My Thoughts on AI

AI is a tool built by humans for humans. It can make me faster, but it does not replace my responsibility as a developer.

I mainly use AI to challenge my ideas, write repetitive code, and handle administrative work such as documentation and email drafts. This can make me up to 40% faster, but speed is useful only when I still understand and control the result.

I use AI to deepen my thinking

Before implementing a feature, I decide what needs to be built. I then ask AI to grill me: question the idea, identify trade-offs and edge cases, challenge my assumptions, and continue one question at a time until the solution is clear.

I want AI to understand my coding style, but I do not want it to agree with everything I say. It should follow my conventions while remaining critical of my architecture and decisions. I also challenge its suggestions in return.

The final judgment remains mine because I understand the actual problem, the users, the client, and the context around the work. AI does not participate in those human relationships for me.

I generate only what I can review

I use AI during implementation, especially for repetitive code, but I generate one feature at a time—not an entire application. I never want to produce more code than I can honestly review.

This limit helps me avoid technical debt and preserve my identity as a developer: the way I think, name functions, organize code, and choose the right level of abstraction. I want AI to increase my output without taking those decisions away from me.

If I cannot understand, question, and maintain generated code, then the productivity gain is not real. I am the person on the job, and I remain responsible for everything I deliver.

Fundamentals matter more than productivity

My biggest concern is becoming unable to work without AI—sacrificing skills on the altar of productivity until I can no longer think critically or judge its output.

That is why I believe developers should first learn how to do the work themselves and use AI for tasks whose results they can evaluate honestly. I still learn through documentation, videos, and courses. I may ask AI specific questions about a bug or error, but I try the solution, test it, research further, and continue until I understand it fully.

AI companies are selling a product, so it is not surprising that they present it as all-powerful. But proper understanding and technical expertise matter more than ever. A developer is not simply someone who produces code. A developer understands the job and can deliver a solution.

The human remains responsible

I also use AI to draft documentation and emails after it asks me enough questions to understand what I want to communicate. Even then, I review everything manually. AI can prepare a draft; it cannot take responsibility for it.

The same boundary applies to privacy. I do not share sensitive information. If AI is treated like a colleague, it should receive only the resources needed to perform the task, with the same attention to privacy and security.

Without AI, I would write code more slowly. But the essential parts of my work would remain: understanding clients, interacting with users, deciding what should be built, and taking responsibility for what is delivered.

AI remains AI. The human remains human. The value is not in confusing the two, but in using the tool without surrendering the craft.