So there is definitely still an art to coding with AI tools then. It’s not as simple as just asking it to create anything you want and sitting back expecting it to do all the work. I’ve not cracked it yet, and I experience all the same issues discussed here. For this reason I don’t use Replit agent, but just rely on Assistant instead. I don’t think agent is brilliant at being collaborative at all. Can’t explain why, just not had great results with the Agent. The assistant however is the perfect companion to me. Things that work well for me are first of all having a solid understanding of what your code is doing, in terms of app flow / pipelines. Before it applies changes, check the code changes it’s proposing, and if something doesn’t make sense, ask the assistant about it. Half the time it’s possible that the assistant has interpreted something wrong you want, which you can then correct and get a better fix. The other half of the time, it’s something you don’t understand, and the assistant can then explain its proposed changes and you in turn learn to be a better engineer / developer. I also find that when I want to I troduce a new functionality, you can ask the sssistant to set it up first as a standalone working script in your project. Get it working independently as a script file, and then when it works well, you can try and integrate it into the rest of your build. Isolate things, get them working, then integrate. I think the assistant is fantastic. I get all my prototypes up and working quickly in Replit, then pass them of to someone else to productionize in AWS or whatever.
When you say productionize what do you mean
I just mean deploy. I know you can do this on Replit, and I often do, but we prefer to deploy to AWS if there is an inkling something might scale or get some serious use.
Oh nice. I’ve been trying to deploy outside replit but to no luck. Could you send a link to a YouTube video on how to deploy outside replit to aws or write a small thread on it here