I noticed that some features are available inside the code editor, but not in the agent window, such as /best-of-n and others.
This creates an inconsistent experience, because users have to keep switching between the code editor and the agent window just to access certain features.
It would be much better to unify the experience so the same tools and features are available in both places, allowing users to continue working smoothly whether they are in the editor or the agent window without constant context switching.
The goal with the Agents Window is to start from a blank page and build an experience that isn’t constrained by everything that was built with a different workflow in mind. The editor and the agent window are designed around different assumptions: the agent window is built to let agents do most of the work, where you’re not focused as much on the code itself, while the editor is optimized for hands-on coding. We’re supporting both, and keeping them distinct is part of what lets each one be good at what it’s for!
The tradeoff, as you’ve noticed, is that some features haven’t reached parity yet. Some things will make the jump, some won’t, and some will look pretty different on the other side. We’ll continue to iterate on this and close the gaps over time. The public release of the Agents Window is only about two weeks old, so there’s a lot still coming.