I finally started using @v0 by @vercel and it's single-handedly the best vibe code editor that exists, bar none. It's got the comprehensiveness of @Replit with the AI model intelligence of @convex_dev Chef. Crucially, it has both built-in integrations with various other tools and a great version control system for the iteration you need for any codebase building. Interestingly, I previously never considered it because all the reviews and AI summaries incorrectly labeled it as a one-off components editor rather than a full-scale IDE. This isn't correct - it is a true IDE with prompt to app deployment capability. But AI uses preponderance of data to assess data accuracy so rapidly changing tools that are described by AI will have a bias to their earlier and more reviewed iterations. Has anyone noticed this for other applications of AI research? Is there a way that model training attempts to correct for this pernicious error?