x.com/NytroRST/status/2049931767004447035
1 correction found
Finding Zero-Days with Any Model
The linked article does not show zero-days can be found with *any* model. It reports success with a few specific models and explicitly says at least one tested model, Qwen 3.5 distilled, could not run the workflow.
Full reasoning
The post's wording is universal: "any model" means the result should hold across models generally. But the linked article describes successful runs only with specific models — Opus 4.6, Sonnet 4.6, and Z.AI's GLM 5.1 — and then directly states that a different tested model failed: "a smaller distilled candidate I tested (Qwen 3.5 distilled) could not run the workflow."
Because the article itself identifies at least one model that could not perform the workflow, the title/post overstates the evidence. A more accurate description would be that the author found zero-days with some non-frontier and open-weight models, not with any model.
1 source
- Finding Zero-Days with Any Model
Using a specialized vulnerability discovery workflow, I ... discovered new zero-day vulnerabilities ... with commercial models like Opus 4.6 and Sonnet 4.6, as well as open-weight models like Z.AI's GLM 5.1. ... Truly local inference on commodity workstation hardware remains an aspiration rather than a demonstration here: a smaller distilled candidate I tested (Qwen 3.5 distilled) could not run the workflow.