Raw LLM Responses

I started a project as a test back in the summer with no prior knowledge/experience of coding with AI. Lots of amazing realizations, and also some serious challenges. However, as I became more familiar, in 6 months I was able to bring an app to alpha, in python, typescript and deploy it on a droplet, none of which I have ever used in my professional career. Because I know and understand system architecture, I was able to get AI to do what I wanted, but that also was a process of creating my commands and guardrails. I am also paying for a $200 month subscription in order to process all my requests. The flip side, by AI's estimations, I did in about 1000 hours what would have taken a team of 3 (senior developer and 2 competent coders) appox 2.5 years. So I would say the acceleration rate in my experience is somewere between 12:1 andd 18:1 based on the analysis both myself and the AI have done in human hours. Yes you can get it to do some jobs autonomously, but the less you are engaged with what AI is doing in each process, the more likely you are going to end up with a mess and iterating over the bad code to either remove or fix it. Also I can confirm that AI is EXTEMELY good with db/backend. This process went so fast it literally blew my mind. Where it struggles is with frontend development, and things slow down a bit. Unit and Mock testing is also very fast and again I was quite impressed. My takeaway from this video is the overall analysis as to where and how AI fits in is very accurate and I have over 1000 hours in. As for languages I am familiar with, or projects I have coded myself, it's an invaluable tool for debugging and optimizing. But again, you have to understand architecture, all the actors involved and where your separations of concerns lie. AI likes to make assumptions about what you are doing and thats where it gets you in trouble.