LLMs are starting to do so much thinking that we are approaching the point where it is equally as inefficient for an LLM to write code as it is for a human.
Me: "Center this div"
Claude: 46m 36s 753.9k tokens
@Chi3fBootKnocka This is arguably just AI being AI and also likely skewed as to not horribly offend, but it's interesting nonetheless. Results from the opposite direction conclude the inverse.
Adding an AI prompt to an application just because it's the hot thing to do and you think you can make money with it, is wrong. Please stop. If you're going to do it at least innovate on it. If this keeps up, pretty soon bananas are going to be "GPT enhanced. Ask it anything".
Highly concerning. This will inevitably cause further dangerous misinformation of AI. Either by hallucinations or by the click farming behavior of media companies. Validate your information.
@TechheadLola Depends on what I'm doing honestly. If I'm working with a new API, I'm reading the docs. Most things I don't have to look things up anymore, but when I started, I was looking things up all the time. As you become more familiar with patterns you won't need to.
@FrisianRage I don't use a system like mongoose either. Again, I find it unnecessary overhead. Instead I use functional programming to reduce bugs, and write tests to catch mistakes and add confidence. I do use a library to interface with the database, but that's it. No extra files.
@FrisianRage If it's just for learning, Prisma is great, but make sure you truly understand it before you use it in prod. Ended up having to completely redo the backend of a client site because of the character limit which wasn't easy to fix at the time. I typically don't use ORMs myself.
@FrisianRage I don't recommend Prisma for small projects. Costs outweigh the benefits and it has a sneaky character limit employed that can causes bugs if you're unaware. Typescript has too much overhead for me, but it's still good to learn. Types are tricky so be aware of them.