BUILD STORY
Yes, This Was Built With AI. Here Is What That Actually Takes.
I have seen the argument a few times now, and it is a fair one to make, so I want to answer it properly. If an AI wrote the code, then anyone could have knocked this out in an afternoon. I understand why that sounds true. It is not true, and the reason why is the most interesting thing about this project.
What AI actually gives you
AI will write you a drone flying around a browser in an hour. I am not going to pretend otherwise, and it is genuinely remarkable. I build with Claude Code and I lean on it hard. What it will not do is tell you that the drone is wrong.
It cannot tell you that a cinewhoop washes out later than that coming out of a hard turn. It cannot tell you that haze on a festival stage sits in layers rather than filling the room, or that those layers move when the pyro goes. It cannot tell you that a lighting rig at that height has a cable run under it, and that the cable run is the exact thing you clip when you get greedy on the exit. It does not know that the drop is where you commit, or what it costs you when you commit half a beat early.
I know those things because I do the job. That is the part of this that did not come out of a prompt, and no amount of prompting gets you there.
Prompting is not the method
The way most people use these tools is to ask for a thing and take what comes back. That works fine for a script. It falls apart by the fourth venue, because nothing you built is related to anything else you built, and you end up with ten levels that each look like they came from a different game.
So the venues here are not prompted into existence one at a time. They are assembled from a kit.
There is one shared set of parts underneath all of it. Truss, barrier, crowd, speaker stacks, moving heads, fog, and the materials that go on top of them. Every part carries settings that stay fixed across the whole game. How tall a barrier stands. How bright a fixture gets before it blooms. How far haze drifts and how quickly it clears. When I build a venue I am assembling it out of that kit under those rules, not generating a new one from a blank page.
The payoff is that improvement compounds. When I fix the way metal reads under stage light, it does not fix one venue. It fixes every venue that uses that material at the same time, and afterwards they all still look like they belong to each other. That is the difference between a pile of levels and a world.
Seeing the change is most of the work
The second thing is duller, and it matters more.
Every venue has a set of fixed camera positions built into it. Not gameplay cameras. Review cameras. When I change something I jump straight to the same handful of angles inside the real level, lit the way it is actually lit, and see what the change did within a few seconds. Same shots, same time of day, every time, so I am comparing like with like instead of trusting my memory of how it looked an hour ago.
That sounds like a small convenience. It is the entire quality loop. Without it you are editing blind and hoping, and hoping is how you end up with something that is technically finished and looks wrong in a way nobody can quite name. With it, I can make forty small corrections in an evening and know that every one of them was actually an improvement.
I did not get that from asking an AI to build a game. I got it from deciding how I wanted to work, then building the tooling to work that way. That part is engineering, and it is the half of this that gets left out when people say anyone could do it.
What it actually costs
Money first, because nobody says this part out loud. This is not a free hobby. Running the models costs real money every day, and the days I get it wrong cost exactly the same as the days I get it right. Then hosting, domains, payment processing, and the hardware I test on. It adds up quickly, and none of it is covered by a $9.99 game.
Time second. Nights and weekends around a job that regularly has me at a venue at four in the morning. This is what I do instead of watching television.
Patience third, and that is the one that actually stops people. Most of the work is not building. It is testing, flying it, finding the thing that feels slightly off, failing to fix it, fixing it, then finding that the fix broke something else. There are more than a thousand automated checks running against the game so that I find out when I break something, and I still break things.
The tools got faster. The work did not get shorter. It got deeper, because the bottleneck moved to knowing what good looks like, and there is no model for that.
Where it honestly stands
The flying is good. I am happy with the flight model, and so are the pilots who have flown it, and that is the part I care about most because it is the part that transfers to a real quad.
The venues are not finished. Some are close, and some of them still look built rather than shot. I know which ones and I know why, because I fly them, and a stage that would not survive a real broadcast does not survive in here either.
So I stopped adding venues. No new levels until the ones already in there are right. That was not an easy call, because new venues are the thing people ask for. But stacking another one on top of a foundation that is not finished is exactly how you end up with the kind of project people are right to be suspicious of. Everything is going into what already exists: the lighting, the haze, the geometry, the interface, the moment you spawn in and look around.
It is far enough along to be genuinely useful. If you fly events, the line of sight trainer on its own is worth your time, and flying a stage to a track you know is the closest thing there is to rehearsing the job. It is not far enough along for me to call it finished, and I am not going to call it finished until it is.
I do not shelve things
Here is the thing about me that matters most here, and it is not technical.
When I start something, I finish it. I get obsessive about it and I do not put it down. There are a lot of projects being built with AI right now, and most of them will be abandoned inside a month, because the first 80% is fast and fun and the last 20% is neither. I know that. I am in the last 20% on purpose, and I am going to be here a while.
Come in at the ground floor
This is the beginning of it, not the end of it.
If you fly, come and fly it. The Mainstage and the line of sight trainer are free, and they stay free. Then tell me what feels wrong on the sticks, or which venue looks fake to you and why. Be specific rather than polite, because specific is the part I can use.
The pilots flying it now are shaping what it turns into. That is not a marketing line. It is just what has happened every single time somebody has told me something was off.