Anthropic and the Impact of AI on Engineers' Intuition

AI: A Double-Edged Tool for Engineers

In today's software development landscape, engineers are under increasing pressure to deliver results more quickly through the use of artificial intelligence tools. These tools, capable of generating code, analyzing data, and designing solutions in record time, raise concerns about their impact on the fundamental skills of engineers.

According to a study conducted by Anthropic, the company behind the AI tool Claude, excessive use of AI could hinder learning, particularly among novice software engineers. Indeed, when AI fills gaps too quickly, it allows for the production of a functional result without the user developing a deep understanding of the underlying process. More experienced engineers, on the other hand, fear gradually losing the valuable intuition that has made them effective.

The Illusion of Quick Competence

Historically, engineers have always borrowed solutions from various sources, whether from manuals, forums, or online code snippets. However, the speed and scale at which AI operates today are unprecedented. It can provide plausible solutions in seconds that, while functional, may be poorly understood by those who use them.

One engineer shared his experience after joining a new team and using AI to quickly adapt to an unfamiliar codebase and language. While AI allowed him to become productive almost immediately, he recognized the risk of settling for a simple "it works" without seeking to understand the details.

From Generation to Inquiry

To counter this trend, he chose to use AI not only to generate solutions but also to deepen his understanding. After achieving a functional change, he would ask AI to help him explore the code step by step, posing questions about model choices and the implications of certain abstractions.

A particularly effective technique was to rephrase in his own words how a part of the system worked, then ask AI to correct his mistakes. This process helped him develop solid mental models, essential for debugging and designing effectively, and not just for copying and pasting.

Towards Informed Use of AI

The real challenge for engineers in the age of AI is to use it to accelerate learning rather than replace thinking. The temptation is strong to take the easiest path: ask, accept, ship, repeat. However, this approach leads to superficial knowledge and increased dependency.

The most sustainable approach is to use AI to progress quickly while ensuring a full understanding of what has been built. Engineers who succeed in maintaining their sharpness will not be those who avoid AI, but those who integrate it as a partner in their ongoing learning. AI can not only make us faster, but when used well, it can also make us better at our jobs.