“Games are grotesque.”
So begins Ian Bogost’s essay The Squalid Grace of Flappy Bird, which remains lodged in my brain from when I first read it a decade ago. I was working in mobile gaming which was then still ascending towards the height of its power—before the Kate Upton Game of War Super Bowl ads, before Tencent’s $8.6B stake in Supercell, before Activision’s $5.9B acquisition of King.
Bogost’s piece was about Flappy Bird, a game featuring “a bird so cute as to signal deformity” that briefly catapulted into the public consciousness for a couple weeks, prompting its creator, Dong Nguyen, to remove it from the app stores.1 I remember reading and rereading The Squalid Grace of Flappy Bird (hereafter referred to as TSGoFP), trying to understand what Bogost was going on about:
To understand Flappy Bird, we must accept the premise that games are squalid, rusty machinery we operate in spite of themselves. What we appreciate about Flappy Bird is not the details of its design, but the fact that it embodies them with such unflappable nonchalance. The best games cease to be for us (or for anyone) and instead strive to be what they are as much as possible. From this indifference emanates a strange squalor that we can appreciate as beauty.
Immediately after this passage, Bogost compares playing Flappy Bird to fixing his bathroom cabinet drawer handle, and then later to a meteorite crashing through a desert motel lobby, “hot and small and unaware.” (Really, if you haven’t yet, stop and go read the whole thing. It is one of my favorite pieces of game writing of all time.)
I was extremely confused at the time, but the idea that “games are squalid, rusty machinery we operate in spite of themselves” stuck with me. And I occasionally return to reread TSGoFP as if it were a Zen koan, hoping to unearth some hidden insight.
So it is today—ten years later—that I think I finally understand what Bogost was trying to say, as the specter of Flappy Bird suddenly reappeared unprompted in my mind as I find myself thinking about the latest “viral” trend in gaming: AI.
AI is the unavoidable buzzword in every industry, and games are no different. There is no shortage of think pieces on how AI will usher in a revolution in gaming. And AI will undoubtedly disrupt the industry, like it will every software-based industry. From using generative AI to speed up the creation of concept art (see Blizzard Diffusion), to the spread of coding co-pilots, to AI-based 3D models and animations, AI will make it easier, cheaper, and faster to produce games.
But there is a vast difference between the tools used to create games and the technology inside the games themselves.
For the former, if it’s cheaper or more efficient, it’s a no-brainer to use AI (and many studios already are for things like concept art generation). I have no qualms with AI-enhanced game development, and you’d be silly to think it won’t happen.
For the latter it’s instructive to look at the practical history of technology inside of gaming: Games are usually viewed as “high tech” because they push the boundaries of what is possible for things like 3D graphics or networking, but these are (in Clayton Christensen’s term) “sustaining innovations”. They are linear increases to performance: graphics get more realistic, worlds get bigger, player counts go up. But games do not usually embrace brand new tech in order to try and find some new product-market fit, unless there is some new distribution advantage attached.
Consider the last decade of most popular games: League of Legends, Fortnite, Roblox, endless Call of Duty sequels. None of the biggest games of the recent decades adopted any disruptive technology. Improbable has been promising to enable new forms of gameplay through new multiplayer server tech since 2012 with nothing to show for it. VR growth has stalled out. AR has one success story in Pokemon Go, and a host of failed games, including Harry Potter: Wizards Unite from the same developer, Niantic.2
The only tech disruptions that were successfully embraced came with some sort of distribution advantage: new platforms like Facebook (e.g., Zynga), and mobile, (e.g., Supercell, King, etc), or new business models to reach more players like F2P (e.g., League of Legends on PC, and also ultimately Facebook and mobile).3
Game developers (especially the large ones) are conservative when it comes to tech. AAA games are already massively risky undertakings—they have all the problems that plague traditional large software projects (Baldur’s Gate 3 had a team of 400), plus compounded risk due to art, IP, and game design, and top of a development cycle that is measured in years. Why make things worse by stacking on additional tech risk? If you don’t have a clear view on how some new disruptive tech improves your distribution, you had better be damn sure the benefits to the game mechanics justify the risk. And the truth is no game developer can accurately judge this at the start of a project—it’s only through extensive trial and error that games are honed into something fun.
To make things worse, the long development cycle means developers are also remiss to change the tech stack partway through development. AI is evolving quickly right now, what studio wants to slap in some unproven LLM gaming model that’s likely going to be outdated in six months in the middle of a five-year timeline?
And then there is the cost: much has been made of how AI agents will revolutionize gaming, but given the astronomical costs of training and running models today, are LLM-powered agents so much better than existing game AIs or NPCs that they justify the significant increase in cost? If even AI companies are struggling to turn usage into profits, why would a game developer choose to spend their budget here? The games are expensive enough—GTA 6 has a reported budget of $1-2B—studios don’t need to take on more cost as it is.
This is not to say there will be zero new AI-enabled game experiences. Some indie studio will build something interesting with AI (e.g., AI Dungeon), but we are a long way out from a massive new AI-powered hit, despite what some VCs might believe. Too many starry-eyed calls for the AI gaming revolution are being written by people who have not actually worked as game makers. Again, AI-enabled tooling and development will happen quickly, but the stuff inside of games will take much, much longer.
So what does any of this have to do with the squalid machinery of games?
The lesson of TSGoFP is that games are imperfect (Bogost would say “broken”) experiences, and that simply does not matter.
That a game like Flappy Bird can go viral is instructive: a game can be the most squalid machine of all, and that can have zero bearing on its success. (Fall Guys is deliberately squalid.)
When I first read TSGoFP all those years ago, I remember being annoyed at the idea that games were stupid. How could Bogost—who himself is a video game designer, and who taught game studies and digital media at the Georgia Institute of Technology—be so condescending towards his own medium? I realize now that the stupidity is a feature, not a bug, and is something that makes games unique:
Flappy Bird is a game that accepts that it is stupid to be a game. It offers us an example of what it might feel like to conclude that this is enough. That it’s enough for games just to be crap in the universe, detritus that we encounter from time to time and that we might encounter as detritus rather than as meaning.
If I could snap my fingers, I’d love to live in a world where we have affordable, unbelievably immersive games, powered by AI agents, capable of creating a Ready Player One-esque experience on day one for the Apple Vision Pro. But we don’t yet live in that world.
For now, our squalid little machines float on as crap in the universe, and that is more than enough.
Thanks to Sebastian Park for reading and providing feedback on this piece.
You might recall when iPhones with Fortnite installed popped up on eBay for sale after the game was forced out of the App Store, but Flappy Bird did it first.
Pokemon Go achieved success via novel gameplay enabled by AR, but given that no other games from Niantic have succeeded (even with a massive IP like Harry Potter), Pokemon Go is more a reflection of the power of the Pokemon brand (see also the demand for Palworld) than a desire for new forms of gameplay enabled by AR.
What’s more, Niantic is highly atypical for a game studio in that it started as an internal team at Google in 2010 led by John Hanke (founder of Keyhole, which was later acquired by Google and the basis for Google Earth and Google Maps) and was only later spun out in 2015. Most studios don’t have this luxury.
The lack of new distribution channels or new business model also partly explains why new technology that was embraced elsewhere in tech like HTML5 and the resurgence of JavaScript (node.js, React, etc) has barely made a dent in gaming. Crypto, for all its faults, at least hits the “new business model” aspect of technology, and studios have been much more excited to adopt web3 because of it (even if gamers are not so keen).