I still remember the bookstore. I was holding a 600-page brick of a book on how to build Windows applications, trying to convince my mother that I really&nbs...

Ayende is a software engineer, blogger, and open-source contributor known for his expertise in software architecture and database development. Readers can learn about database design, NoSQL databases, and domain-driven design (DDD). With blog posts, tutorials, and open-source projects, Ayende shares  insights and best practices for building robust and scalable software systems.

Ayende @ Rahien

A veteran developer reflects on how learning to code has changed from the 1990s to 2026, tracing the evolution from physical books through Stack Overflow to AI coding agents. The core concern is that each abstraction layer has reduced the conceptual depth developers carry — but AI agents are uniquely different because they abstract away the reasoning itself, not just mechanical work. The author wrestles with practical consequences: how to evaluate junior developers when AI can ace any take-home task, and what foundational knowledge still matters. The conclusion is that developers should use AI tools but deliberately go deeper on occasion — understanding DNS, servers, and what frameworks actually generate — because that mental model is what separates engineers who grow over a career from those who get stuck when abstractions leak.

Learning to code, 1990s vs 2026

Hidden costs for reduced conceptual depth

What is the role of a junior developer now?

<blockquote>
“It changes the depth of knowledge a developer carries, and that has consequences.”
</blockquote>
I’m a forever advocate for progress, but overly relying on AI is a malpractice in itself. It’s the same as too much is always wrong. For example, social media. It lowers the IQ of its users and makes them a slave to consume endless content (which unfortunately, most of them are bad even before AI).
Look no further with AI in development. If we evolve to just pressing tab and accepting every AI suggestion into our codebase, then our brain will not evolve progressively. Our brain cells needs its daily exercise like any other muscle in our body. I’m all good with AI, but AI malpractice is one road I fear a lot of developers go down to in this modern era.

In the 90s, everything was written from scratch. Only a programming language was used.
In the 2000s, due to the growth of the codebase and the complexity of logic, libraries and frameworks began to appear.
In the 2010s, autocomplete and other automation tools appeared.
In 2022+, code began to be generated completely automatically.
But what remained constant at all stages is fundamental knowledge, which distinguishes a coder who knows the syntax from an engineer who understands all the stack limitations and understands how everything should function.

well Billy, its supposed to be faster and better, having so much access and knowledge, streamings and AI, internet… to learn, people are not more absorbing just learning, we living in the COW model behavior. sad.