Claude Code Is Not the Same as ChatGPT: One Month of Real Use
My friend is an independent developer. He takes freelance work and runs a few small projects of his own.
Last month he switched his daily workflow to Claude Code. I sat next to him a few times and watched. Today, the real story.
First, Let Us Be Clear: It Is Not ChatGPT
Most people think Claude Code is "ChatGPT but in the terminal." It is not.
With ChatGPT, you copy code in, get suggestions, copy them out, and edit the file yourself. That gets old fast. After three or four files you want to throw your keyboard.
Claude Code runs right in your terminal. It reads your files, edits them, runs tests, checks git diff, and even helps you commit.
My friend says the first time he saw it run npm test on its own, then fix the code based on the errors, he paused. This is not chat. This is real work.
Pricing
Claude Code Pro starts at $20/month with a subscription. If you use the API, pricing varies by usage. He has not tracked his exact spend, but says it is "way cheaper than hiring a junior engineer."
What My Friend Built With It
He took on a small project: a backend booking system for a local restaurant brand. Tech stack was Node.js plus PostgreSQL, medium complexity.
His approach: let Claude Code scaffold the project first. Database schema, base API routes. Then have it write test cases, run them, and iterate based on failures.
The whole backend, from zero to a working baseline, took him about two days. By hand, he estimates four or five days.
One detail stuck with me. He asked Claude Code to refactor a messy middleware file. It did not just rewrite the code. It also pointed out a potential SQL injection risk he had missed.
Plenty of Downsides
First problem: it only works inside one repo. Cross-project work is out.
Sometimes he needs to reference utility functions from another project. He has to copy them over manually. Claude Code cannot reach them on its own.
Second problem: limited context.
As the project files pile up, it sometimes "forgets" earlier changes, leading to duplicated effort or conflicts. He got into the habit of asking it to refresh its understanding of the current state every so often.
Third problem, and the biggest one: bad fit for complex monorepos.
Another project in his studio is a large monorepo with a dozen interlinked packages. He tried Claude Code on it. Results were poor. It kept editing the wrong files or missing cross-package dependencies. For that kind of work he went back to doing it by hand.
Remote server work is also weak. When he needs to SSH into a server to debug a production issue, Claude Code cannot help much. It is built for local repos.
Compared to Cursor
He has used both. The differences are clear.
Cursor is an IDE. You see the cursor move, watch code change line by line, good for fine-grained edits where you want to stay close to the work.
Claude Code is a CLI. You throw a task at it, it works in the background, and gives you a result. Good for "let me grab a coffee, see what you changed when I am back" tasks.
His current habit: small edits go to Cursor because the feedback is direct. Big feature builds or repetitive refactors get handed to Claude Code.
I have not used Cursor's team collaboration features. I have heard the team version supports shared config, but neither of us has tested it deeply.
Who It Is For
If you spend a lot of time writing backend logic, running tests, or doing refactors, Claude Code will save you real time.
If you mainly work in large monorepos or debug production servers remotely, it will not help much.
My friend's verdict: it changed the rhythm of how he writes code, but it has not replaced manual work entirely.
I maintain an AI tool directory at saas.pet, updated automatically every day.
Next up, I will write a deeper look at Cursor and whether it actually holds up on large projects.