Learning better decision tree splits – LLMs as Heuristics for Program Synthesis | Hacker News Reader