“The Llama 3 paper is slightly less actionable than DeepSeek's papers, offering fewer details on training specifics and plots.”