When do you scrap a project and restart?

Sup! So check this out. I’m a construction guy with minimal coding knowledge and experience. I have, however, an idea that has the potential to make a niche area of construction 10x better.

Simple in concept: a marketplace app where Contractors post Jobs, which get broadcast to Workers who can signal interest, Contractor hires Worker, then we finish the process with 2-way reviews. Repeat.

I got an almost-MVP working. Got Contractor and Worker user types down, authentication, Job Posting function, rudimentary comms thru Jobs and a very basic UI. Separate development and production environments.

And that’s where progress stopped. To make an MVP that actually works, I need a real-time notification system and a review system.

I attempted to get Agent to implement the notification system. Knowing nothing about this, I trusted Agent to implement Websocket notifications. It kinda works, but it feels like development has slowed to a crawl. I think maybe it’s because the app is now too big for the LLMs to manage (too much context), or because I gave inconsistent instructions and now the thing is a Frankenstein.

Reading thru this forum I’ve learned some good guidelines, such as establishing a DevGuideLines.md document right off the hop, testing and debugging like a psycho, documenting like a psycho, building in reusable components as much as possible, committing to Git and branching like a psycho, basically be psychotic or obsessive compulsive with organization and cleanliness. Basically the exact opposite of my personality type. Ugh.

Anyway, the point is that I feel like I put in so much work to this project but I also feel like starting over might be better in the long run because I can build it cleaner with the knowledge I have.

The other part of me thinks that maybe this is too complex of a project for an LLM to manage without an experienced developer at the helm??

Anyway, if you guys want to check out the mess so far, you can go to the demo environment (frameworkdemo.replit.app) where you can apply for Jobs such as ‘Twerk dat azz’ or ‘Get a boner’, and feel free to create Contractor accounts, Worker accounts, post Jobs, etc. PS If those Job titles offend you, we’re not a good fit to work together :).

Someone help!! Maybe @robinbobinpopin ? You seem like a cool individual :slight_smile:

That was the same experience I had. The moment websockets for notification and OTP integration became too complex for the LLM and we got tons of issues. It would be great to figure this out because we do have an enterprise application concept to get build out asap.

Hey @brandon263, I was thinking of scrapping the Websocket implementation in favor of long polling. If there isn’t too much network activity it could be a good, less complex base for a notification system. Unfortunately I didn’t branch off in Github so I can’t roll back.

What are your thoughts on Long Polling?

Hey man! Send me a DM! Let’s schedule a time and maybe I can help you!

1 Like

You might need to hook it up to zapier to trigger the notifications. Maybe try having it build a module around that. I’m on my 2nd build of my diamond website and I’m about $400 in a believe. The first one was so hard to scrap, I was honestly depressed for a few days and then I took a week off, got sleep and I was able to rebuild the second version in 48 hours from the previous 28 days. I was able to remedy some of the issues I had “lived” with on the first build and then after that agent 2 came out in beta and was able to solve some of the other problems for me. I have gone pretty far, I’m actually held up right now due to a weird billing issue so I’ll report back if I crash and burn on this build. I will be honest if the app is working, try running it on agent 2 to give it a wack, Claude 3.7 is pretty good.

1 Like