps software in that the location data (and often metadata like time, heart rate, etc.) is personal, it mostly is useful for you alone. As such most of these apps will guarantee an amount of privacy in relation to your location data. However, a distinction of this kind of software is that it doesn't allow you to “own” your information fully. Since the data is uploaded and managed on an external server, or saved in a format that is hard to view/edit as a regular user, you can't choose how you use that data. You are restricted to a particular set of “views” on the data, defined by the goals of the application. So a fitness app is likely going to give you information in a form that is helpful for tracking and encouraging exercise. You are generally not able to create your own custom views of the data. Usually the “raw” data is inaccessible to the end users. Additionally, since servers are involved these applications are very often subscription services to offset the price of the servers. And the guarantee of privacy I mentioned earlier is not always to the extent you would want.

• TODO: Software like ArcGIS
• TODO: Software like fantasy map creators https://inkarnate.com/
• TODO: Software like JOSM
• TODO: Python packages for interacting with GIS data

• First proper use of double
• Remembering my focus (end user, produce and consume, game-y)
• Offline rendering of tiles. Downloading subsets of data from API. Map display expectations from users that are hard to implement in real-time.
• The size of OpenStreetMaps data. Space partititioning. Back of the envelope math for computational capability for an average laptop/desktop. How much distributed computing do we need to solve the problem?