4/9/2026 at 2:09:09 PM
> Our tests gave models the vulnerable function directly, often with contextual hints (e.g., "consider wraparound behavior")."Often with contextual hints" is doing some heavy lifting here, IMO. I agree with the article's premise -- you don't need Mythos to use AI to find novel, complex vulnerabilities -- but these results as presented are somewhat misleading.
by tao_oat
4/9/2026 at 4:03:15 PM
AFAIU, their claim is that Mythos is in reality used in a framework that builds such contextual hints, and that their (Aisle's) own framework does the same:"(...) a well-designed scaffold naturally produces this kind of scoped context through its targeting and iterative prompting stages, which is exactly what both AISLE's and Anthropic's systems do."
by akavel
4/9/2026 at 7:52:27 PM
All evidence is point to LLMs not being sufficient for the taks everyone want them to do. That harness and agentic capabilities that shove them through JSON-shaped holes are utterly necessary and along with all the security, that there's no great singularity happening here.The current tech is a sigmoid and even using the abilities of the AI, novelty, improvements don't appear to be happening at any exponetial pace.
by cyanydeez
4/10/2026 at 7:55:48 AM
> The current tech is a sigmoidWhat makes you say that? I'm only asking because the data I've seen looks pretty cleanly exponential still, e.g. https://metr.org.
by tao_oat