Demolition is the most honest word for what I did, and I have used it deliberately. I did not drift out of 3D design, and I did not fail at it. I tore it down while it was standing, while I was good at it, and while it was still paying me.
That is the part people find strange, so it is the part worth explaining first: the dream was to be able to make the things I could picture, and 3D was only ever the vehicle. I picked it when I was young and did not know many vehicles. It was a good one for years. Then I priced it properly, and it could not get me where I said I was going. Confusing the destination with the vehicle is how people spend a life servicing something they chose at twenty-two.
Pricing the thing before you defend it
I got properly good at 3D. Modelling and rendering, the whole chain from an idea to an image convincing enough that someone would sign off on a building because of it. Real skill, real years.
The version people see is that output. The version you live in is the cost structure, and mine stacked up in four layers.
The fee did not reflect the skill. Rendering is priced as a deliverable rather than as expertise. The client is buying an image, and images have a going rate set by whoever is cheapest and adequate anywhere on earth. Being significantly better than adequate moves that rate less than you would expect, because the buyer often cannot tell the difference until the building is finished, at which point nobody is paying for images any more.
Revisions had no end condition. This is the one that grinds. Software work has a definition of done you can write in advance, argue about, and check against. Visual work is signed off on taste, and taste moves. Make the light warmer. Now it is too warm. Try the previous one. Each round is hours of your life, none of it is billable in any real sense, and there is no test you can run that says the work is finished. I have written since about why defining done is the leverage point in any work, and I first learned it as its absence.
The hardware was a subscription with extra steps. Render time is bought with a workstation, and a workstation stops being competitive on a schedule set by other people. Everyone else upgrades, everyone else’s turnaround shortens, and the client’s expectation follows the fastest supplier rather than the average one. So you re-enter the race every few years, and the money for it comes out of the same fees that were not moving.
The software billed monthly regardless. A slow quarter does not produce a discount. Fixed cost against variable income is a bad shape in any business, and a particularly bad shape for one person.
None of these on their own is a reason to tear anything down. Together they describe a structure where working harder produced more work rather than more leverage, and where a good year and a bad year sat on the same fixed floor: a ceiling, and a fact about the vehicle rather than about you.
The renovation I tried first
The first move was a renovation, because that is what everyone tries first: keep the structure, change its use.
I went sideways into construction — detail drawings, the documents that tell someone how to actually build the thing. The geometry knowledge was the durable part of what I had, and here it was structural rather than decorative. A detail drawing is an instruction that a person with materials will follow.
What I found there ended the argument, and it was not what I went looking for.
Every line I drew was somebody spending money. A line on a DWG is concrete, or a week of a crew’s time. Change your mind after it is built and you are demolishing.
So the thing I had actually been chasing — the ability to make what I could picture — was limited by capital. I could design anything. Realising any of it took thousands of dollars per idea, and somebody else’s permission for every one of them. The skill had never been the binding constraint, and I had spent years sharpening it anyway.
That is the finding that turns a renovation into a demolition: the problem was in the foundation, whichever room I was standing in.
What made the decision urgent rather than interesting
Plenty of people notice their ceiling and stay anyway. I nearly did. What moved it from an observation to a decision was the arithmetic of how many attempts a person actually gets.
Getting genuinely good at something takes years. Finding out where its ceiling sits takes a few more, because a ceiling is invisible until you are near it. Which means a serious attempt at a domain is close to a decade, and a working life holds a small and countable number of those, few enough that the number should frighten you slightly.
Once that is the unit, the sunk-cost argument inverts completely. “I have already put ten years into this” is the strongest possible reason to stop, because the ten years are gone either way and the only live question is what the next ten buy. Continuing doubles the investment.
The years I had left were the asset, and they were the only input in the entire equation I could not get more of.
Rebuilding on the destination instead of the vehicle
Generative design was the bridge, and I did not have to go looking for it, because it was already the half of the work I found most interesting.
Computational design is coding: you write scripts that generate geometry, define constraints, run loops, read results. The people doing it well in architecture are programming, and the barrier that kept it rare in ordinary practice was economic rather than cultural.
So the change that looks like a career switch from outside was, from inside, following the most interesting part of the work and noticing it led somewhere with a different cost structure attached.
The cost structure was the whole point. In code, the distance between an idea and the working thing is time and attention. There is no crew, no materials order, no line that costs money the moment it is drawn. Being wrong costs a rebuild rather than a demolition, which means you can be wrong far more often, which means you get to try things that are unlikely to work — and unlikely-to-work is where most of the interesting outcomes live. That single difference is worth laying out properly, because it explains most of what separates the two disciplines.
The destination had not moved at all. Make things I can picture. Have that ability not be rationed by what I can afford to build. Every one of the four costs above was a tax on that, and none of them were laws of nature — they were properties of one particular way of delivering it.
What survived the demolition
Almost everything, which surprises people who assume a domain change is a reset.
The geometry is in daily use, and not as nostalgia. Thinking in three dimensions, and reasoning about constraints and tolerances and what happens where two things meet — that goes straight into infrastructure work, and it is not evenly distributed among people who arrived through software. Neither is the habit of being right before you commit, which construction beats into you because it charges thousands of dollars per lesson.
What I demolished was the label and the cost structure. What I kept was everything that made the label worth having in the first place. The identity was the expensive part, and it is expensive for everyone — “I am a 3D designer” is a sentence that quietly answers your next several decisions, and it keeps answering long after it stops being right.
Three generations of my family each worked under a constraint they could not see from inside it, and the constraint moved every time. Mine was the price of realising an idea. Once it was visible, staying was a choice rather than a career.
I would rather demolish something I built than spend the remaining attempts maintaining it. That is arithmetic about the future, and the arithmetic only runs one way.