In my daily workflow as a developer, I rely on several AI tools to boost productivity, solve coding problems, and explore new ideas. Over time, I’ve developed a kind of personal “AI stack” — each tool has its own place depending on what I’m working on and how complex the task is.


My Approach to Problem-Solving

For small, quick questions or one-line fixes, I still use ChatGPT occasionally — though less and less nowadays. It’s great when I just need to double-check something simple or clarify a concept.

When I want to work within a web environment and use Claude with OPUS mode, I switch to Claude in the browser. I usually do that when I’m not dealing with multiple files and when the project context is clearly defined. It’s lightweight and efficient for focused coding sessions.

But most of the time, I use GitHub Copilot Agent, combined with Claude Code, which I’ve been studying in more depth lately. That combination feels like the best of both worlds — Copilot gives me excellent in-editor support, while Claude helps me reason about structure, patterns, and larger architectural decisions.


Copilot Agent: My Main Companion

One of my favorite things about the Copilot Agent is how well it’s integrated into my workflow. It’s not just about generating code — it’s about reviewing, applying, and managing changes directly in my editor. The agent offers a clear interface to choose models, including Claude 4.5, which is available via paid tokens.

I really appreciate how natural the process feels. The accepting and reviewing of changes is smooth — no copy-pasting or switching between tools. You can even set restoring points, which act like local checkpoints. This means I can experiment freely, knowing I can easily roll back if something goes wrong.

Here’s a quick look at what that restoring point feature looks like with Copilot Agent — it feels like having version control for every small AI-assisted session.


Exploring Claude Code 2.0

Recently, I started diving into Claude Code 2.0. The team behind it has released a VS Code extension, so now you don’t have to use it from the command line anymore — which is a big improvement in terms of usability.

The extension allows you to reference files directly, navigate project structures, and use commands like you would with any other assistant. You can ask it to explain, refactor, or even write custom commands.

That said, I’m still somewhat limited by experience here — I haven’t had enough large or complex tasks to truly test all the advanced options that Claude Code offers. But I’m exploring it step by step.

If you’re curious about how to create custom commands, I recommend checking out this video — at around the 13:00 minute mark, there’s a detailed walkthrough of how it’s done:
👉 How to Create Custom Commands in Claude Code


Final Thoughts

AI tools are evolving incredibly fast, and the key isn’t just knowing how to prompt them — it’s knowing when to use which tool. For me, Copilot Agent has become the backbone of my workflow, while Claude Code 2.0 is something I’m still learning and integrating gradually.

The balance between automation and understanding is what matters most. These tools don’t replace the developer — they extend what we can do.