2026-02-03
What's the difference between coding with an AI and having an AI code for you?
In this walkthrough, I build the same small Python project two ways. In Part 1, I write starter code myself and iteratively refine it by copying prompts and responses back and forth with Claude and ChatGPT — classic AI-assisted coding. In Part 2, I hand requirements to Claude Code and watch it scaffold the project, install dependencies, write the code, and run it — agentic coding.
The example is deliberately simple (an LLM that answers questions in different "personalities"), so the focus stays on the workflow, not the problem.
Both approaches produce working code, but the tradeoffs are real: more control and understanding with AI-assisted coding, more speed and less friction with agentic coding. That speed is seductive — agentic coding is accessible to people with little developer experience, which is great, but also fuels the rise of "vibe coders" running code they don't fully understand.
Knowing when to reach for each approach, and when to slow down and actually read the code, is the skill that matters most.