I've been using AI tools in my day-to-day work for months now and honestly, I can no longer imagine working without them. But not in the way you might think — it's not "AI does my job." It's more like having a pair programming partner who never gets tired of searching through documentation or explaining why that regex doesn't work.
The difference in workflow is real:
I'm going to tell you about the tools I actually use, not the ones I tried once and abandoned. This is my current stack.
For development, my main combination is Cursor and Claude Code. They're different tools for different moments. Cursor is my editor — basically VS Code with superpowers. I use it for everyday work: smart autocompletion, quick refactoring, and that "edit with AI" feature that lets you select a block of code and ask it to modify it. When I'm writing code and need something specific, Cursor is instant.
But Claude Code is something else entirely. I use it from the terminal for bigger tasks — when I need to understand a new codebase, implement a complete feature, or make changes that touch multiple files. The key difference is that Claude Code understands the full context of the project. I can tell it "add authentication with magic links" and it will explore how the project is structured, what dependencies I already have, and propose an implementation that fits with what already exists.
Not sure which one to use? Try this:
Which tool do I use for...?
Select a task to see my recommendation
What I like most about Claude Code is that it works iteratively. It doesn't spit out 500 lines of code at once hoping it works. It reads files, makes small changes, runs tests, and adjusts as it goes. It's like watching someone actually work, not like asking an oracle to guess what you need.
A concrete example: last week I needed to migrate some components from one UI library to another. With Cursor it would have been tedious — select each component, request the change, review, repeat. With Claude Code I explained what I wanted to do, showed it a couple of examples of the migration pattern, and it processed the entire directory while maintaining consistency. An hour of tedious work turned into ten minutes of supervision.
For images I use Freepik with its AI generator. I know there are many options out there (Midjourney, DALL-E, etc.), but Freepik has something that's key for me: clear licenses for commercial use and a style that works well for UI. When I need illustrations for a landing page, custom icons, or placeholder images that aren't the typical stock photos, Freepik gets the job done quickly.
The typical workflow is: I generate several variations with a simple prompt, pick the one that fits best, and if I need adjustments I either edit it directly in Figma or run it through another model for variations. It's not perfect — sometimes I need three or four attempts to get something usable — but it's infinitely faster than searching through image banks or commissioning custom illustrations.
I also use ChatGPT for things that aren't code: drafting tricky emails, reviewing text, brainstorming ideas, explaining concepts I don't understand. It's my instant "second opinion." When I'm designing and I'm not sure if a UX decision makes sense, I explain the context and it gives me perspective.
What I've learned after months of using these tools is that the key is in how you ask for things:
Same request, different result
"Improve this code"
Result: random changes, may break things, you don't know what to expect
"Refactor this function to make it more testable. Extract the validation logic into a separate helper"
Result: predictable changes, easy to review, aligned with your intent
The difference between a vague prompt and a specific one is the difference between frustration and productivity. You don't tell it "build me an app" and go grab a coffee. You tell it exactly what you need, review what it does, and constantly course-correct.
The mistake I see many people make is treating these tools like magic. They expect perfect results on the first try and get frustrated when they don't work. The reality is that it's an iterative process — just like working with a talented junior. You have to review their work, give feedback, and guide them. But that process is still much faster than doing everything yourself.
My productivity has gone up noticeably, but not because I work less. It's because the time I used to spend on mechanical tasks now goes toward design decisions, architecture, and the parts of the job that truly matter. AI takes on the boring work. I keep the interesting stuff.
If you still haven't integrated these tools into your workflow, start with just one. Cursor is probably the easiest entry point — it's VS Code, you already know how to use it, it just has extra features. Once you get used to having contextual help while writing code, you won't want to go back.
The future of development isn't "AI replaces programmers." It's programmers who know how to use AI doing the work of three. And that future is already here.