I saw that the other ones were behind a pay wall and for the most part look like rulers with extra steps, so I started putting a few small datasets together and trained some bots how to spot different graded cards. That part was surprisingly easy compared to the coding part. I was able to hold it together for a little while but I kept getting back tracked and turned around. Half fixes I didn’t know were working or not.
After doing that for two weeks and getting a working web app going I heard about Cursor. I dropped my project into it and watching it make sense of my spaghetti code was the most satisfying thing ever. It took two or three hours to get through it all and I watched it get through everything I got stuck on.
So cursor took my app and made a useable version. Needless to say I kept using it. Ran through five hundred prompts real fast because I didn’t know what I was doing. So I started getting gpt 4o to prompt it for me. Made a lot of progress really fast, then plateaued.
So I started seeding prompts in my project and telling cursor about them. Worked great. Prompting from both ends gets it way closer to a one shot than just doing a really complicated front end version.
If you join that and get into the group the only clickable link inside is to its store page.
