❓ Composer 2.5 vs Opus for complex planning, how are people finding it?

❓ Composer 2.5 vs Opus for complex planning, how are people finding it?

I wasn't impressed at all with Composer 2 to be honest. It didn't really change my working pattern. I love cursor and especially 2 but was still using Claude/Opus for more complex planning and thinking work, then moving into Cursor auto for plan execution. Composer was there under the hood, but Opus was doing a lot of the "brain" work.

Composer 2.5 though, I've been giving a bit more of a chance.

I've done a few small tests on dece t size work over the last week, definitely nothing scientific. My normal pattern is that when things are getting harder, more complex, or more architectural, I'll use Opus to build plans, the serious hard big chnages it fairs the best.

So I started swapping Composer 2.5 into some of those tasks and then got both models reviewing each other:

• Opus plan → reviewed by Composer 2.5 • Composer 2.5 plan → reviewed by Opus

I found it surprisingly difficult to separate them on review and end result plan.

Both found issues. Both got pretty nit-picky. But nothing felt fundamentally wrong. It was more along the lines of:

"This could go deeper." "This assumption could be stronger." "You could add more detail here."

Honestly, not much different from asking the same model to take another pass and think harder.

One takeaway for me so far is that my AI usage and subscriptions are stretching far further because of this. If I'm not seeing an obvious degradation in quality, then getting more value out of the same spend seems like a pretty sensible trade-off.

So opening it up:

How are people finding Cursor with Composer 2.5 compared to Opus when it comes to planning more complex work?

Not smaller tasks, more architecture, larger implementations, and work where you want the model thinking ahead rather than just generating code.

#AI #Claude #Cursor #SoftwareArchitecture #GenAI #Engineering