Anthropic has submitted a revised S-1A filing. Beyond financial data showing sharp surges in both revenue and spending, the filing details reveal that the company is very close to developing a model with recursive self-improvement capabilities. Recursive self-improvement refers to an AI's ability to improve itself during training to a certain point, triggering a dramatic leap in capabilities. Views on its impact vary, and Anthropic has incorporated the associated risks into its Responsible Scaling Policy monitoring. Previously, many dismissed recursive self-improvement as merely a buzzword or something difficult to achieve. This time, Anthropic added an explanation stating that it has used AI to develop a new model with capability jumps that went unnoticed at the time. According to internal estimates, the company may soon meet the threshold defined in its relevant policy as "a sharp acceleration in progress brought about by AI," and there is a possibility that this acceleration could be discovered only after it has already occurred. Currently, no model has crossed that threshold. Anthropic has promised to keep humans in the R&D process for the foreseeable future, not allowing Claude to fully autonomously design and build the next generation of frontier models without effective human oversight, and is considering revising the relevant policy to ensure that frontier model development always remains under human control.