What's this about?
A developer's writeup of deliberately switching their primary coding agent for a week, with concrete comparisons. It is the hands-on counterpart to the FT story, and one of several defection narratives this week.
What exactly is a "coding agent"?
Not autocomplete. A coding agent reads your repository, plans a change, edits multiple files, runs the tests and iterates on failures — you review the result rather than the keystrokes. Because it operates on whole codebases, switching between them is a real workflow change rather than swapping an autocomplete engine.
What's the community saying?
HN 243 points, 280 comments
- Many similar reports of switching, usually citing price or usage limits rather than raw capability.
- A persistent counter-current insisting the capability gap is real on genuinely hard tasks and worth the money.
- Frustration with usage limits as the actual deciding factor for both tools — which is why the restored Codex limits item below mattered.