After 1.5 years with Cursor, I still don’t think you understand your own differentiation.
I’ve been using Cursor for a year and a half now, and I feel like the team still hasn’t figured out what truly differentiates Cursor from competitors.
As an aggregator with access to so many models, it’s baffling that subagents can only be swapped in Max mode. When new models drop — like the latest ones — we literally can’t try them unless we’re on Max. If a model requires a 1M-token context window, then don’t show us the 300K and no-thinking options in the UI. It’s misleading and frustrating.
On the workflow side: I’ve spent significant time building my own pipeline-based AI workflow within Cursor. Meanwhile, subagent model selection remains locked down — and then Claude Code ships a similar “workflows” feature. What exactly is Cursor’s moat going forward?
The Agent Windows experience in 3.x has also been rough. It wasn’t until 3.3 that the issue of being unable to view subagent activity in the Edit window was finally fixed.
At this point, I genuinely wonder: what do I still need Cursor for?
Hey, thanks for the detailed feedback. There are a few different points here, so I’ll go through each one.
On choosing the model for sub-agents: yep, right now it’s tied to Max mode, and I can see this is a pain point for more than just you. The same topic is being discussed in a few threads:
The request to control the sub-agent model outside of Max makes sense, and I’ve passed it to the team. I can’t share an ETA.
On the picker showing options that aren’t available, like 300K or no-thinking for models that need 1M context: agreed, that’s misleading. Those options shouldn’t show up if you can’t select them on your plan. This is a helpful, concrete report.
On Agent Windows in 3.x: glad the sub-agent activity view in the Edit window was finally fixed in 3.3. If anything else in that flow still feels rough, please make a separate thread with steps and we’ll take a look.
On differentiation and workflows: this is a more strategic topic, so I can’t answer on behalf of the product team, but the feedback is valuable and I’ll pass it along. Detailed posts like this from long-time users really help.