The Dev World – Sergio Lema

Category: AI

  • Why AI Won’t Replace Developers Who Lead the Tool

    Why AI Won’t Replace Developers Who Lead the Tool

    —

    by

    Sergio Lema
    in

    I use AI code assistants every single day. My net output hasn’t changed. Not because the tools are overhyped — but because most developers are using them wrong. If you can’t explain the code an AI gave you, you didn’t write it. You just hosted it. Here’s how to stay the architect, not become the…

  • Why AI is Replacing Junior Developers and How Mid-Level Engineers Can Survive

    Why AI is Replacing Junior Developers and How Mid-Level Engineers Can Survive

    —

    by

    Sergio Lema
    in

    Many junior developers struggle to find jobs as AI handles basic tasks efficiently. However, companies now seek senior engineers to rectify poor AI-generated code. Developers should use AI for problem-solving rather than for coding directly, ensuring they understand and own their work to effectively address future issues and enhance system design.

  • Choosing the Right AI Coding Agent: Claude Code vs. GitHub Copilot vs. Gemini Code Assist

    Choosing the Right AI Coding Agent: Claude Code vs. GitHub Copilot vs. Gemini Code Assist

    —

    by

    Sergio Lema
    in

    The market for AI coding tools has shifted from simple chat interfaces to agentic workflows. After extensive testing across large-scale repositories, Claude Code is currently the superior tool for complex engineering tasks — though it comes with specific trade-offs in resource consumption. A modern development workflow shouldn’t rely on a single tool: it requires a…

  • The Art of Prompting for a Developer

    The Art of Prompting for a Developer

    —

    by

    Sergio Lema
    in

    Effective communication with LLMs is crucial for maximizing their utility in development. Clear, specific prompts guide LLMs, helping them produce relevant, quality outputs. By assuming the role of a senior developer, defining best practices, and critically reviewing outputs, users can improve interactions with these AI models and streamline their coding processes.

  • Architecture for my Generative AI Application

    Architecture for my Generative AI Application

    —

    by

    Sergio Lema
    in ,

    In this article, I explore using AI image generation models accessed via platforms like Hugging Face but faces challenges with high GPU costs and inefficient resource management. The solution involves a web application and a backend service that auto-scales GPU usage via asynchronous messaging, ensuring resources are only used when necessary.

  • I’ve Created a RAG in Two Days

    I’ve Created a RAG in Two Days

    —

    by

    Sergio Lema
    in

    In the AI era, web developers must integrate AI into projects to stay relevant. I’ve built a Retrieval-Augmented Generation (RAG) system in two days, combining a knowledge database with an LLM. Key components included data chunking and retrieval. The project showcased AI’s ability to provide specific responses, using creative datasets for testing.