It's different, I wouldn't necessarily say it's worse than Copilot, and I have been happy making the switch given my usage is included with the Jetbrains All Products subscription I already pay.

If you're expecting inline completion that reads your comments and shits out a fully formed function like Copilot, you'll be pretty annoyed - but that's been a positive for me as the completion doesn't constantly try to get in my way and suggest things 10 steps ahead of me. I'm much happier with my assistant letting my continue to steer when I'm actively coding, and instead just focusing on reducing keystrokes.

Chat works great, and I have a whole host of models from multiple providers to pick from. If I actually do want to have it generate some garbage for me, do code review, or simple refactors then I'll whip it out and it does the trick. Being better integrated into the IntelliJ Platform, actually applying changes from the chat window works more often than Copilot did in my experience.

And then there's Junie, which also benefits greatly from integration into the IDE. Since the PSI syntax tree you get from native IntelliJ language plugins gives the IDE a lot of data on whether code is actually valid or not (missing references, hallucinated packages or standard library functions, straight up invalid syntax, etc), you're not relying solely on tests to make sure the end result of a task you set the agent off to do is actually correct. I can come back after asking her to do a large-scale refactor, or implement some (part of a) feature, and generally be confident I'm not wandering into a bunch of garbage.

If you already pay for an All Products Pack license you've got JetBrains AI Pro included, if you pay for a single IDE from them and are already at the 3-year continuity discount then the cost of upgrading to the All Products Pack and getting the AI Pro license is a couple USD difference (depending on the specific product, could be cheaper). Really not much for anybody to lose trying it out, at the least.