Has Cursor Gotten Worse Over the Last 4 Months?

When I first started using Cursor, I was blown away. With a single prompt, it generated clean, multi-file codes that mirrored exactly how I would have written it myself.

It understood my stack: GraphQL for APIs, Prisma as the ORM, and Next.js with Ant Design on the frontend. It generated the Prisma schema, GraphQL schema and resolvers, and even the frontend components, all in my style. Like magic.

It felt like a senior engineer who could read through my codebase, understand my coding patterns, and implement a new feature without much guidance.

This was before Cursor implemented the ‘Agent’ mode.

Then I took a break from coding to focus on business.

Now that I’m back, the magic seems to have disappeared. Cursor still helps, but I have to guide it step by step, constantly remind it of my preferences, and nudge it to follow my conventions.

It feels more like a junior engineer who needs a lot of hand-holding.

It feels… less intelligent.

Curious if anyone else has had the same experience?

#AI #Cursor #DeveloperTools #CodeGeneration #LLM #SoftwareEngineering #BuildInPublic