When Everyone Can Build, What Is Left?

An old colleague told me something recently that would have sounded absurd five years ago.

He came from consulting and design, with no traditional software engineering background. Today, he manages a small team of AI agents. He speaks requirements into them, judges the output, pushes prototypes forward, and gets working software out the other end.

That is a market structure story disguised as a productivity story.

For the last generation of software, the old advantage was that you could build. You had engineers, budget, a roadmap, a backlog, a product manager, a sprint process, and enough capital to survive the distance between idea and release. That distance kept most people out.

The moat was partly taste and partly capital, but it was also brutally simple:

Making the thing was hard.

That distance is now collapsing.

For the deepest infrastructure, the hardest engineering systems, and messy enterprise platforms, reliability, security, integration, and maintenance still matter enormously. But for a large and growing class of software, the first version has become much easier to reach.

You can already see the cheap version in the wild. Electricians, builders, coaches, consultants, creators, and local operators are launching little coded funnels, booking tools, quote calculators, internal dashboards, and landing pages in weekend bursts.

Some are ugly. Some are obvious AI outputs. Some are barely software.

Some will work anyway.

Why? Because they are close to a real customer, a real pain point, and a real moment of need. They may lack perfect product craft, but they have something many polished startups lack: proximity to demand.

This raises the uncomfortable question.

If many more people can build, what is left?

Building was the old risk

The last generation of software process was built around an expensive truth:

Changing direction late was painful.

That is why teams talked so much about de-risking upfront. Write the requirements. Validate the user need. Design before engineering. Catch the wrong assumption early, because catching it after three months of implementation meant wasted salary, wasted opportunity, and a painful meeting with the board.

That world still exists.

But the centre of gravity has moved.

When the first version of a product can be generated in days, while the old version took quarters, the expensive part shifts. The new risk is whether the thing matters.

Is the workflow real?

Does the buyer trust you?

Is the timing right?

Do you have distribution?

Will anyone care once the demo novelty wears off?

AI moves product risk closer to judgement.

Ronald Coase asked why firms exist when markets can coordinate production. One answer was transaction cost: search, coordination, bargaining, enforcement, and trust. Software teams were partly internal machines for lowering the cost of making and changing digital work.

AI lowers one of those costs.

The model can write the code. It can generate the interface. It can produce the first version, the second version, and ten variations after lunch.

Customer truth, trust, timing, market memory, and the courage to kill a clever but distracting feature still come from human judgement.

The bottleneck moves.

Advantage shift

Slide the cost of building down.

As execution gets cheaper, the value moves into the layers people and markets create.

Cost of building

Expensive Old software world

cheap expensive

Build Can you make the thing?

Judgement Can you decide what matters?

Distribution Can the market remember you?

Status Does choosing you raise the buyer?

When building is expensive, execution itself carries status. Lower the build cost and watch the scarce layer move upward.

Taste Has to Mean Judgement

The fashionable answer is taste.

There is truth in that. When tools get easier, judgement matters more. But “taste” has become too comfortable a word. People use it to mean visual preference, brand polish, nice typography, or the ability to say “make it more premium” until the website looks like every other AI-generated website.

That is template selection with better lighting.

Real taste is compressed experience.

It is the ability to notice that a product is solving the wrong emotional problem. It is knowing which detail gives the buyer confidence and which detail only impresses the maker. It is feeling when a landing page sounds like it was written by a committee trying to sound brave. It is the discipline to make one promise clearly and cut the ten weak ones hiding behind motion graphics.

This is why prompt engineering is misunderstood.

People ask how to prompt better as if there is a hidden sequence: use a role, add a loop, demand three options, ask the model to critique itself, then press enter.

Those tricks help around the edges.

But the real prompt is the mind behind it.

What comes out of the machine depends on what you have learned to notice.

If you read the same articles as everyone else, borrow the same startup language, and ask the same safe questions, the model will give you a cleaner version of the same average thought. It may look polished. It will still be average.

Better prompting starts before the prompt.

It starts with reading outside your lane, collecting odd references, testing small ideas before you fully understand them, and getting comfortable with the possibility that the first version will be strange. Curious people already do this naturally. They follow a thread, hit a paper, try a tool, compare the result with reality, then ask a sharper question the next time.

That way of thinking develops slowly.

The prompt is judgement, curiosity, experience, and taste turned into language.

If your judgement is generic, the output will be generic.

If your assumptions are lazy, the model will decorate them.

Everyone gets the machine.

Fewer people have trained themselves to ask interesting questions.

The prompt exposes the advantage: a way of thinking precise enough to become language.

When products get cheap, status gets expensive

People like to pretend markets are clean contests of utility.

Better product wins. Lower price wins. More features win.

Sometimes.

But many markets are also contests of meaning.

Who looks credible?

Who feels inevitable?

Who makes the buyer look smart for choosing them?

Who gives a company, investor, worker, or customer a story they are happy to be seen inside?

Thorstein Veblen saw the status game hiding inside consumption more than a century ago. People buy function, visible position, association, seriousness, and a place inside a social story.

That logic gets stronger when software gets cheaper.

If ten teams can produce roughly similar dashboards, agents, apps, workflows, and internal tools, the buyer starts asking a different set of questions.

Which one works?

Which one feels safe to choose?

Which one will be remembered in the board meeting?

Which one makes me look forward-looking, credible, and protected?

Which one have my peers already heard of?

Which one carries the least career risk?

That is brand: stored trust, borrowed status, and the shortcut a buyer uses when the product space becomes too noisy to inspect from first principles every time.

In a noisy market, people reach for memory.

This is why familiar global brands keep paying to be seen around major events, broadcasts, sports, and cultural moments. They are buying attention, memory, and repeated association with importance.

The ROI often lives outside a neat spreadsheet cell.

The ROI is that when the shelf, search result, procurement list, or boardroom shortlist appears, the brand is already inside the buyer’s head wearing a suit.

Software is moving into the same world.

When making becomes easier, being chosen becomes harder.

Distribution is memory with plumbing

In the old software world, a company could hide behind the cost of execution for a while. If the product was hard to build, simply reaching launch was a signal.

That signal is weakening.

A working demo used to be proof.

Increasingly, it is the start line.

The question becomes who can earn repeat attention through memory, channel, and trust.

Ads, followers, SEO, partnerships, salespeople, newsletters, communities, app stores, marketplaces, procurement lists, and personal reputation are the plumbing.

Distribution is the ability to enter a market’s memory at the moment of choice.

Byron Sharp’s brand-growth work uses the language of mental and physical availability: being easy to think of and easy to buy.

That sounds simple because the hard part is hidden.

Being remembered is expensive.

Being trusted is expensive.

Being distinctive while staying serious is expensive.

Showing up long enough for the market to believe you will still be there tomorrow is expensive.

AI makes building cheaper.

Market memory stays expensive.

In fact, it becomes more valuable because the market is about to drown in adequate things.

This changes how you should look at companies.

Underwrite the thing around the demo.

Who owns the customer relationship?

Who has distribution before the product is inspected?

Who has proprietary context beyond the model?

Who understands the workflow deeply enough to know which features to cut?

Who has brand memory in the buyer’s head?

Who has an operator with judgement under pressure?

The dangerous company in the AI era is the one with a sharp understanding of the customer, a trusted path to market, and the discipline to turn cheap software generation into expensive market belief.

The product still matters.

But the product is one part of a wider argument.

The company becomes a judgement machine

Michael Hammer warned companies to rethink the process before automating old work. AI makes that warning sharper.

A company with weak judgement will use AI to generate more noise: more prototypes, more campaigns, more dashboards, more internal tools, more mediocre options to debate.

It will look busy.

It will feel modern.

It will become faster at producing things nobody was brave enough to reject.

A company with strong judgement will use AI differently.

It will test more ideas, but it will also kill more ideas. It will turn customer language into product decisions. It will make the brand sharper because the product promise is sharper. It will use software generation to shorten the distance between belief and evidence.

The scarce skill is decision.

Can you name the problem in your own language?

Can you see the status game inside the buying decision?

Can you build something that makes the customer feel more competent, more secure, more early, more respected?

Can you tell the difference between a useful product and a product-shaped argument the market ignores?

That is what separates builders from people who merely produce software-shaped objects.

What is left

So what remains when everyone can build?

Judgement remains.

Distribution remains.

Trust remains.

Brand remains.

Status remains.

The ability to turn lived experience into precise language becomes more valuable. The person who can describe the real constraint will get better output from the machine. The business that can describe the real customer will build better products. The founder who can describe the real status ladder will market better than the competitor still selling features.

Building continues.

Treating building as proof fades.

The future will have more software, more products, more agents, more landing pages, more workflows, more companies, and more noise.

The cheap thing will be making something that works.

The expensive thing will be making something the market chooses.

“AI lowers the cost of making. It raises the value of meaning.”

Johnny’s verdict

If your competitive advantage is “we can build this”, assume it is already decaying.

Ask the harder questions.

Who trusts you before the product is inspected?

Who remembers you at the moment of choice?

What status does the buyer gain by choosing you?

What private context do you understand beyond the model?

What judgement have you earned through lived experience?

When everyone can build, the winner is the person who knows what should exist, who it is for, why it matters, and how to make the market feel that before the feature list even starts.