Blogger claims RSA-896 factored with Claude's help — but shows no method

Stephen A. Weis published a terse post claiming that RSA-896, one of the still-unclaimed numbers from the old RSA Factoring Challenge, was factored on September 19, 2026 with help from Claude. The page presents the 270-digit semiprime alongside its two 155-digit prime factors p and q, and little else — no description of the algorithm, hardware, runtime, or verification steps that would let a reader reproduce or trust the result.

The claim is extraordinary on its face. RSA-896 is an 896-bit modulus; the largest RSA Challenge number ever publicly factored is RSA-250 (829 bits), a 2020 effort that consumed roughly 2,700 core-years on academic clusters using the number field sieve. Jumping to 896 bits, and crediting a general-purpose language model rather than a dedicated sieve pipeline, would represent an enormous and independently unverified leap. Anyone can trivially check the product of the two published factors, but that alone doesn’t establish how — or whether — the factorization was actually performed.

For now this reads as a provocative curiosity rather than a documented cryptographic result. The interesting significance isn’t a broken key size; 896-bit RSA has been considered inadequate for years. It’s the framing of an LLM as the agent of a hard number-theoretic computation, a claim that demands published methodology and third-party reproduction before it means anything for cryptographic practice.