Anthropic's Claude Mythos LLM has demonstrated new cryptanalytic capabilities, specifically targeting post-quantum and symmetric primitives. The model successfully identified a key-recovery attack on the HAWK post-quantum signature scheme, reducing the security level of HAWK-512 from a 128-bit target to a range between 81 and 108 bits, though the attack remains impractical for the full version. Additionally, the LLM achieved an improved key-recovery attack on a 7-round version of AES-128, though it failed to find more significant vulnerabilities in the full 10-round implementation. Despite these findings, the author argues that LLMs are unlikely to break established symmetric cryptography like SHA-3 or BLAKE3. This conclusion is based on the "messy" design of symmetric ciphers, which deliberately lack the clean mathematical structures found in public-key schemes, and the fact that most modern attacks rely on differential cryptanalysis—a field that has already been extensively explored. Because symmetric primitives like BLAKE3 rely on sequences of XORs, modular additions, and bit rotations to destroy linearity, they offer no high-level mathematical structure for an LLM to exploit. Consequently, while LLMs may assist in identifying errors in security proofs or attacking reduced-round variants, the inherent complexity and empirical robustness of established symmetric algorithms remain secure against large language models.