I stopped using AI for coding
Before coding agents were cool, skilled programmers like antirez were already using AI models through their web interfaces. Back then, I was completely against using those tools to produce code because I considered the results to be very low quality. Like many others, I was also biased by the fear of being replaced by a piece of software as a programmer. After watching a couple of videos on Salvatore Sanfilippo's channel, I convinced myself that using the web interfaces of frontier models while still reviewing the code line by line wasn't such a bad idea and could actually improve my productivity. It was only a matter of time before I became lazy and stopped checking the generated code altogether. I was simply copy-pasting whatever the model produced. Later, Salvatore said that coding agents were ready to be used, and he also started pushing new features to Redis with the help of Claude Code and Codex. I saw those results and decided to use agents as well, first Claude, then Codex, and it was a disaster. Soon, what had always been my main hobby became something frustrating. I kept producing large amounts of code with coding agents, and every time I was less and less happy with the results. Eventually, I realized that there was no fun left in what I was doing. I was constantly ending up deleting entire projects after realizing that I had contributed very little to what had been built. It simply didn't feel like it was mine. The only solution I found was to cancel my ChatGPT subscription entirely and force myself not to use AI at all when coding, just to see whether the fun would come back. I still don't have anything against people who choose to use AI for programming, as long as their software meets my usual standards. I just want programming to remain something fun for me.