On Thursday 12 March, Frederick Vallaeys of Optmyzr took to the stage at Friends of Search in Amsterdam. He recounted an anecdote about his neighbour in Silicon Valley — someone who works at OpenAI.
That neighbour described what the workplace sounds like these days. In the past, a room full of developers had one sound: the tapping of keyboards. Now it sounds different. Developers speak their prompts aloud. Out loud. All at once.
No more tapping. A cacophony of voices.
That single image says more than a slide deck full of charts. The way software is made is changing — not gradually, but now.
And Vallaeys went further. He explained how his team works: they send the same prompt to four AI agents simultaneously. Four different outputs. They pick the best one. The rest is discarded.
Not one developer thinking and typing. But one person directing four agents — and deciding which output is worth keeping.
That is not workflow optimisation. That is a fundamentally different way of building software.
From permanent code to throwaway code
For twenty years we treated code as architecture. You built for the long term. Refactoring, technical debt, maintenance — that was part of it, because writing code was expensive and slow.
Vallaeys mentioned it in passing: "throw away code."
Take this example: you build an application for a specific purpose. It works. You use it. And then: delete. The next need calls for a new prompt and a new application.
This is not Software as a Service. This is Software on Demand — you ask, it appears, you use it, it disappears.
Sam Altman said this before. But when the CEO of a company that has built twenty years of codebases says the same thing, it is no longer hype.
Lovable, Bolt, Flutterflow: it already exists
These tools exist. They take your text and turn it into working applications. No API documentation. No boilerplate. No waiting.
They still feel like toys. Until you realise they remove the bottleneck.
Example: a marketer has an idea for a lead magnet. In the past that meant: a dev request, backlog, sprint planning, maybe live in two weeks. Now: type a prompt, have lunch, go live, iterate.
That is not a small step.
Where is this heading?
No wild predictions, but what I expect over the next 12 months:
1. The developer becomes a director
The developer role does not disappear. It becomes one thing: director.
Not the person who types, tests or maintains. But the person who decides what gets built, asks the right question before the prompt, and decides which output is worth keeping.
The rest — reviewing, testing, managing legacy — is either automated or disappears along with the codebases that require it.
2. Small teams gain ground
Today, small companies are barely able to build software (opent in nieuw venster). Too expensive, too slow. Soon one person will build three applications per week, each targeting a specific segment. No-code is already outdated — this is no-friction code.
3. Code becomes a commodity, quality of thinking does not
When everyone can generate code, the value no longer lies in writing it. It lies in: what you ask before your prompt, which output you discard, how sharply you assess the result. The builder becomes a curator.
4. Large organisations fall behind
Too much legacy, too much risk aversion. That gives fast, small teams a window of 18 to 24 months to get ahead.
5. Regulation is coming
No git history, no code review trail, no "who wrote this." Audit departments will start asking questions. GDPR teams too. Regulation slows adoption, but does not stop it.
What this means for you
You said last night: "I'm not a coder, but I want to build things myself now."
This is precisely the moment for that. Twenty years of insight into how teams and software work is now enough to build something yourself that goes live next week.
The barrier does not disappear entirely. But quickly enough that it truly matters.
Vallaeys already said it: you are no longer dependent on developers who type. You are dependent on what you ask and how sharply you assess.
The uncomfortable side
I saw it in the room last week: disbelief. Caution.
Understandable. Developers have spent twenty years building their identity around writing code (opent in nieuw venster). This feels like an attack on that.
But photography did not replace painting either. It changed what painting meant.
Developer discipline does not disappear. It shifts from execution to thinking. The best developers in 2027 are not the fastest typists, but the sharpest thinkers.
Start here
Take something you need today. Something that keeps getting shelved because it "costs too many dev hours."
Open Bolt or Lovable. Write your prompt. See what you get. Iterate four times. Choose the best version. Discard the rest.
You are not a developer. You are a curator of possibilities (opent in nieuw venster).
Do you know someone who has been having software built for years and now wants to start themselves? Pass this on. Or respond: what does Software on Demand mean for your work?