The US stock market faced a $1 trillion loss, prompting President Trump to label it a wake-up call, while hype grew around DeepSeek's advancements. Marc Andreessen, a prominent Silicon Valley investor, praised DeepSeek R1 as an extraordinary breakthrough and significant open-source contribution. DeepSeek's publication of R1 and its predecessor, V3, reveals that building reasoning models is simpler than previously thought, positioning the company as a top competitor against leading labs. This announcement spurred rivals like Alibaba and the Allen Institute for AI (AI2) to unveil new language models, claiming superiority over DeepSeek's offerings. Sam Altman, OpenAI's CEO, acknowledged R1's affordability but promised enhancements with upcoming models. OpenAI quickly introduced ChatGPT Gov, catering to US government security needs – a response to concerns about DeepSeek's data security. DeepSeek's innovative training methods, particularly its use of automation in reinforcement learning, allow it to develop effective models without heavy reliance on human feedback.

While this method excels in math and coding, it still needs human input for subjective tasks. The company employed cost-effective strategies in data collection and hardware optimization, leveraging existing technologies rather than acquiring the latest equipment. Although DeepSeek claims to have trained V3 for under $6 million, experts suggest the overall investment, including research and infrastructure, likely exceeds that figure significantly. The recent surge in reasoning models, including DeepSeek R1, is attributed to the emergence of robust base models that facilitate this development with minimal human oversight. By sharing its methods, DeepSeek has opened the door for future advancements, potentially leveling the playing field among smaller companies and reducing the dominance of larger firms in AI development.