From the Zed Blog: Alberto Fortin shares his honest reflection on the reality of using LLMs in production code and why he's taking a more measured approach.

Zed's platform  offers insights and resources on technology, software development, and entrepreneurship. Through articles, tutorials, and personal stories, Zed shares  knowledge and practical advice for individuals looking to succeed in the tech industry. Readers can learn about startup culture, product development strategies, and career growth opportunities to navigate the complexities of the tech landscape effectively.

A seasoned software engineer with 15 years of experience shares his honest reflection on using LLMs in production code. After initially embracing AI tools with enthusiasm, he encountered significant challenges including poor code quality, cascading bugs, and productivity illusions. He advocates for a more measured approach where developers maintain architectural control and use LLMs only for small, well-scoped tasks like refactoring rather than large features. The key insight is treating LLMs as assistants rather than replacements, emphasizing that senior developers should trust their skills and use AI to leverage existing knowledge rather than offload critical decision-making.

Why I'm Dialing Back My LLM Usage — Zed's Blog

The Reality Check: When AI Hype Meets Production Code

Point about productivity illusion hit me.
I am working with a timer (not exactly pomodoro but it kinda helps maintaining a sense of time for me). It is amazing how often I cought myself forcing an LLM to do something for like 45 minutes whereas I would be able to do it myself within similar timeframe or less.
Plus, being in constant review-mode is tiring if you want to make sure you have everything under control.
I still think LLMs are a powerful tool but you gotta recognize when and why to use it.

Always felt this hype was blown out of proportion.

I’ve pretty much always shared this approach, constantly even just reviewing and breaking into the tiniest of tasks, avoiding agents doing stuff cause they go off the rails and I have to refactor it anyway. Focus more on ask. I’ve been learning new languages and staring at it and CONSTANTLY asking idiomatic approaches etc. I will say that has definitely speed up learning new language and concepts behind them etc. Constantly have to ask questions and break things into tiny pieces to make sure it’s accomplishing what I intend though. So amazing tool but over reliance was clearly something people were falling into a trap of.