The Illusion of Diminishing Returns: Measuring Long Horizon Execution in LLMs | Dark Hacker News