Large language models can't think. The "thinking" it spits out to explain the other text it spits out is pure bullshit.