Recent LLM models have provided “reasoning” capabilities. Birgitta Böckeler asks what role these play with coding tasks. She doesn't have an answer, but does have questions and thoughts - and has not found such capabilities worthwhile so far.