Use AI to Get Smarter, Not Lazier

The article advocates for using AI not just to complete tasks faster, but also to help understand what those tasks entail. The author explains that while AI can indeed speed up work, using it improperly leads to a lack of comprehension, which they term "cognitive debt". This occurs when tasks are completed quickly without a proper understanding of how they work or why they were done in a particular way.

The author illustrates this concept by using the example of triaging static analysis findings, specifically using the Semgrep tool. The task involves identifying whether certain code snippets contain security vulnerabilities such as SQL injection, path traversal, XSS, and CSRF. The author initially uses a simple AI prompt to quickly triage these findings, but this approach lacks understanding and context.

The author then demonstrates a more effective approach, which involves using AI to help understand the code behind each finding. By employing a skill in the Claude Code tool called "code-walkthrough", the author is able to gain a deeper understanding of the code's context and flow. This method involves creating visual diagrams, like sequence diagrams, that illustrate the code's structure and how different parts interact.

This approach not only helps the author to quickly triage the findings, but also allows them to validate the AI's suggestions, understand any false positives, and even provide explanations and fixes for the issues. The author emphasizes that this "balanced shortcut" still allows them to complete their tasks quickly, but also results in a deeper understanding, which helps them avoid cognitive debt.

This way of working with AI is therefore recommended over simply using it to complete tasks as quickly as possible, without understanding the underlying processes.