From Stack Overflow to AI Agents: Why I Stopped Fighting and Started Orchestrating in 2025
Source: Dev.to
In my journey as a web developer, I always felt comfortable studying, practicing, and coding. It was a familiar cycle: get frustrated with documentation, struggle with code that didn’t work, and try to find the solution in the official docs or through our best friend, Stack Overflow.
All of that changed for me in 2025. We no longer need to be stuck in that endless loop, and it’s all thanks to AI. When ChatGPT exploded onto the scene in 2022, we all started seeing this tool as something that would change the way we work—and it has.
Now many of us can’t imagine day‑to‑day tasks without AI assistance. AI models have evolved to the point where they can code several times better than most developers, know more languages and syntax, and have “seen” more code than we could in ten lifetimes. Rather than fearing job loss, I’ve started adopting AI into my development and creation process.
The Rise of Agents and MCP
Claude Code
Claude Code is a terminal tool that integrates with Anthropic’s AI models. It lets you develop, perform any type of task, and automate workflows. Agents take AI models to the next level, enabling orchestration and well‑planned work.
Model Context Protocol (MCP)
The Model Context Protocol (MCP) provides “plugins” for AI models, granting them access to functionalities they can’t reach on their own—such as databases or cloud‑hosted documents. The possibilities are virtually infinite.
Agents can also specialize: you can have dedicated agents for different task types that understand how to outline plans and implement them.
My Current Stack
- ✅ ChatGPT Pro
- ✅ Eleven Labs (for audio/content creation)
- ✅ GitHub Copilot Pro
I still feel I’m only leveraging 20‑40 % of what AI can do for me. My goal for 2026 is to unlock more of that potential.
The Hardest Question
What percentage of the development process should we delegate to AI?
This is the toughest question I’ve faced. Two years ago I imagined the AI usage process as “you writing and AI autocompleting.” That fantasy faded quickly as AI began planning for itself, code‑generation models improved dramatically, and “Vibe Coding” emerged.
When agents arrived with advanced capabilities, my ideal of AI usage collapsed. I realized how wrong I was about the limits of AI.
I don’t yet have a definitive answer for how a programmer’s day‑to‑day should look, but I believe AI will not replace us—at least not in the immediate future. The landscape can shift dramatically in a couple of years, so we’ll see how things evolve by 2028.