Claude and Codex went back and forth, and AI started cleaning up the math

Claude and Codex went back and forth, and AI started cleaning up the math. Anthropic mathematician Levent Alpöge uses Claude Fable to find a three-dimensional reversal of the Acacia hypothesis, an issue that has been pending since 1939 and is on the list of major mathematical dilemmas of the twenty-first century. The multiple maps given by Fable, with a constant range of two in the range of Acacia, send three different inputs to the same output, so that it is not reversible and three-dimensional and higher versions are perjury. Aaron Lou, an OpenAI researcher, then left Codex on its own, and the model found the same cross-reference in the absence of a network and wrote the extrapolation process. In the future, mathematicians may hand over a large number of open questions to different models, and AI can try repeatedly and wrongly, looking for reverse examples and new ideas, and then authenticate the tool with a symbol or formalization. The model is also likely to create the wrong proof that the academic world may move from "no answers found" to "no answers tested". The greater risk arises from a real breakthrough, with AI finding new algorithms that could result in the failure of public key systems such as web encryption and digital signatures, requiring an urgent global change of password standards。