Software Eats the Rest of the World

That forgetting is the most expensive thing in software, and nobody has ever counted it. We have all done it. Every one of us has wanted our tools to do one more thing, found that they could not, and quietly bent our work around the gap until the gap disappeared from view. The wish did not die. It went under.

On August 20, 2011, Marc Andreessen published an essay with a line that became the creed of a generation: software is eating the world. He was right. Over the next fifteen years software ate the bookstore, the video store, the taxi stand and the travel agent. But read the essay again and you find he also named the thing that would slow the meal down. It was not money or technology. It was people. He wrote that “every company I work with is absolutely starved for talent,” and his only remedy was education, which is to say, waiting.

This year the markets gave the opposite idea a name, the SaaSpocalypse, and marked down software companies as if AI were the end of the meal. I think AI is the cook.

The build line

Every wish for software has two numbers attached to it. One is what it is worth to everyone who shares it. The other is what it costs to write it, and there are really two costs inside that one. There are the hours. And there is translation: the distance between the person who knows exactly what they want and the person who knows how to build it. Anyone who has tried to explain their business to a developer knows that second cost. It is often the larger one.

Software exists only where worth clears cost. Payroll, shared by millions, clears it with room to spare, so payroll software is everywhere. Your rain-aware crew text, wanted by one company, does not come close, so it does not exist. That is why the software world has the shape it has: a tall center of tools that huge numbers of people share, and around it a vast, dark ring of wishes that were real and never built. We never measured that ring, because unbuilt software leaves no record. The workaround is its only fossil.

The software industry is not the size of our needs. It is the size of the part of our needs that could afford a programmer.

There is an old law about what happens when the line moves. In 1865 the economist William Stanley Jevons noticed that Britain’s steam engines had become far more efficient with coal, and that Britain was burning more coal than ever. Cheaper power did not mean the same work for less fuel. It meant a thousand jobs that had never been worth powering suddenly were. When something useful gets cheaper, you do not get the same amount for less. You get the uses that were never worth it before.

Andreessen’s wave was Jevons applied to running software. He noted that the monthly cost of operating a basic internet application had fallen about a hundredfold in a decade, thanks to the cloud, and a flood of companies followed. But the cost of writing software barely moved. It still needed a scarce person who understood both the code and your business. The line dropped for the companies that could hire that person. For everyone else it stayed where it was.

The cook arrives

Now the programmer is no longer scarce. AI can be the programmer, the integrator, the back end developer, the front end developer, the person who reads the documentation at two in the morning. And it collapses both costs at once. The hours shrink toward nothing. More important, translation disappears, because the person with the wish is now talking directly to the thing that builds. Nobody has to explain the business to someone else. The owner is the one in the room.

Here is what most people miss. The obvious prediction is that fewer engineers will be needed, because each one is now so much more productive. That is the coal-saving prediction, and Jevons already showed us why it fails. When writing software gets cheap, the dark ring lights up, and every point that lights up is a piece of software someone has to want, shape, check and own. We do not get fewer software engineers. We get ten times as many, then a hundred times. They just will not look like the engineers we know, and they will not appear in the org charts we know.

We have seen this once before, on a small scale. When the spreadsheet arrived in 1979, accountants, analysts and shop owners began writing formulas, and within a decade there were millions of them. A formula is a program. Those people were programming every day, and almost none of them were ever counted as programmers. The same thing is about to happen with whole systems: databases, apps, automations, agents. The title will not follow the work.

The person who gets this power has a name I have been using for a while: super admin. Not a user who edits the rows inside someone else’s software, but the one who can change the table itself. We have all had the intuition toward it. We wanted the stack to do this or that, and when it could not, we worked around it or forgot. That was the bottleneck that kept software from finishing its meal. AI removes it, which is why I wrote Can You Delete the Table? and the course Super Admin.

1. More software will be written this week than in all of history before it.

   Count programs, not products. Most of it will be small, used by one person or one shop, and some of it thrown away by Friday, the way most writing is. That is not a sign the software is unserious. It is a sign the line has fallen below the size of a single wish.

2. Software engineers will multiply by ten, then a hundred, mostly without the title.

   The census will miss them the way it missed the spreadsheet authors. They will be called owners, office managers, chefs, coaches and contractors. What they do all day will be engineering: deciding what the system should do, checking that it does it, and owning it when it breaks.

3. The workaround dies first, not the vendor.

   The sentence “the system won’t let me” will start to sound as dated as “the computer is down.” What disappears fastest is the sticky note, the double entry and the habit of not asking, long before any big software company does.

4. Rented software becomes the floor, not the ceiling.

   Large shared systems will become the foundation that personal software is built on: payments, records, identity, compliance. Each owner commands a layer on top that fits their business exactly. Some vendors shrink inside this shift. The total amount of software in the world grows enormously.

5. The scarce skill moves from writing to wanting.

   When building costs almost nothing, the bottleneck becomes knowing precisely what you want and being willing to take the seat. Clarity of intention becomes the new technical skill, and the people who have it will build circles around the people who are waiting for a vendor to notice them.

Where I sit

I have built software the old way. When Paul Stern and I started E2open, we were its first two employees, with venture money behind us and some of the largest technology companies in the world as founders on paper. Everything about how we worked assumed that writing software was the scarce, expensive thing, and that it had to be organized like a cathedral. That assumption was correct for its time.

It is not correct anymore. At Charleston AI, I watch local owners describe what they want in plain English and get working systems back: a reservation tool for a restaurant, a daily market scout for a custom car shop, a CRM a contractor owns outright instead of renting, which I wrote about in Own Your CRM. None of these would ever have cleared the old line. None of these owners think of themselves as engineers. All of them are.

How this could be wrong

Already true

A person with no training can describe a piece of software in plain language and get working code back the same day. Small businesses already run tools that no vendor ever sold them, built by someone who simply asked.

What has to happen

The unglamorous parts have to catch up: backups, permissions, security and data that does not leak. The big shared systems have to stay open to the personal layers built on top of them. And owners have to actually take the seat, which means treating their own wishes as worth building.

Where I am probably wrong

The deepest risk is not cost. It is attention. If most people, handed the power to build exactly what they want, prefer the rented default and never ask, then the dark ring stays dark for a new reason, and I will have predicted a hundredfold when the truth is threefold. And if you only count software someone uses for more than a day, “this week” may turn out to be “this year.” The test I will watch: whether a typical small business with no developer is running software it commissioned in its own words within two years.

Go back to the landscaper at 6:40 on a Tuesday, the note he made and forgot. That wish was never too strange or too small to matter. It was simply below the line, and the line has moved. The thing you wanted your software to do is still there, under the workaround, waiting for you to say it out loud.

It was never too small. It was only too expensive.