I’m sure everyone has been hearing the new buzzword lately — AI. From productivity apps to development tools, most online applications we use now include some form of AI assistance. In the world of software development, two tools garnering significant attention are Cursor and Claude Code. Along with the excitement, there’s also growing anxiety that AI will replace jobs. Whether recent layoffs are truly driven by AI or simply used as a convenient explanation is debatable.
Whatever the case, AI today feels a lot like the early internet era. And the internet wasn’t a bad thing, right?
The internet made our daily lives easier — online shopping, instant access to knowledge, digital banking, communication — the list goes on. AI feels similar in that it’s not just a flashy feature or a chatbot; it’s fundamentally changing how we work.
In many ways, AI feels like a co-worker — one you’re comfortable asking for guidance when you’re stuck on a problem. We’ve all been in situations where we hesitated to ask for help, worried about interrupting someone or being judged for not knowing the answer. AI removes that friction.
AI development tools, in particular, feel like an advanced version of Google search — or even another teammate — helping you explore solutions, validate ideas, and move forward faster.
What are AI Development tools?
AI development tools are software tools that leverage artificial intelligence to help developers build, debug, test, and understand code more efficiently.
AI Development Tools in Action: Cursor & Claude Code
Two AI tools gaining significant attention among developers are Cursor and Claude Code. Both are designed to make development workflows faster and less frustrating. I used them regularly while working as a Full-Stack Engineer at my last company, and after nearly 6 months of real-world use, I feel comfortable sharing my experience and reviewing how each tool performs.
Cursor
According to the Cursor’s website, built to make you extraordinarily productive, Cursor is the best way to code with AI. Here is the pricing. I initially tested the free version and later upgraded to the Pro plan, mainly because it allowed me to upload a Figma feature design and have Cursor generate a UI that looked almost identical to the original design. I was very excited about the results and had a very good experience.
However, when it came to making iterative changes, the experience became frustrating. Asking the Cursor to modify existing implementations often resulted in unintended side effects, impacting previously working code. I found myself spending more time fixing issues and undoing changes than making progress, creating a loop in which each attempt to move forward made the codebase messier.
Cursor works well as a blueprint generator — it helps you define structure and get momentum early on. Beyond that stage, however, it requires a high level of caution. In one instance, while attempting to resolve a database issue in local development, the Cursor deleted the entire database. After that, I learned to be very explicit: I would ask for suggestions only, without allowing it to make direct code changes. Whenever I allowed automatic changes, things quickly became chaotic.
Cursor also struggled with more complex feature implementations. For example, while building an inventory system, it had difficulty handling a workflow where an order required switching between multiple forms — one to define the base order and another to specify individual items. Managing this kind of state and interaction logic proved beyond its current capabilities.
When to use Cursor?
Use Cursor to:
- Quickly scaffold new features
- Generate UI directly from designs
- Establish an initial code structure
- Create documentation for your codebase
- Add new features safely by specifying in your prompt to check existing components, reuse similar functionality, avoid duplication, and follow the project’s coding practices
Avoid Cursor
Avoid relying on it for complex logic, iterative changes, or automatic code modifications, where it can introduce unintended side effects.
Claude Code
According to Claude’s website, it is a next-generation AI assistant built by Anthropic and trained to be safe, accurate, and secure, helping you do your best work. Claude Code is a bit more expensive than Cursor, but it’s worth it 🙂 I am paying $25 Cad plus tax for the Pro Plan, though I do wish the usage limits didn’t run out so quickly.
I like that Claude Code doesn’t require switching editors. It runs as a VS Code extension, while Cursor is a separate, AI-first editor built on top of VS Code.
Claude Code really shines at solving complex problems. It doesn’t just provide a solution—it explains the issue and confirms whether to proceed with changes, making it feel like we’re solving problems together. The interface is clean and intuitive, which makes the experience even better.
I find it particularly useful for creating queries. For example, when we were using Prisma, Claude could scan the prisma.schema file and generate SQL queries I could use directly. I also use it for code refactoring — I ask it to check that I’m handling undefined properly, using the latest JavaScript syntax, and following best practices. It then provides actionable suggestions to improve and clean up my code.
I was able to quickly build features that would have taken me many hours if I hadn’t been using Claude Code. Thanks to Claude Code, I was able to quickly develop a proof of concept and demo it to the team.
The only downside is that Claude isn’t great at making style changes. It can fix issues, but it doesn’t generate a blueprint or scaffold a UI the way Cursor does.
Because of the great experience I’ve had with Claude Code, I even encouraged my team members to start using it — and they’ve confirmed just how helpful it is. Just like with Cursor, it’s important to be explicit in your prompts with Claude as well: make sure it checks the existing codebase standards and ensures any new changes follow established practices.
Thank you for reading 🙂