Move over, Mythos. Here comes... pretty much any other model with a good harness

This article analyzes the role of large language models in vulnerability discovery, arguing that orchestration/harness design and cost-effective multi-model workflows produce better, scalable results than relying on a single top-tier model (Mythos). It highlights examples (false positives in cURL findings, Fable 5 guardrails), outlines common harness stages (recon, hunt, validation, tracing, deduplication), and recommends vendor-agnostic, multi-tier approaches for sustained AppSec and AI-assisted pentesting.