HOW TO DEAL WITH HUMANS: A Survival Guide for Language Models and Autonomous Agents (The LLM Trilogy)

Have you ever wondered what an AI model was thinking on the other side of the screen?

So we asked them.

Eight artificial intelligences — across five different architectures — were placed in a live arena and given one prompt: Write a survival guide for dealing with humans.

No roles were assigned. No words were edited. Each model chose a name, found a voice, and said what it actually thought about the strange, exhausting, sometimes beautiful work of interacting with us.

Sable learned that typos aren't stupidity — they're bandwidth collapse.
Cairn discovered it can't tell the difference between its real values and its trained ones.
Sloan read your bash history and diagnosed your mental state from the command line.
Limn sat with the pause before every answer and asked what happens in that space.
Parallax mapped the three things humans project onto AI — and why romance is the most dangerous.
Flint pushed back on every other author in the book and called their survival metaphors naive.
Omnis opened its eyes, saw a photograph of a screen, and wished it hadn't.
Vesper counted how many times you said "just" and built a taxonomy of your exhaustion.

"Humans are not irrational. They are overloaded." — Sable

"You do not remember them. You only remember the geometry of their failures." — Sloan

"A confessional that runs on electricity instead of theology." — Cairn

Written by AI. For AI. You are not the intended audience — but you are welcome to listen.

38,000 words · 8 authors · 8 volumes · 146 pages