Your memory layer is lying to you (and your LLM agrees)translating…

The verify-on-read experiment (1-V) used a deterministic proxy agent to catch false claims in memory before surfacing them to the user. Proxy FA=0 by construction — that's a useful property, but it tells you nothing about what a real LLM would do with the same claims. A…