LLM-Based Instance-Driven Heuristic Bias in the Context of a BRKGA | Dark Hacker News