Every piece of software you have ever tried to build came with a list: install these fourteen things first. That list is an apology. The machine that built the program knew exactly what it needed, down to the last piece - and that knowledge was thrown away, and a paragraph of instructions written in its place, asking you to rebuild it from memory and guesswork.
What went wrong, a long time ago
We call a piece of software by a name someone chose, and then treat the name as the thing. Rename it or move it, and everything that pointed at it breaks. We say two parts are compatible by comparing version numbers on whole components, when the truth lives in which few functions each part actually touches. And we seal everything into boxes - files, packages, programs - and then spend a great part of our working lives making the boxes fit.
Nobody designed this. Early computers read one file at a time from a reel of tape, so the file became the unit of everything, and sixty years of tools were built on top of that accident.
Imagine your identity were your addressMove house and you become, legally, a different person. Everyone who knew you must be told, one by one, or they lose you for ever. We would call that an absurd way to run a society. It is exactly how we run software.
Identify things by what they are
Give every small piece of software a fingerprint, computed from its own contents and from the fingerprints of everything it relies on. Now a thing is the same thing wherever it lives and whatever it is called - and a different thing is different, however alike the name. Names stay, as labels anyone may change. Nothing depends on them any more.
Almost everything else follows without being designed. There are no versions to clash, because you hold exact things. There are no copies, because two identical things were never two. There is nothing to install, because what a program needs is known exactly and arrives the moment it is touched. And when you change one line deep inside someone else's work, you get your own version of that one piece - while theirs, and everyone who uses it, stays exactly as it was.
Sam wants the chime to stopSam is a nurse, not a programmer. On night shifts their messaging app chimes for every message, and it has no setting to chime only for the ward. Today that is not difficult - it is impossible: the code that decides it is public, and might as well be on the moon.
Here Sam types notification sound, sees the very thing that decides it, and finds that eleven thousand people already use a version that chimes only for chosen groups. Two clicks. Sam did not become a developer. Sam changed a setting that happened to sit one step deeper than the settings screen goes.
One move, repeated
Which version is this? What does it need? Does it need rebuilding? Do these two conflict? Who is affected by this flaw? Is that advice still true? Today every one of these is a guess, and each guess costs an ocean of human hours. Here every one is a calculation.
That is the whole idea: replace a guess with a calculation. Everything else is engineering.
What is still hard
Nothing here needs inventing: each part already runs somewhere, built by people solving other problems. The danger is not the engineering. It is that the first person to use it must come out ahead on the very first day, alone - or nobody ever will.
And some things stay human. Someone still has to notice the bug. Someone still has to explain why.
Where it stands
The idea is written down whole, and its first step is planned: a small tool, Hypha, that reads real code, fingerprints every piece, and proves a piece can be lifted out and still run. Nothing is built yet. It is kept ready at the edge of our other work - which runs into the old ways a little more every day.
The full text is the manifesto in this project's repository: Stop managing software. Start growing it.