← All writes

AI / Culture / Networks

When LLMs spread like an idea

The virus metaphor is useful only if we keep it a metaphor: it describes patterns of copying, not biology or destiny.

Large language models are beginning to travel through products, classrooms, offices, and group chats with the speed of a good idea. One person finds a prompt, shares it, and a small behavior becomes a team habit. A template becomes a workflow. A generated sentence becomes the next person's starting point.

That is why the language of spread can be illuminating. It points to replication: patterns moving through a network, changing a little at every handoff, and gaining force because they are easy to reuse. It is not a biological claim. It should not make us imagine that people are passive hosts. People choose, adapt, reject, and remake what reaches them.

The question is what kinds of habits deserve to spread. A prompt that helps someone inspect an assumption may travel differently from one that makes every document sound the same. A copied workflow can free time, but it can also carry hidden errors or flatten local judgment.

I want to notice the social life of an AI tool, not only its capability score. Who teaches it to whom? What gets repeated without scrutiny? What would make a good practice easy to pass on? The answers matter because the next interface may become a tiny institution before anyone calls it one.