I quit using ChatGPT for coding
I quit using ChatGPT for coding.
Not because it's bad. Not because I'm an AI snob.
Because I kept switching tabs.
Here's what happened:
I was building a trading CLI. Python, API integrations, 13 files, timezone logic that kept breaking.
Every time I needed help, I'd copy code into ChatGPT in the browser. Paste the response back. Lose context. Repeat.
Then I switched to Claude Code in the terminal.
Same place I write code. Same place I run tests. No tab switching.
In one session I:
- Refactored 13 files for timezone-aware functions
- Wrote 49 unit tests across 3 test files
- Fixed a bug where the AI was leaking its own thinking into responses
That last one is worth explaining. The AI was outputting its internal reasoning ("Let me think about this...") directly into the code responses. One line in the system prompt fixed it. But I only caught it because the AI was right there in the terminal, showing me everything in real time.
The model matters less than the workflow.
ChatGPT in a browser tab is a conversation.
Claude Code in a terminal is a pair programmer.
The difference isn't intelligence. It's proximity to the work.
I've since used the same setup to build newsletter automation scripts, content analyzers that read my notes and generate ideas, and a Telegram bot migration. All from the same terminal window where the code lives.
If you're still copying and pasting code into a chat window, try moving the AI into your editor. The speed difference will surprise you.
P.S. I teach my exact AI prompting and workflow frameworks inside Prompt Writing Studio. It's how I build tools like this in 30 minutes instead of 3 hours.