Sebastian Gomez

Sep 01, 2026

Updated: Sep 11, 2026

The AI Learning Paradox: Why Current Convenience Could Cost Your Engineering Future

A few days ago, I was listening to a reflection by Andrew Ng that resonated deeply with me, both as an Engineering Manager and as a university professor: the way the vast majority of people use artificial intelligence today is fundamentally detrimental to learning.

In an era where any tool can generate a React component, a backend script, or a complete algorithm solution in three seconds, a widespread illusion of competence has taken hold. It feels as though everyone has become more productive, but beneath that superficial speed, a quiet problem is emerging: the atrophy of a software engineer's most critical muscle, the ability to solve complex problems independently.

Whether you are starting out in programming or leading engineering teams, this is an urgent conversation we need to have openly.

The Classroom Reality: From Critical Thinking to Blind Copy-Pasting

Teaching at the university, I have had a front-row seat to how student dynamics have shifted over the last few semesters.

The pattern repeats itself constantly: when faced with a problem requiring abstraction or algorithmic design, the first impulse is no longer to sit down with pencil and paper to map out data flow. The first impulse is to throw the prompt at a language model. The most concerning part is not that they use the tool, but that they frequently do not even read the response the AI provides. They copy, paste into the editor, and if it compiles or passes basic tests, they consider the work done without having built a mental model of what just happened.

University programs and the early years of a developer's career are not designed solely to deliver code that runs; they exist so your brain experiences the friction of facing uncertainty, making mistakes, debugging, and structuring a solution.

When you outsource that initial cognitive struggle to a machine, you are outsourcing the exact biological process that builds engineering judgment.

The Great Asymmetry: Why AI Empowers Seniors but Makes Juniors Fragile

There is a fundamental difference in how an experienced engineer interacts with AI compared to someone still in training:

• The senior developer uses AI as a velocity multiplier on top of an already consolidated foundation. The senior knows exactly what architecture they want, recognizes design patterns, spots vulnerabilities instantly, and possesses the judgment to discard code when a model hallucinates or suggests inefficient solutions.
• The junior developer, having not yet built those fundamentals by hand, lacks the reference framework to evaluate what the AI produces. They end up operating as a prompt-suggestion consumer rather than a software architect.

The paradox is that AI is incredible for those who already know how to program, but can become a dangerous trap for those who are just learning.

Even senior engineers must stay mindful of this risk: if we stop grappling with raw code and manual problem solving, our own judgment will gradually rust over time.

Who Will AI Actually Replace?

There is endless discussion about AI eliminating junior developer roles across the industry. My stance is straightforward: AI will not replace junior developers who are eager to learn; it will replace developers (junior or senior) who limit themselves to being passive intermediaries between a prompt and a repository.

If your only value proposition is asking a model for code and committing it without understanding its memory footprint, concurrency behavior, performance implications, or security posture, your role is easily automated.

By contrast, the engineer who masters the fundamentals, understands underlying architecture, and uses AI as a Socratic tutor (interrogating its answers and demanding alternatives) becomes exponentially more valuable and difficult to replace. The real technical revolution is not about coding less, but about thinking deeper.

Whiteboard Discipline: The Irreplaceable Value of Deliberate Effort

To counteract this cognitive trap, there is one habit I consistently recommend to my students and the engineers I mentor: whiteboard (or blank notebook) discipline.

When approaching a new feature or a complex bug:

• Step away from the prompt during the first few minutes: Before writing a single line of code or opening an AI chat, sketch the flow on paper or a whiteboard. What are the inputs? What are the outputs? What intermediate states exist?
• Struggle with the problem first: Try to work out the core logic yourself. Experience the roadblock, test hypotheses, and consult official documentation. That brief moment of mental friction is what creates lasting neural connections.
• Use AI as a Socratic tutor, not a shortcut generator: When you turn to AI, do not ask "give me the code to solve this." Ask "review my reasoning, what edge cases am I overlooking?" or "explain why this algorithmic approach carries higher time complexity."

We do not build things by hand because AI is incapable of doing so; we do it because it is the only proven way to cultivate rigorous analytical thinking and sound engineering judgment.

How are you or your team managing the use of AI tools day to day? Do you feel it accelerates your learning or encourages convenient shortcuts?

I would love to read your thoughts and experiences in the comments. Let us keep learning and building with purpose!