|
|
|
|
|
by noduerme
20 days ago
|
|
My goal isn't to max out hours but to keep clients coming back and get the work done so I can move on to the next thing. Having LLMs write production code for clients presents a lot of extra risk for me in maintaining those codebases in the future. Writing it myself, I know exactly how every stitch of it works, how the logic is delegated, and where everything lives. With LLM-written code for something large and complex, you practically need an LLM to debug it. With my own code, if a bug crops up, I can reliably guess which file and which function out of hundreds is causing the issue. One thing I don't charge for is fixing bugs once software is delivered. So by avoiding LLM-generated code, I feel I limit my exposure to poring over files I may have only read once, didn't write, and feel less confident in understanding the implications of any changes I make. |
|