WHEN CORRECT CODE IS NOT ENOUGH: ASSESSING THE PRODUCTION READINESS OF LLMGENERATED SOFTWARE
Large language models (LLMs) increasingly generate executable software, but functional correctness alone does not establish production readiness. This study evaluates GPT-5.6 Sol, Terra, and Luna across 15 algorithmic and production-oriented tasks using five generations per model-task pair, yielding 225 generated progr...