Backtracking Through Agent Memory – Adventures with LLMs, Prolog and DeepClause | Hacker News Reader