Great Things Rarely Start Pure
A great deal of energy is spent on the purity of intentions — on making sure a project begins from the right motives before it is allowed to begin at all. I wasted years on this. Purifying intention before acting is one of the most expensive forms of procrastination, and it disguises itself as virtue. There is even a vocabulary that launders it: stubbornness rebranded as consistency, sunk-cost fallacy experienced as steadfastness.
The historical record is not kind to the purity requirement. Computing and the internet owe enormous parts of their existence to war; without military use cases, it is hard to see how that generation of technology would have found its footing. Adult entertainment quietly funded much of the early internet’s infrastructure and drove adoption in ways no respectable institution would have championed. We can debate how significant those contributions were, but the pattern is hard to deny: some of the most consequential systems in modern life did not start good. They started useful to someone with unlovely motives.
Gaming tells the same story from the consumer side. Games were long dismissed as a waste of time, yet the demand for ever-better graphics pushed hardware and algorithms far beyond what “serious” computing required. Without gaming, there is no NVIDIA as we know it. Researchers borrowed games as experimental environments for decades; much of what modern AI labs build — world models, interactive simulation — descends directly from game-like setups. The “useless” thing turned out to be the substrate the useful things grew on.
Blockchain fits the pattern too, with an interesting twist. Today its dominant use case is speculation — gambling, essentially. But the underlying wager is coherent: decentralized systems matter most where institutions fail. A trustless ledger is nearly worthless under a competent, honest government and nearly essential under a corrupt one — which is why so much crypto enthusiasm is, at root, a bet against institutional integrity. Verifying that a state is actually trustworthy is itself enormously costly, so people hedge by building systems that don’t require trusting anyone. The moral valence of the whole enterprise depends entirely on facts about governance that nobody can fully verify.
Step back and two lessons emerge.
First, good and bad are mostly social constructs applied after the fact. Many things labeled destructive contain the seed of something constructive, and the classification often says more about the classifier than the thing. This is not an argument for cynicism or for abandoning ethics — it is an argument against letting moral bookkeeping veto curiosity. Insist that every beginning be pure, and you will simply never begin, because nothing important began pure.
Second, hold categories loosely. Things are rarely mutually exclusive; causes and effects tangle; the same phenomenon can be vice and substrate simultaneously. The practical discipline that survives this uncertainty is old and unfashionable: carry an umbrella even when rain seems unlikely. What’s the cost? Perhaps it shields you from the sun instead. Preparing for outcomes you can’t cleanly classify — morally or otherwise — beats pretending you can classify them.
Great things rarely start pure. They start somewhere messy, get used for purposes their creators didn’t intend, and become respectable only in retrospect. Judge them later, by what they made possible — but bring the umbrella.
Drafted with AI assistance from my personal journal (2025-08-07), then edited by me. See On AI Assistance.