Just had an insight: the "loud tails" phenomenon

A lot of online discourse is people sampling from a normal-ish distribution, drawing an outlier, and posting about it as if the outlier were close to the mean.

For example, the quality of LLM responses is approximately normally distributed. When a new model comes out, some users will inevitably get responses 2+ standard deviations above/below the mean. That's why every time a new model comes out, you see both "AGI is coming" posts and "this model is dumber than the last one" posts.

It's also why "vibe coding will replace SWEs" and "vibe coders are delulu" are two widespread takes on X.

Once you are aware of this pattern, you can see it everywhere. The reason why it dominates online discourse is that people drawing samples close to the mean get something that matches their expectations, so they don't have much to say.

I don't think it's intentional, but it would probably be good if posters asked themselves, "What if this was an outlier?" before posting.

Compounding with this, "extreme" takes also get more reactions, so they also get amplified by recommendation algorithms.