1. Actual coding skills decline When I started my studies in 2022 I had general knowledge of the web: HTML, CSS, a bit of JS and some Python because I'd built a web scraper once. During the first semester ChatGPT launched. I was skeptical and two feet deep in my weekly Java coding assignments, spending about 20 hours each week on easy, medium, and hard Java programming tasks alongside introductory CS lectures. I'm extremely happy that I started adopting AI a little later, probably around Q2 2023. By then I had already built my first solo side projects and thought: that's it, I now need to build a business using AI! I still watched a lot of YouTube videos about Spring Boot back then and really taught myself the basics, aided by ChatGPT here and there. Fast forward a few years: my university dropped the coding homework tasks because everyone did them with AI and introduced 3 in-person exams during the semester inside lecture halls where you had to solve coding exercises without AI. I personally noticed my coding skills declined significantly because of heavy AI-assisted coding adoption. I used to be very skeptical of AI directly editing my code and used the ChatGPT on one screen, code editor on the other approach for a long time which was the sweet spot between increased productivity and still understanding everything I wrote. Nowadays I just prompt and accept.
2. Long-term retention of knowledge drops Through tools like NotebookLM and Google Gemini (which was the go-to education AI at my university and from what I heard at others too) you don't even need to read actual sources, scripts, or papers anymore. You just upload everything and have it summarized. Upload your lecture scripts, homework assignments, old exams, and ask the AI to explain everything and create exercises for you. This works. But you only optimize for the exam. Like fine-tuning a model for a specific task, you fine-tune your knowledge input to pass the exam rather than building a holistic, deeply rooted understanding. When I used to write my own summaries, read through lecture scripts, and actually attend tutorials, the knowledge I gained stayed with me much longer.
3. You solve tasks instead of learning Convenience is the enemy of every student. AI has become so convenient that you really have to kick yourself to do anything without it. Reading something, writing something... at every task you tend to think: "I know I could do it, I'm capable enough, but it's just not worth my time when I have other assignments." It has genuinely become too convenient. It's like your brain builds up friction against using itself and would rather outsource to AI because it's less demanding and exhausting. I recently heard that Google researched something similar and is now trying to build friction into their products that sometimes guides users toward an answer instead of just giving it to them.
What I observed during my studies mirrors what I now see in my professional life: using AI for coding has become so convenient that you just prompt nearly everything and gradually lose the ability to code (and maybe even to think independently).
That convenience problem is actually what I'm trying to solve in a different context with my current project.