AI writes script in seconds, sparking debate on coding jobs

AI generates in 30 seconds the script a programmer charged €400 for

A person with no programming knowledge had 7,000 compressed files containing folders, subfolders, text, and videos. They needed to extract only the videos to upload them to a streaming platform, as the server rejected other formats. Their initial plan involved paying several people to open files one by one for days, or hiring a developer to write a custom script. The estimated cost of the latter option, according to their own calculation: between €300 and €400, with a timeline dependent on the professional's mood. They solved it by asking a conversational AI for the code. Thirty seconds for the script. Another thirty to explain where the paths went and how to run it.

The episode, told in the first person, triggered a discussion of nearly two hundred replies about whether this is a labor revolution or an after-dinner anecdote. The author’s thesis is that generative AI has just given anyone the power to solve tasks that previously required hiring a third party. The tech sector’s response came on three fronts: that this is not programming, that generated code requires review, and that the work disappearing is the most repetitive, not the complex.

What that script actually did and why it isn't programming

The task consisted of traversing compressed files, ignoring text documents, and keeping the videos. A filter by extension within a loop. The comparison circulating in the discussion was that of the electrician: changing a lightbulb doesn’t make anyone an electrician, and watching a tutorial doesn’t eliminate the profession. Those holding this view add contextual data: thirty years ago, anyone with a computer knew how to compress files onto floppy disks from the command line. Today, you have to ask a machine, and it’s presented as a milestone.

This is the point that most unsettles technical profiles. They do not deny that the tool works. They deny that the merit is comparable. The line separating “using AI” from “programming” is the same as separating “using a calculator” from “being a mathematician,” they argue. And much of the debate hinges on this distinction.

Hallucinating code: why review remains mandatory

The most repeated argument among those who work with these tools is that generated code requires review effort. One participant summarizes: current versions are tools that do many things quickly, but they hallucinate and do not generate productive code without supervision. Another recounts having to test several versions of a Visual Basic script in Excel until it worked, drawing a conclusion from this: knowing how to talk to the machine matters.

That gap between what AI delivers and what the system needs is the terrain where the sector defends its longevity. It’s not that AI can’t write a file filter. It’s that the result must be integrated into a specific environment, with its particularities, and someone must take responsibility when it fails. The discussion drifted toward a deliberately extreme example: asking an AI to simulate in real time, at millisecond steps, a complete combustion engine, including temperatures, pressures, air and fuel flows, electronics, and mechanics. The answer was that very few programmers know how to do that. And that those tasks should be left to those who remain.

The precedent nobody wants to look at: tractors and construction

The historical comparison appeared early. One participant recalled that tractors didn’t send farm workers home; they became drivers, mechanics, and factory operators. Another countered with the construction sector, where mechanization did leave people behind. The discussion got tangled in whether previous changes were slow enough to allow an orderly transition and whether the timeline is compressed this time.

Some argue that the percentage of the workforce affected in a short time may be too high to absorb without conflict. Others respond with the argument that all automation generates resistance because it keeps more people employed than necessary. The discussion turned, in several stretches, into a clash between two ways of reading economic history: the one trusting in reabsorption and the one remembering that it didn’t always happen.

The broken business model: freelancers, dev factories, and clients who stop paying

The economic edge of the issue lies in the segment that lived off basic tasks. The author of the initial message states it bluntly: the freelance programmer who made simple scripts and charged for them loses that volume. The formulation used was that of the professional dressing up as a mechanic to rip off the client who doesn’t understand: “I have to make you a customized script, twenty hours, and a thousand euros.” If the client can get the script in a minute, that margin disappears.

The technical counteroffensive has two legs. First: AI makes the programmer more efficient, it doesn’t replace them, citing the example of development factories. Second: the freed-up work is dedicated to more complex tasks, and not everyone is suited for those tasks. The problem with that argument is that it excludes precisely the junior segment and the small-contract freelancer, which is where the hit is felt first.

The intern, the analyst, and the project manager

The most repeated prediction speaks not of disappearance but of selective replacement. AI can execute concrete tasks and will do so better over time, but it does not understand a complex problem involving many different tasks with a meta-problem to solve. Hence the conclusion circulating: it doesn’t replace a project manager or an analyst, but it does replace the intern given assignments that later need reviewing.

That hierarchy by complexity levels orders almost the entire debate. At the base, repetitive and well-defined tasks: automation advances there. At the top, open problems requiring judgment: replacement is not on the horizon. The uncomfortable question is how many jobs live at the base and what happens to them.

Productivity rises, the question is for whom

The broadest consensus, even among skeptics, is that productivity will increase. The discrepancy lies in distribution. If fewer people are needed for the same work, the logical consequence is less hiring, and hardly anyone disputes that. What is disputed is whether the savings translate into more projects, more profit, or simply fewer payrolls.

One participant added a derivative often overlooked: if developments accelerate, deadlines shorten, and environments change faster, with psychological and lifestyle implications for everyone. Estimulante ilegal isn’t free. Meanwhile, the author of the initial message announced he would save the discussion to revive it in five years. He believes laughter will be universal then. He might be right. So might those who replied that he himself works with a computer and keys.