The leap from Agent v1 to v2 feels like an order of magnitude improvement. The chain of thought visualization with occasional interjection around next steps is such a huge improvement. The speed at which I can create a fully functioning prototype that is deployed to the web feels unparalleled to anything I’ve done before. 
I would consider any comments about Agent from 2/24/25 or previous to be outdated.
My aha moment 
I’m not a developer, I don’t really know how to write code. I was having an issue figuring out a data related item and I asked Agent v2 to create a page on my app with an ERD (diagram of the database tables and relationships), schema, api’s and explanation of the data flows of the app. In one good prompt it got 90% of the way there in maybe 2 or 3 minutes. Then I spent another 10-15 minutes tweaking but still…MIND BLOWN. 
RANT: If you made it this far, thanks!. Most complaints I see I attribute to poor prompting and explanation of the desired goal. How often does a fast moving person explain in crystal clear details with full requirements and consider edge cases the first time? Right. Most times I have issues it’s because my prompt stinks and I didn’t take the time to think it through. We can all do better.