AI is finally getting good enough to code with. It still however makes horrendous mistakes. After many many hours of dedicated persistence and baby sitting this can occur. AI is still not at the point where it is viable to use to create projects like this on it's own or without dedicated patience to watch it or back it up or to concoct ways of making it behave in a way that is going to create progress. I think for the most part that the owners of LLM's Need to Restructure their pricing in a way that makes sense to the end user. I tried paying for it once and that only gave me about 5 hours of programming and that was mostly used to fix mistakes the ai made.  Some things I have observed is that you need to give AI the chance to create plans to fix the mistakes and to propose 3 different methods of fixing the problem then choose the best method. It becomes more like teamwork that way. So this is why we must stay  on the free tier and the ones that were used to create this. were like this.

I used some Copilot as much as I could in MS Code but it has the added issue of when you want to chat Pilot eventually wants money to use and then you have to pay the models on top of that so No! not good enough Microsoft. Then I finally found the Terminal it is where the magic happens AI cli is the greatest. If your not coding this way then you are coding wrong. I find that Googles Gemini has the best free tier but you have to code in streaks as it will shut you down in a few hours of programming but it is able to handle the complex tasks. I also used Kero and was able to get a lot accomplished on the free trial but their free tier sucks so I quit using it. I think the best is Claude 4.5 but it is way to expensive and not worth the price. So google ended up Kicking Claude's Ai in Value and Usability. I didn't get to try GPT-CLI  Yet but will when I decide to make a new project but none work long enough to give you a really long thought out process of programming. 

It will be nice to find out what is like to have a LLM actually think through and entire project I can see it is finally getting to the point where is even slightly usable for programming but it still messes things up in a grand scale by piece milling code.