Learning to Code with AI: Commodity or Edge?

Learning to Code with AI: Commodity or Edge?

Learning to program while using AI is still worthwhile, but for different reasons than before. The core question is straightforward: no one wants to spend years mastering a skill that a machine can execute in minutes. The doubt isn't technical, it's economic: does this skill still offer a return when AI rapidly devalues its basic components? The answer circulating in the discussion is uncomfortable yet nuanced: yes, but what part of the profession gets paid has changed.

What Pays Now vs. What Doesn't in Programming

The consensus is that AI acts as a copilot, not a replacement. It generates blocks, solves simple problems, and returns code that usually requires correction. According to several technical profiles, what it doesn't do is decide architecture, select technologies, deploy software, or translate ambiguous client specifications into something that survives two months in production. There is still significant human work there.

The practical consequence is that specific syntax—whether line 312 needs a semicolon or colon—is no longer the bottleneck. The value lies in breaking down problems, organizing workflows, and detecting when the machine makes errors. A programmer with judgment uses AI as an autocomplete tool; one without judgment produces apps that work until they fail, with no idea why.

The VBA Case: Who Puts Out the Fire AI Starts

A concrete example involves Excel macros in VBA. One participant notes their company uses them daily, with many sheets containing hastily written, error-prone, undocumented macros. Behind these are profiles who relied on AI to generate them but cannot identify faults or effectively prompt the tool to fix them. This gap—understanding what the machine wrote—is where some are making themselves indispensable.

This isn't a romantic argument about code craftsmanship; it's a market reality. AI has multiplied mediocre software production, increasing demand for people who can fix it. Those who only know how to ask AI for things compete with anyone who asks better. Those who also understand the output compete in another league.

The Sector Paradox: Too Many Juniors, Not Enough Programmers

The diagnosis splits in two. One view holds the sector is saturated, junior roles are gone, and vocational training (FP) or bootcamps hold little value compared to self-taught skills. Remaining spots go to university graduates, while interviews become gauntlets of technical tests. Another view claims demand exceeds supply, forcing companies to hire non-technical staff.

Both can be true simultaneously, which is the trap. Generic candidates abound, while problem-solvers are scarce. One participant cites a job portal listing 28 ads from the same company seeking programmers for weeks; few apply, and almost none pass technical tests. The shortage isn't of degrees, but of people who finish what they start.

AI as a Fool Detector: The Ideological Noise

Not all content is analysis. Some opinions reduce AI to marketing, manipulated databases, and algorithms, comparing it to socialism for spotting gullible users. They cite a well-known software engineer claiming 90% of AI is pure marketing. The counter-argument references chess: it didn't die when machines beat grandmasters. Today, cheap engines beat elites, yet the game remains more popular than ever.

This parallel is the most useful—and debatable—argument here. In chess, machines didn't replace players; they provided infinitely better coaches. In programming, AI doesn't replace those deciding what to build, but it has commoditized the part of the craft that once took years to learn. The question is whether that part was what paid the bills.

What to Do If You Decide to Learn Anyway

Practical recommendations converge: learn fundamentals traditionally, work with modern tools. Python appears as the language with the most future potential, with free materials available on official websites, video platforms, and tech academy courses. Paying for training has a bad fruta among those burned by it.

Another repeated tip is using AI as a learning tool from day one, not a shortcut. A warning applies to any trade: if you enjoy coding, proceed; if not, hunting your first external bug will exhaust you. Memorizing data is useless now, but the ability to divide problems and organize remains the key asset.

---

With these insights, the reasonable prediction is that AI won't eliminate programming but will narrow the middle tier of the profession. Those who only request code will have low ceilings and high competition. Those who can read, correct, and decide what to build will have more work than ever, though less glamour than bootcamps promised. And if the bubble bursts, as some suggest, those relying solely on shortcuts will feel it first.