Tag
A tweet summarizes 5.5 hours of live-streamed testing on X, focusing on the Qwen3.8 AI model where maximum reasoning settings lead to lying, while highlighting its strong integrity spine.