The Entrance Exam Is Gone.

The tools did not simply make the existing players faster. They started creating new ones.

This is not a story about developers getting more productive. It is a story about who is allowed to participate.

Six months ago, someone might have been poking at prompts, barely able to name the architecture under a first screen. Today they can have a product, a site, a payment flow, customers, and a paid campaign.

They have not acquired ten years of engineering. They have acquired something that used to require access to people who had: the ability to enter the market.

That is a redistribution of agency.

The six-month competitor

Six months ago, they were not in the industry.

Now they are buying ads.

That was the moment it landed. People promoting products they had built with a model. People who, not long before, would not have called themselves capable of making software. And now they were in the market.

Not studying for a qualification. Not waiting to be hired by somebody who already belonged. Participating.

Some of those products will fail. Some are held together by prompts, optimism, and an alarming lack of security. Some of their makers do not yet understand what they shipped. Karpathy named the careless end of this in February 2025: forget the code exists, accept all, paste the error back in, ship the weekend project. ^[1]

That is not the observation that matters.

The time between being an outsider and becoming a competitor has collapsed.

For decades, a technology business had to assemble the means of production before the idea could be tested. Expertise, funding, developers, designers, infrastructure. Even a modest product was a chain of people. Those dependencies were not only costs. They were a barrier to entry.

Then the first wave arrived, and the industry laughed. Broken apps. Sprawling generated code. No discipline. Often the criticism was fair.

While the argument was about whether they were real developers, they were becoming founders, designers, distributors, competitors. They learned by entering the market, not by asking permission to enter it. And the tools kept getting better.

Everyone is not an engineer

Two claims get collapsed, and they are not the same.

The revolution is not that everyone is now an engineer.

It is that engineering expertise is no longer always the prerequisite for having a software product.

Building something is not the same as operating it. A deployed application is not a business. Judgment still matters, especially where there are customers, money, privacy, or safety. Fowler's May 2026 distinction is the one worth keeping: vibe coding forgets the code exists; agentic engineering still reviews it, tests it, and owns what it does. ^{[2] [3]} The popular word ate both. Do not let it.

Shipping is also not success. Greta.sh published its own telemetry in August 2026. One platform, not a market survey. Roughly four in ten apps ever built there reached a public URL, and the report is explicit that a live URL is a low bar. It is not users, revenue, or a security review. ^[4]

The absence of expertise no longer stops the first meaningful step. The market is replacing the institution as the entrance exam. That changes who gets to have an idea, who gets to test it, and who gets to challenge an incumbent.

The next competitor may not come out of another technology company. They may be a teacher, an operations manager, a designer, or someone who spent the last six months discovering they could make things.

Meanwhile, inside the building

This is the contradiction.

Outside, someone can conceive, build, launch, and advertise with a laptop and a modest budget. Inside, a capable employee can still spend weeks waiting for a development environment.

The outside is removing the barriers to participation. The enterprise is still administering them.

New entrants are using the tools to invent products. A lot of established businesses are using the same tools to automate the processes they already have. One group is finding out what is possible. The other is making yesterday more efficient. That is the internal danger in The Bureaucracy Is About to Become Autonomous: extraordinary capacity, spent preserving the structure that already exists.

So this is larger than an adoption programme. It is a question of who is allowed to create, who is allowed to decide, and which institutional steps still earn their keep.

When the ability to build is no longer scarce, the liability is how hard you make building.

For decades, technology was an industry you had to qualify to enter. Now you can begin by making something. That is not the democratisation of engineering. It is the democratisation of participation in the economy built around it. The disruption may not come from what the existing experts can now do. It may come from the people who never had the chance to do it before.

Related: The Bureaucracy Is About to Become Autonomous, As Fast As You Can Think, and The Individual Contributor Age.