Graph memory and the token tax nobody is paying attention to

Stuffing an LLM's whole memory into every prompt is the most expensive habit in agent engineering. Here is how the memory crowd is fixing it, and the four-way retrieval I keep reaching for.