My experience with agentic coding has pretty much been:1. This is fun and fast and the code is good enough2. I feel anxi...

My experience with agentic coding has pretty much been:1. This is fun and fast and the code is good enough2. I feel anxious sitting here waiting for an #AI3. I can’t really write code anymore4. I don’t trust my PR reviews because I don’t remember the API5. I am really uncomfortable asking the LLM questions about the approach because it always tells me “it’s good” and I am starting to just accept that.On the flip side going back to #tradcoding is also uncomfortable:1. It’s fun to code to solve programming tests2. I feel anxious taking all the time and effort to write something by hand when an LLM can do it fasterOverall I feel like #LLMs have destroyed me. They feel like an addiction. I don’t feel more productive using them. I think Language model tech is cool and could have good uses but using it for learning and producing stuff lacks efficacy.And that’s on top of how untrustworthy the owners are, and how exploitative the training is, and how environmentally unfriendly they are.