I can put a number on it now: about a week.
A project goes fully operational — the engine runs clean, the pipeline publishes itself, the loop closes — and within a week, on average, the questions run out and the steam goes with them. From there it splits two ways. Either the next wall has already shown up and I jump to it, or I go through a mental burnout pause until one does. Those are the only two exits I’ve ever taken.
Each project dies its own way. rpgCore was hard to let go of — a design lab I kept circling back to long after the circling stopped producing anything. OperatorGame was easier; it ran its course, and we both knew it. VoidDrift was the slow, gradual, painful one — a live game, real players, a sprint queued with all the fun stuff, and a creeping loss of interest I could feel happening and couldn’t argue with. There’s a specific grief to losing interest in a thing once it feels like it’s reached its natural conclusion. It’s taken years, but it gets less painful. The first ones were work projects, and those were the hardest to resolve — because with work projects, letting go and rebuilding are both on the table, and you have to pick.
I used to read all of this as a character flaw. Lack of follow-through, a discipline problem, pick your self-help framing. I’d force the polish phases, schedule them like chores, and watch them die anyway — not dramatically, just a repo going quiet. Then I’d feel bad about it, which helped nothing, until some new wall presented itself and the cycle restarted at full intensity like the last project never existed.
It took an embarrassing number of repetitions to see the shape. The pull was never the project. The pull was the *problem* — the moment an ugly specific friction reveals itself as a general class I can name and build. That recognition is rocket fuel, and it’s fully spent by the time the engine first runs clean, because the question — the only part I was actually there for — has been answered. What remains is content, polish, marketing: real work, valuable work, and work my engagement system treats as someone else’s job.
The surprise wasn’t the pattern. It was accepting that it’s structural, not moral. I don’t have a discipline problem any more than a sprinter has an endurance problem. I have an engagement curve with a known shape and a known duration — about a week of fumes past ignition — and for years I scoped projects as if I had a different one. Long polish tails. Content phases stretching months past the interesting part. Plans that required a person I’m not.
So now I design around it instead of against it. New projects get scoped so “engine works” and “shippable” land in the same week. Reskins of proven chassis instead of new engines when the mechanic already exists. Every certified build converts into its write-up and demo at the moment of certification, because I know exactly what “I’ll do it later” means coming from me. Polish only happens against a concrete external trigger — a date, an actual player asking — never as an open-ended phase.
Stop scoping projects for the person you wish you were. Scope them for the engagement curve you can actually demonstrate.
The relief is that the pattern stops being a verdict and becomes a parameter. The graveyard of quiet repos was never evidence that I can’t finish things — it was evidence that I kept defining “finished” as something a week past where the fuel runs out. The fix was never more fuel. It was moving the finish line.
Next is the purest test of the theory I’ve run: a themed reskin of the VoidDrift chassis, where the engine is already done on day zero and the entire project is the part I historically abandon. If the thesis holds, it ships inside the window. If it doesn’t, the pattern is worse than I thought — which is also worth knowing, and worth a post.
