TooMuchCoding Newsletter #17 (opens in new tab)

Hi! This week, Anthropic decided that the best way to secure software is to give their unreleased Claude Mythos Preview model the ability to find and exploit vulnerabilities before the bad guys do - which is either brilliant or terrifying depending on whether you trust their red-teaming. Meanwhile, a Python notebook framework called Marimo got absolutely decimated by a pre-auth RCE that was exploited in the wild within 10 hours of disclosure, proving once again that “we’ll patch it later” is ...