Snoop as the Unsafe Live Eval

Daniel deliberately gives Snoop aspirin and monitors the dog’s collapse to test whether an earlier aspirin exposure could explain Snoop’s illness and support Sandra’s account. The experiment yields decision-relevant evidence, but it is performed on a living subject without a sandbox, safety review, or reliable rollback.

It is a vivid example of obtaining useful causal evidence through a high-stakes live evaluation whose harm is part of the measurement process.

This is a desperate forensic experiment rather than an AI evaluation, and Snoop survives.