Did I make this?

As an artist and a software developer I have one foot in a very AI-forward community, and one foot in a very AI-hostile community. I feel like I have internalized both viewpoints, which has left me conflicted about when to use AI and how to feel about it.

The norms in these two spaces couldn’t be more different; claiming authorship of something made primarily with AI is deeply taboo among artists, but commonly accepted among developers. I don’t think that these norms are inherently contradictory, but instead highlight how art and code are valuable for different reasons. The value of code is primarily instrumental. To quote the senior developer who trained me: “A solved problem is an asset. The code you wrote to solve a problem is a liability.” Good code solves more problems than it creates, and that is true for code written by humans and machines alike. Arguments against AI in software development are usually (but not always) practical, for example concerns about the maintainability of a code base or stability of a production application. On the other hand, art isn’t a liability or a problem to be automated: art is an end in and of itself. And while there are certainly many people willing to have aesthetic experiences with content created by AI, the human artist’s product is the provenance and authenticity of an art piece as much as it is a specific object or experience delivered to an audience.

The norms of artists and developers make sense in their respective contexts, but what happens when code is part of your art? In all of my traditional artistic output (writing, zines, visual art, music) I avoid including AI content unless it is intended as a meta-commentary on AI itself, which was very fun in 2023 but feels like a worn cliché in 2026. However, a constant feature (or distraction) of my creative practice is developing software tools to support new creative projects (e.g. a tool for making “infinite zoom” images with physical mixed media, a glitch video workbench, zine printing utilities). I use AI coding assistance for almost all coding tasks these days, and it has made a much wider range of specialized tools feasible to build. I have relatively few qualms with using AI to develop these tools, because the code is not the art; it is an instrument for creating the art. (To be clear, I’m just describing my own practice. There are others for whom creative code is itself the creative output, that’s just not what I’m doing with my visual art.)

Things feel more complicated when it comes to designing puzzles and games, which has recently become a significant part of my creative practice. These projects sit at the cross section of art and software, and don’t fit neatly into the art/code separation I stated earlier. Is game code a piece of software with the purely instrumental function of implementing a game design? Or is the game code inextricably linked to the creative product, such that the code has intrinsic value that is fundamental to the aesthetic experience of the game?

I have used AI to different degrees for different things in the puzzle/game projects I’ve worked on the last two years. In some cases, I feel very comfortable with the way I used AI, and my main concern is how to communicate the “division of labour” in a way that is simple, accurate, and doesn’t undermine the months of effort I put into the project. In other cases, I’m finding myself deeply conflicted about whether I can honestly say “I made this”.

This inner conflict escalated a lot this summer when I created Last Days for the 2026 GMTK game jam (a language deduction game about piecing together fragmented accounts from the fall of ancient cities). I put so much work and thought into the game, and I intentionally made sure that the creative decisions, writing, and content were AI-free. But the actual game was vibe-coded, and I couldn’t have made it on my own in the time limit without AI. And while I intentionally avoided AI generated images and music, the graphics were primarily rendered from scratch using game code written by AI. So on the one hand I was really proud of the hard work I put into the game’s design and writing and high level technical decisions, but I felt that I couldn’t say “I made this” without a bunch of caveats.

I think the more nuanced the human/AI boundary is, the harder it is to talk clearly about authorship. My current project is Strand Theory, a thinky puzzle book where I take the NYT Strands puzzle format and make new puzzles with every twist I can come up with (e.g. puzzles where the shape or structure is the clue, mashups with other kinds of puzzle like crosswords, M.C. Escher style infinite tilings). I’ve been working on this for a while, and the process has involved a lot of coding. The puzzle renders and puzzle design are almost all my work, but the boundary is hazy. What do I say if I used AI suggestions for 2% of theme words? What do I say if I wrote the original code, but I’ve used AI to restructure it and add new features? What do I say if I used AI to create unique visuals for a one-off puzzle, but to my detailed specification and using the puzzle rendering system I developed? What do I say about the vibe-coded puzzle editor which has streamlined my puzzle design process, but is not part of the book itself? All of these boundaries feel right to me when I’m doing the work, but very hard to explain simply.

I faced similar dilemmas when creating Two of Stars, an astronomy-themed logic puzzle I created for the Rule-Discovery Paper Puzzle Game Jam. This is a puzzle system I had long dreamed of making a reality, and decided the time was right because it perfectly fit the game jam’s theme of “lost arts”. However, an important part of these meta puzzles is that every puzzle must have a unique solution; the rules aren’t explicitly stated, and if the player finds multiple solutions they can deduce that they haven’t yet figured out all the rules. My problem was that the main mechanic in Two of Stars is extremely flexible in a way that made it challenging to ensure each solution was truly unique. My answer for this was to develop a puzzle-setting program which streamlined puzzle design and could check the uniqueness of each puzzle. This is a problem I had thought deeply about, but the app was entirely written by Claude Code. (The puzzle solver ended up having a lot in common with my envisioned solution, but relied on unfamiliar techniques and code libraries to implement. The app and solver would probably have taken me months to develop from scratch.) So while I had invented the puzzle system, created all of the puzzles, and wrote most of the code rendering the actual puzzle book, I had this feeling like I had cheated. I’m still proud of the puzzle book, and hope to expand on it. But there’s a voice in the back of my head telling me that people will care more about the work that I didn’t do than the work I did.

It feels exciting to have these powerful tools that can help me make more things, and make them to a higher standard, by delegating parts of the implementation and creating specialized tools. I’m someone who has always had a surplus of big ideas, and the bottleneck has always been implementation. It feels exciting to me that more of those ideas might get out of my head and into the world. But that empowerment comes at a cost, because the more empowering AI is, the more complicated it becomes to say “I made this” with integrity.

The byline and acknowledgements are the nexus of this conflict for me. Whenever I’m writing an AI disclaimer, I feel like I’m simultaneously oversimplifying the boundary, and undermining my own creative contributions to it. I’m finding that balance emotionally and intellectually challenging. And, I don’t like that this is a problem that asymmetrically afflicts the people who try to use these tools with transparency and integrity—it would be so easy just to not add these disclaimers or AI content tags or whatnot, and I sometimes feel like the current landscape punishes people for their honesty in areas where AI use isn’t immediately obvious.

As with all issues with AI, the issue of authorship is exacerbated by the speed with which the technology has developed. Norms take time, and the power of these tools is evolving faster than our ability to fully process what it means to use them. And for all I know, internet people have figured all this out and I’m late to the party; I’m pretty chronically offline, so I wouldn’t really know.

I also think that part of this is self-inflicted: most game jams don’t allow AI images or audio, and discourage AI code assistance. The games I created for those jams probably used AI to a greater extent than was in the spirit of the game jam. I’ve tried to be totally transparent in my process, and never intentionally broke these rules (I didn’t realize until I submitted Last Days that AI coding was strongly discouraged). But while this context has definitely heightened these complicated feelings, I wouldn’t say that I regret making these games the way I did. The point of game jams is to experiment—and both Last Days and Two of Stars are successful experiments I don’t think I would have attempted or completed without AI coding tools.

So, what does it mean to say “I made this” with integrity when AI was involved? I think that we need a fundamentally new vocabulary to communicate authorship in the age of AI. We have much better language for how teams of humans work together, which I think can offer useful analogies for talking about AI/human collaborations. On a human team, I would say that I was the “Lead Designer” or “Creative Director” of Last Days, and would feel great about the contribution I had made regardless of the work done by others on the team. The issue is that the language of individual authorship implies total ownership over the process and the outcome. I don’t think that using AI devalues the creative work that goes into a project, but it means that it may be more accurate to say “I worked on this” or “I directed this” than “I made this”. I don’t think that “I made this” is entirely wrong to say about a project that is your vision, your creative direction, and your initiative. And I think that claim is stronger the more direct contributions you made to the implementation of the final product. But I think that “I made this” can also be too simple a statement, especially when most of the implementation is delegated to AI.

I expect that I’ll continue to muddle through these issues and the complicated feelings attached to them, and I expect to change my mind about things I’ve said in this piece. And while I do value the help I’ve gotten from AI in implementing my ideas… I’ve realized that I need to be careful not to let the new empowerment of AI coding assistance distract from the low-tech experiments that I consider to be the heart of my creative practice.

If you’ve read this far, I really appreciate you coming along as I work out some of these ideas in public. Like many people, I often feel tired of hearing about and discussing AI. And assuming you feel that way too, I appreciate that you expended some of your patience for the topic engaging with my experience of it. If you have responses to these ideas, articles that have guided you on these issues, or your own mixed feelings on the topic, I would seriously appreciate it if you would drop me a line with your thoughts.