Sonic Fallacy #
Gotta go fast!
The Sonic Fallacy is when someone uses or advocates for a specific language or framework simply because "it's the fastest!" without profiling or otherwise attempting to discern whether it's even required.
Of course, no one wants to have slow code. However, the Sonic Fallacy generally describes the choice being made primarily because of speed, regardless of whether the language or framework is easy to use, memory safe, or any other number of trade-offs that can occur in the name of speed.1
Claude Fallacy #
Everyone writes slop AI code except me because I Know How To Use It Properly™️
The Claude Fallacy could be described as "The Dunning-Kruger effect for LLM users".
It is the dichotomy among certain LLM users that LLMs produce slop when other people use them, however their own LLM does not produce slop because it is being used in a superior manner.
While there is some discussion to be had about the efficiency or results of various prompting techniques, harnesses, agents, etc., overall in my (completely non-scientific) experience the average LLM project is the same.
-
I manage to support an entire state using Python/Django. I'm not saying that means Python is some perfect language, I'm just saying your project being in Rust doesn't automatically make it a unicorn. ↩︎