The coding job market is splitting: AI cracks junior employment

AI promises five times the output and leaves juniors out in the cold

A technician with a vocational degree in multiplatform application development is clear about it: with a well-structured prompt, AI hands him reliable code and cuts his project times down to a fifth. Another profile, with a degree in management informatics, qualifies the miracle: at best he became twice as productive, and in exchange had to review and fix a lot of code. Between the two accounts slips the question running through the sector: if output is five times faster, are five times fewer programmers needed, or is five times more software being built?

The answer, as of this discussion, is not settled. What does emerge clearly is an uneven distribution of the impact. Profiles with years of experience and job stability are holding up; those finishing their studies or recently laid off face a market where most IT job postings demand three to five years of experience, technical tests, psychological assessments, references and a portfolio. AI has not eliminated programming, but it has moved the ground under the feet of anyone trying to get in.

Productivity that doesn't add up

The dominant narrative says AI multiplies output. The numbers circulating in the conversation range from double to a fifth of the time. That range is not a detail: it completely changes the diagnosis. If the saving is fivefold, demand for new programmers collapses; if it is twofold, the effect gets diluted among the reviewing, fixing and architecture work the machine cannot solve.

Some argue the big difference lies not in writing code but in diagnosis. Hard-to-fix bugs and assistance for junior profiles —the part of the job that used to eat up the hours— disappear because AI detects failures with a reliability described as brutal. JavaScript components with multiple edge cases and events come out at a estimulante ilegal that used to take a full working day.

The flip side is less publicized. One programmer admits he tended to overcomplicate the project when leaning on the tool, stopped analyzing on his own and got stuck in loops. AI generates code for projects that, as they grow in complexity, it cannot handle itself. The result are monsters that someone will have to maintain with funding from a client who believed the matter was solved.

Why do job postings ask for three years of experience when medicine asks for none?

The labor market paradox is hard to swallow. IT postings demand three to five years, technical and psychological tests, references and a public repository. In healthcare, even for physician positions, not even one year is required. The comparison, raised in the conversation, is not meant as consolation: it describes a sector that has stopped training those who come after.

The effect on small businesses is ambiguous. A design and development firm noticed a drop in work in the first half of the year and now sees a strong recovery. Its diagnosis is pessimistic: a monumental shakeup is coming, and lots of digital product will be garbage until AI learns to program and organize itself. On the web, the result can look good visually and be a disaster in the code, forcing you to steer the tool through sheer persistence.

The other side is the falling cost. A CRM that lived in spreadsheets because its owner could not afford custom development has received offers far below previous ones. Software is reaching people who previously could not afford it. That multiplies projects, yes, but it also multiplies competition among those who build them.

From typing code to supervising the machine

The idea that the job consisted of typing is a ghost. For a long time the value has been in architecture, database modeling, optimization and new antiestéticatures. What changes now is the recipient of those decisions: AI executes, the human chooses requirements, components and strategy. And there appears the risk flagged in the conversation: the machine will make mistakes on requirements, pick more expensive or misaligned pieces, and do so with the confidence of someone who never looks back.

The most cited scenario is not mass layoffs but concentration. The top profiles will remain, those who have spent a decade building things at home on weekends and deliver a masterclass when you ask them a question. The ratio being thrown around is brutal: one in twenty. The rest, onto something else.

And those who learn with the tool without understanding what it generates? The warning is clear: if free tiers are closed or made more expensive, anyone who only does vibe coding is left without a trade and without foundations. The craft of programming from scratch, that mental agility lost to the magic of autocomplete, may regain value when the bill arrives.

---

With these materials, the market is not dead: it is split in two. One side concentrates stability, salary and judgment; the other piles up degrees, personal projects and technical tests for a job that no longer exists. How long will it take cheap software to generate the maintenance demand that brings programmers back to the table?