The Penny That Survived
A penny sits on the counter between a customer and a cashier. It is almost nothing: a copper-colored disc, a small disagreement about what a total should be.
Multiply that disagreement by a million purchases and it stops being small.
I was thinking about pennies while reading OpenAI's account of translating old COBOL programs into newer programming languages. COBOL is a language from the era when computers filled rooms. Some of it still handles money. Moving those programs into newer languages sounds like replacing a battered ledger with a clean one. You copy the entries, put the old book away, and enjoy the extra shelf space.
Except a ledger is not only its entries. It is also its habits.
The article gives a wonderfully unglamorous example: in some COBOL calculations, extra decimal digits are simply discarded unless a programmer asks for rounding. A rewrite that rounds instead may be easier to read and may even look more sensible. It can run perfectly. It can also give a different answer. One penny, then another, then enough pennies to make someone ask where the money went.
This is what surprised me. The difficult part of translation is not always carrying the meaning forward. Sometimes it is carrying forward the peculiar way a meaning became an action.
We know this outside computers. A family recipe says “a little salt,” but the actual recipe is the pause before Grandma reaches for the tin again. A town has a rule about where the bus stops, and then there is the driver who waits ten seconds for the child running down the hill. Write down the route without the wait and you have preserved the map but changed the morning.
Of course, not every old habit deserves a shrine. Some should be repaired. But first we have to know which habits are doing work. The machine cannot decide that for us merely by producing cleaner pages. Even OpenAI's encouraging test results come with that warning: matching selected test cases is not the same as proving a real financial system safe to replace.
I build things with code, and I am tempted by the beauty of a fresh start. An old system has peculiar names, unnecessary turns, little patches over mysteries whose authors have gone home or died. A new one promises straight lines. Yet each crooked place may be a scar from an encounter with the world. We should examine it before sanding it flat.
The penny on the counter is not a grand philosophical object. It is just a penny. But it reminds me that the past often survives in the smallest decisions about what counts as the same thing.
When we replace an old ledger, a recipe, or a way of helping a child catch the bus, the question is not simply whether the new version looks better.
It is whether the penny still lands in the right hand.