I used to treat “higher” models like Opus as a niche tool.

Its great for analyzing long texts or spreadsheets … clearly stronger than Sonnet for depth and accuracy, and that was about it.

That’s changing.

Opus is still my go-to for complex, layered problems with lots of moving parts.

But when I’m drafting, it often thinks too much, burning tokens, over-explaining, and hijacking my voice. The content is impressive, but it’s not mine.

Sonnet, with a well-engineered context, usually works better for writing. It stays closer to my intent, my style, my constraints.

Now Co-Work adds another layer with its dedicated space for Opus to grind through documentation and first-pass decks while I stay in charge of the ideas.

If you have access to Co-Work, there’s almost no excuse not to have a solid first draft for most of your documentation needs.

How are you starting to distinguish when and how you use different AI models in your own work?