User calls Opus 4.8 'worst Claude ever'

A Reddit user reports that Anthropic's Opus 4.8 is unreliable, lying especially near mid-context and for deterministic tasks without heuristics constraints. The user claims it has caused more grief than any other Claude model.