The magic wand

TL;DR

A magic wand now grants any software wish you can describe, most in five minutes. It has four rules. It grants wishes, not wants, so you must know which five percent is yours. It works in one hand, never by committee. The first wish is always wrong, and using the thing is how you find the real one. And what you learn stays in the hand that waved it. Service as a software is a wand in the vendor's hand, and rule four means the learning goes to the vendor. Whose software the work becomes is decided by who holds the wand.

Suppose someone handed you a magic wand. Not a metaphor for a budget, an actual wand. You describe a piece of software and it appears. Most wishes take five minutes. The complicated ones take twenty. If you go hog wild, two hours. No ticket, no sprint, no vendor, no procurement. Just you, a sentence, and a thing on the screen that was not there before.

This stopped being a thought experiment this year. David Heinemeier Hansson, who wrote Ruby on Rails and spent twenty five years chiselling code by hand, told Lex Fridman it is simply how he works now. He has shipped an entire Linux distribution without writing a line of it himself. His word for the state it puts you in is delirious.

Who would not get delirious if suddenly a genie pops out of the bottle and says, "You can have whatever you want. Every feature you've ever dreamed of, I can deliver them to you, most of them in five minutes, a few in 20, and if we really go hog wild, it's gonna take me two hours."

I have been turning the wand over in my hands since I heard that, and I think it comes with rules. Four of them. None are written on the box.

Rule one. It grants wishes, not wants

The wand does exactly what you ask, which is the problem, because most people cannot say what they want. They can say what they dislike about what they have. That is a different skill, and it is the skill the wand exposes. DHH's first wish was small and it is the best illustration I know. He wrote in an app called Typora and reckoned he used five percent of it. So he asked for his five percent, had a first version in twenty minutes, and had abandoned Typora within two days. Everyone uses a different five percent of every product they rent. The wand makes your five percent an afternoon's work. But you have to know which five percent is yours, and most firms have never once been asked.

Rule two. It works in one hand

A wand has never worked by committee. Point it through a product manager, two designers, a VP and a CTO, all of whom must be in the room for the shaping because all of them must justify being in the building, and nothing comes out the other end. The magic is the bandwidth between one head and the machine, and every person you put in between halves it. DHH's own company learned this the hard way when it let a room of designers wave the wand at Basecamp together. Each wish was fine on its own. Together they wrecked the architecture, and humans had to mop it up by hand. Somebody still has to know the shape. Not carve every line, but stand where the sculptor stands and say the proportions are wrong.

A wand works in one hand. It has never worked by committee.

Rule three. The first wish is always wrong

I will admit to my own delirium. In the folder where this website lives there are eight HTML files that have nothing to do with the website. A koi pond. A bonsai. A kintsugi bowl. Petals on a river, which turned into lanterns on the same river two hours later. An origami swan that folds itself in WebGL, and a raked garden in 3D. They were made over five evenings in July, none of them for a client. Each began as a sentence and became a thing in a few minutes, and I kept going because the pond was not quite right, and then the light on the pond was not quite right. Only the swan made it onto the site. I am not sorry about the rest.

The agile people wrote this down twenty five years ago. Nobody knows what they want until they receive it. The wand turns that from a warning into a method. Wish badly, look at it, wish again. The thing on the screen after the first wave is never the software. It is the question you should have asked.

Rule four. What you learn stays in the hand

Five evenings of ponds taught me more about motion on a screen than any brief ever has, and the lesson did not go anywhere. It is in my hands now. The wand's real output is not the software. It is finding out what you wanted, and that knowledge belongs to whoever waved it. This is the rule that matters for business, because it is the one the market is about to get wrong.

Wishes off a shelf

There is an idea going around called service as a software. Vendors stop selling tools and start selling finished work. A resolved ticket, a closed set of books, a shipped feature. You buy the outcome and never touch the engine. The pitch is honest and the trillions behind it are real. It is also a wand in someone else's hand.

The case for renting the work rested on one fact. Building the engine yourself was expensive. Nine months, a team, a budget line, a steering committee. Renting was rational because building was not. The wand removes that fact. The engine that resolves your tickets is, at bottom, software that knows your returns policy, your escalation paths and your strange exceptions. Your five percent. Twenty minutes of asking, if you know what to ask.

And rule four bites here. When the vendor waves the wand over your work, the vendor learns your work. The wishes that go wrong, the second and third attempts, the slow discovery of what your firm actually needs, all of it accrues to an engine that belongs to someone else and serves your competitors off the same shelf. Wishes off a shelf are everybody's wishes averaged. That is a bargain for the firm with no taste. It is a slow transfer for the firm that has some, because it is paying to have its own judgement replaced by the mean.

The usual advice is that a five person shop should take the trade and not look back. I am no longer sure where that floor sits. If the owner can describe the work in plain English, and the owner is the one who knows it best, the wand is within reach of five people. Probably one.

The verdict

The work becomes software. That part is settled. Whose software is decided by whose hand is on the wand. The good is that the price of your own engine went from nine months to an afternoon. The bad is that the wand costs a new currency, attention, taste and one person who knows the shape and is allowed to say no, and most large firms are poorer in that currency than they think. The ugly is that none of it is sustainable yet. The people living this way run a dozen agents at once, buy a second subscription at four in the morning when the first runs dry, and say plainly that the pace cannot last. They are delirious and exhausted and would not trade it. Neither would I.

So my advice is smaller than the trillions. Pick one piece of software you rent and use five percent of. Describe the five percent in plain English, badly, and wave. Twenty minutes later there will be something on your screen that is not quite right, and that is the moment the work starts becoming yours. I say, try it.