In Episode 6, the features Paul wants turn out to be “premium model only.” Model choice matters, but it isn’t always the bigger-is-better story the price list implies.
A simple rule of thumb
- Fast, smaller models: renames, formatting, boilerplate, simple tests, summaries.
- Larger, slower models: debugging across files, design decisions, tricky refactors, reviewing security-sensitive code.
- When results are poor, first improve the request (context, scope, examples), then try a stronger model.
Track which tasks actually needed the bigger model. After a few weeks you’ll know whether the upgrade pays for itself.