
The Prompt Is Not a Security Boundary
Why instructions inside retrieved content must never inherit an agent's authority
Short, words-first books made for curious people. Every edition has one job: teach you something real.
Start with a question, then follow it through a small collection that belongs together.
Browse the whole library, or reshuffle when you want a fresh starting point.

Why instructions inside retrieved content must never inherit an agent's authority

Turn a model benchmark into a map of user promises and failure slices

A fluent answer can hide whether the evidence was missing or merely ignored

Why a longer conversation consumes serving memory even after the prompt is processed

What entanglement changes, and what it never lets you send


A change of measurement reveals what a fifty-fifty result leaves out


A visual guide to how boat engines turn stored energy into moving water

Total energy can remain accounted for while the differences that could drive change disappear

The cabinet cools only because the room receives its heat plus the compressor's work

Electrons take the outside road because ions alone can cross the inside separator