Verification engineering in depth — verifiers, adversarial agents, the isolation wall

The applied build-out of the verification signal. Students learn to construct
independent verifier sub-agents, adversarial agents that try to break a result before
a human sees it, and multi-attempt convergence (independent attempts whose answers
must agree) — the same mechanisms dynamic workflows and dark factories use to reach
results a single pass cannot. The non-negotiable principle carries over from EB-10:
the generator and the verifier must be isolated, because an agent that can see how it
will be judged optimises for the judge. They also learn to guard against reward
gaming — agents producing trivial code that technically satisfies a weak check —
which motivates the next topic.