Publish or Perish: A Text Adventure of Academic Survival

A few weeks ago, I listened to Matt Grossman’s interview with Andy Hall, in which he discussed how he and Graham Strauss used Claude Code to extend a previous study. While Claude did make mistakes, they were correctable for an informed user. They may be a stepping stone towards greater research productivity by reducing time spent on mechanical coding decisions and increasing researchers’ freedom to focus on other important parts of the research process. Their write-up is an informative read on their findings, or you can listen to either of the interviews (1, 2) they did about the project.

The Quantitative Peace has been dormant for about five years. Part of this dormancy was due to backend issues beyond my expertise. While I could troubleshoot them, it would take more time than I had to devote to them. The website not working was a bit of a sadness as well, since we have several popular posts that have been linked to many times over our history (including topics of the Dark Knight, Zombies, Songs that mention a specific age, fictional political scientists, and the Job Market), but those links are now dead (until today). Given both my limited time and Claude Code’s potential for use by researchers in professional environments, I decided to gamble that $20 could bring this blog back.

My first few forays into Claude Code had nothing to do with the Quantitative Peace, but were instead to see if it could code some toy projects for me and deliver outcomes that would work even though I could not troubleshoot them myself. While this is counterintuitive, starting from ignorance seemed like a first step to understanding how much intervention I might need while using the tool.

First, I had Claude create a game that was like rock-paper-scissors, but it was about matching a handshake at random. Do you fist-bump, handshake, or high-five? The game had a simple 8-bit-like interface, and drawing the hands was difficult for the coding engine (as it is for most AI engines). I switched it to using hand emojis, and the core of the game worked well enough. I could deploy the HTML, and it ran. I wanted to create a high-score board as well, so Claude told me how to set up an external database that users could write to and the game could pull from for high scores. All worked pretty well. An inn joke inspired the game with a friend group, so I am not going to share the result here, but it did what I needed it to do.

For project two, I went back to my roots of coding in BASIC. As a teenager, I loved making text adventures. While I was firmly in the SNES/N64 era of graphics, text adventures were easy to code, and I had played enough Hitchhiker’s Guide and Zork to develop a passion for the genre. Using Claude, I made a game about publishing in international relations called Publish or Perish. This project turned out to be a ton of fun to make, as deciding what branches should look like, how deep the game should go, and whether to add more to it was captivating. The project itself is infinitely expandable and begging for more depth in the future, so it may be something I return to later.

With those two projects out of the way, I turned to my primary task: could it resolve my coding issues and help me recover this website? My first goal was to make a static version of the Quantitative Peace so that old links would work again. Claude was able to pull that off with some prompting and persistence. I uploaded that version of the website, and things were solid. Yet, I would rather have a fully functional blog if feasible. Phase 2 involved a bit more digging into the code and troubleshooting, but the site is back up and running as a working product.

This success led me to use Claude to revamp my website and update sections that had languished for over five years. While many of those tasks were small in nature, the number of updates became daunting over time, and having AI provide the code, formatting, and a list of items to add (or consider adding) was a helpful push in the right direction.

In sum: this website/blog exists again (which may be old-fashioned in the age of Substack). I do not know how research and journal standards will adapt to AI-assisted research in the near future. Some journals require scholars to submit a disclosure as to how AI was used, if at all, in the preparation of the manuscript, and that seems like a reasonable way to begin.