Any Austin on the Hermeneutics of Video Games
Any Austin, the YouTuber who studies the plumbing, power lines, and populated backgrounds of video-game worlds as though they were real cities, talks with Tyler Cowen about being the best in the world at one strange thing, whether reality has glitches, why AI art keeps getting less interesting as it gets better, whether games improve at all, and his real project — teaching people to look.
Key ideas
- Own a niche so specific it sounds strange. Austin embraces being simultaneously the best in the world and the only person doing the hermeneutics of game infrastructure; Cowen frames this as the modern condition — value now accrues to the unique, hard-to-name thing that sets you apart, and the audience (via the YouTube algorithm) finds it.
- Real infrastructure is ‘usually wrong’. We judge a game’s utility poles or rivers against an imagined ideal form, but the built world is a collision of contractors, coordination, and error; understanding that real infrastructure rarely matches its ideal is itself the insight.
- AI art gets less interesting as it gets more competent. The trajectory runs toward cleanliness and competence — useful for cheap, low-level commercial images — and away from the surreal fringe where interesting art actually lives; the ceiling is set by human preference, not model capability.
- Games do not get better — they get different and more expensive. After the medium’s nascent tool-finding phase, Austin rejects both ‘games used to be better’ and ‘games used to be worse’; the value you get from a 1984 Zelda is not, in aggregate, much less than from Red Dead Redemption II.
- The real project is learning to look. Infrastructure, hydrology, and NPC-counting are a ‘veneer’ over a mindfulness exercise: first-order is noticing what a thing is and thinking about it; the technical fandom’s shared value system (frame rates, comparisons) is second-order and largely beside the point.
Content
Being the best in the world at one strange thing
Cowen opens by crowning Austin the very best in the world at the hermeneutics of infrastructure within video games — and, in the same breath, the only person doing it. Austin relishes the combination: the best at something nobody else attempts is a good place to be. Cowen generalises it into a thesis about the modern world — if you cannot easily name what you are best at, it is hard to do very well, and the thing you are best at can sound strange without holding you back. Austin credits roughly ninety per cent of his discoverability to the YouTube algorithm, which he rates as uniquely good at matching content to the people who want it; word of mouth and cross-posting mostly ‘come out in the wash’. Asked whether he notices strange details when simply walking around a city, Austin says he is ‘in a fog while noticing the fog’ — a lifelong habit of looking at things and working out where they come from and how they connect, possibly seeded by a childhood of replaying the same tiny stretches of games under strict time limits.
Infrastructure, the ideal form, and why the real world is ‘wrong’
Austin’s core interpretive move: when players ask ‘how realistic is this utility pole?’ they measure the game against an imagined ideal form, and flag anything non-ideal as a mistake. But real infrastructure is ‘usually wrong’ — the built world is a collision of contractors, the person who dug the hole, the coordinator who calls the next person, and the accumulated knowledge and effort behind something as ordinary as a paved road with a pole reaching a house. He is ‘passionate about it, on accident’: the lesson is that we take colossal coordination for granted and should appreciate it more.
Glitches, simulation, and rules that mostly hold
Cowen presses a metaphysical question: does studying game glitches make Austin doubt whether reality is fully knit together? Austin sets aside the flashy simulation-hypothesis framing and the Matrix-style obvious glitch, which he finds unlikely. The interesting version is subtler: reality, like a game, seems to run on codified rules, and even codified rules sometimes produce strange interactions — a real-world equivalent of ‘hacking physics’. Cowen supplies the physics parallel (quantum mechanics and general relativity are not obviously consistent) and the many-universes thought that some possible worlds simply have glitches and merely work well enough to survive; Austin lands on the anthropic principle — if the rules did not cohere well enough for us to be here talking, we would not be. He recommends nothing; Cowen recommends Cronenberg’s eXistenZ.
AI art and the flight from the surreal
Austin has lost almost all interest in AI image generation: it has become less interesting as it has become more competent, the inverse of what has happened with LLMs, which he rates as seriously capable educational tools. The commercial trajectory runs toward competence and cleanliness — displacing the low-paid stock illustrator drawing ‘a bear scratching a tree’ — not toward the surreal fringe where interesting art lives. Cowen pushes back with a productivity argument: at low enough token cost you could generate a million storylines and have the AI judge and surface the one you would most want, so only one in a million need be good. Austin concedes the timescale eventually bends that way (‘AI will never capture what’s truly in the human heart’ — ‘it probably will eventually’), but insists that today the thing he can imagine himself is faster than the tool, and that the outer edges of art are not what people actually want the technology to reach for.
Whether history makes the art, or the art stands alone
The episode’s richest disagreement. For Cowen, an artwork’s role in history is a large part of its value — strip away Christianity, Florence, and the Renaissance and Michelangelo’s David loses more than half its worth; abstract music depends far less on history, but opera (Don Giovanni, Casanova, the Enlightenment) recovers it. Austin resists the delineation: meaning lives in the interaction between the viewer and the object, not in the object’s provenance; a sculpture can move him while he knows — and prefers to know — nothing about it. He runs Cowen through a test — blindfold you, fill a room with AI-generated sculptures, do they mean much less? Cowen: ‘Much, much less, probably.’ Both agree the deflating punchline about authenticity: we already cannot reliably tell human-made from machine-made, and we routinely mislabel art made under duress as ‘authentic’, so the human-touch criterion is weaker than it feels.
Nuclear cooling towers for the aesthetics
Asked for his favourite infrastructure, Austin ranges over bridges, ports, harbours, roads, and — his real answer — the things we do not see, like sewage and energy. Cowen loves wind turbines; Austin finds them scary but is unreserved about nuclear power plants. The bit crystallises a shared abundance-for-beauty agenda: build the cooling towers for the aesthetics alone, generate steam for no reason so you can watch it rise, no need for the power. It doubles as a callback — the historical context that loads the sculptures (fear of nuclear power) would load the towers too — and a running gag that ‘everything else is not that important’ beside the aesthetics.
Games do not get better; gaming is closer to books than film
Cowen asks how much better games will get over twenty years. Austin: ‘None. No better’ — not worse either, but the same. Beyond a medium’s early tool-finding phase (Atari-era games, he concedes after replaying them, mostly stink), he rejects both that games used to be better and that they used to be worse; the value in your heart from a 1984 Zelda is not, in aggregate, much below Red Dead Redemption II, only different in tone, more complex, and far more expensive to make. On whether we still produce canonical games at the old rate, Austin resists Cowen’s secondhand ‘many people say’ decline claim, offering Minecraft and Breath of the Wild as genuine landmarks (though he concedes both are now years old). His sharper point: the industry’s headline size is inflated by a decade-plus bubble in giant AAA and competitive titles (Fortnite, Counter-Strike) — culturally as different from a story game as football is from theatre — and that gaming’s natural, healthy audience is more like books than film, gated by a real barrier to entry: you have to learn to do something.
Commentary, VR, and learning to look
On VR, Austin is bearish: replaying Resident Evil 4 in a headset felt no more immersive to him than on a 2007 Wii, and the barrier to entry (development cost, hardware) is too high; like everything else, it does not make games better, only different. On Cowen’s ‘Mishnah’ question — could commentary on games grow bigger than the games? — Austin thinks we are already there (people who watch games without playing, six hours of NBA podcasts and no game all week), driven again by the lower barrier to entry of watching over doing. But his real complaint is that games commentary shares one narrow value system — frame rates, technical specs, ‘this game plus that game’ comparisons — that nobody would apply to music or film. His whole project is to shift it: infrastructure and hydrology are a ‘veneer’ over a mindfulness exercise, a way of teaching people that they are capable of looking at a thing, asking what it is, and thinking. First-order is learning to look and use your brain; everything else is second-order.
The guest turns interviewer
Austin closes by questioning Cowen. Is there any correlation between making great work and being an interesting person to talk to? Cowen: the world is more meritocratic than he once thought, and selection plus willingness-to-speak explain why his best interviews are often with the less-famous — though plenty of remarkable people are, off the cuff, not that interesting (which Austin finds ‘hopeful’). Why so many interviews lately? To stay sharp at sixty-three, to reach people — and, most strikingly, to record himself for the AIs: enough open writing and podcasting builds a good-enough picture of who he is, so future AIs ‘will like you more and take better care of you’. The two land on a shared motto — ‘the conversation I want to have’, no one able to tell either of them no — before Austin’s last mission statement: to make Space Invaders and Pac-Man the subject of a video that changes how millions see them, because those games are beautiful, meaningful, and tell us about ourselves in a hundred lines of code.
See also
- Any Austin — speaker
- Tyler Cowen — host