# What if I told you that I stopped "coding" and I'm still building apps? 🤯💻

Machine translation from the original post. Not reviewed by a human translator. Historical claims may no longer be current.

What if I told you that I stopped "coding" and I'm still building apps? 🤯💻
I'm not talking about No-Code or Low-Code, but about a complete change of approach.

With the Cursor development tool, which contains a revolutionary feature called Composer Agent, I started writing development tasks instead of writing code line by line.

What does it look like in practice?

🔹 Step one - planning: I ask Composer to write a README.md for the project, including a description of the features, file structure, dependencies, and the right way to develop it.
 🔹 Step two - execution: after it plans everything in an orderly way, I ask it to start developing the project step by step.
 🔹 Step three - improvements and iterations: if something is missing or doesn't work, I ask for changes - and it fixes them.

What is Composer Agent, actually?

When you run Cursor in Agent mode, it doesn't just make small changes, but acts like a real programmer:
 ✅ Plans the code structure and architecture
 ✅ Suggests fixes for errors and works in iterations
 ✅ Writes complete code, including configuration files, documentation and tests
 ✅ Understands which parts of the project need to change so everything works

🚀 The result?
Instead of sitting for hours thinking about how everything is supposed to connect, I just give guidance and approve.

Within minutes I have a working project, which needs a bit more polish, but is already at a very advanced stage.

We are moving from programmers to code reviewers -
In an era where AI can write the code for us, we no longer function only as programmers, but more like Code Review developers.

What does that mean?
 🔹 You need to understand the big picture and manage the product smartly
 🔹 Make sure the code is written correctly, without edge cases, bugs and performance problems
 🔹 Know when to approve code and when to send it back for fixes

But it's not all rosy.

AI isn't perfect - sometimes there are bugs and faults that require several iterations and sometimes even manual intervention.

Here is where real programmers have the advantage: the ability to understand, diagnose and solve complex problems.

And you can see it with other developers too - I'm not the only one who thinks so - Andrej Karpathy (a legendary programmer, worked at OpenAI and Tesla) tweeted about it:
https://lnkd.in/dJym_Jzw

We're approaching an era where programmers don't code, but guide and manage processes.

So yes, it won't suit critical projects, but for weekend products and MVPs? It's crazy.

Original: https://ofershap.github.io/posts/building-apps-by-assigning-tasks-to-cursor/
Source: https://www.linkedin.com/posts/ofershap_%D7%9E%D7%94-%D7%90%D7%9D-%D7%94%D7%99%D7%99%D7%AA%D7%99-%D7%90%D7%95%D7%9E%D7%A8-%D7%9C%D7%9B%D7%9D-%D7%A9%D7%94%D7%A4%D7%A1%D7%A7%D7%AA%D7%99-%D7%9C%D7%A7%D7%95%D7%93%D7%93-%D7%95%D7%90%D7%A0%D7%99-activity-7295819055888789504-yCJR
Published: 2025-02-13T17:00:12.687000+02:00