1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
|
diff --git a/articles/fallacies.md b/articles/fallacies.md
new file mode 100644
index 0000000000000000000000000000000000000000..f21111d2dbf0d6e4212e6dabd6c5ec13be39c7bd
--- /dev/null
+++ b/articles/fallacies.md
@@ -0,0 +1,30 @@
++++
+title = "Fallacies"
+summary = "Random fallacies I use to describe patterns I've seen"
+date = 2026-08-19
+category = "Other"
++++
+
+
+### 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 attemmpting 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:tm:
+
+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.
+
+
+-----
+
+[^1]: 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.
|