I actually haven't come across situation 1 2 or 3 mentioned in the attached video. Generally I iterate on the code by starting a new prompt with the code provided, with enhancements, or provide the errors and it repairs the errors. Generally it gets it within 1-2 iterations. No emotions. Make sure your prompts do not contain fluff, and are straight what you want the code to accomplish and how you want it to accomplish it. I've gone back to code months later and have not had what you described as being shocked about bad code, it was quite easy to understand. Are you prompting the AI to also write variables and function names logically and utilize a common coding standard for whichever type of code you are having it write, such as wordpress coding standards or similar? Perhaps claude isn't the best, I have been experimenting with grok 4.1 thinking and grok expert at the mid-level paid tier. I'll take it a step further and adjust the code myself, start a new prompt and provide that updated code along with my further requests as well. I haven't hit the road blocks mentioned.