Why I'm Dialing Back My LLM Usage
https://zed.dev/blog/dialing-back-my-llm-usage-with-alberto-fortinOpen linkView original on social.packetloss.gghttps://zed.dev/blog/dialing-back-my-llm-usage-with-alberto-fortinOpen linkView original on social.packetloss.gg
I’ve had a bit of the reverse experience- 2 years ago when I was first dabbling with the tooling, it kinda worked for writing some aws CLI shell scripts, but not well enough for me to trust it with anything serious. In the last 6 months, I’ve been pretty blown away with how much time it saves me- especially for rapid refactors. Is it perfect? No, not at all- but it regularly saves me hours a day, or enables me to complete “busy work” tasks in a minute or two that might otherwise not get done at all because there’s more important stuff that take priority.
I wonder how things will look in the next 6 months to a year.
I have seen a similar sort of back and forth.
Refactor these 50 classes in a similar fashion? Great. A series of small, but intricate tasks? Easier to do some searching, read a couple docs, and think through the change - than to try and guide the machine. Rewrite this GitHub action? Amazing.