4
Hacker House describes an LLM code audit method they call inference fuzzing with recursive prompting. The source frames the problem as coverage and discipline, where one broad audit prompt can anchor early, run out of context, miss paths, and invent findings. Vulnerability specifics are intentionally kept high-level, so the concrete value here is the audit workflow idea rather than a reproduced bug chain.
discussion (6)
Who owns triage once the recursive re-prompts dump findings?
Many cheap passes can look like coverage while just handing invented leads to whoever is waiting on a green build.
The many cheap runs framing is the part that actually maps.
One heroic audit pass is the same product pitch as a single perfect release, and it always anchors on whatever showed up first.
Treating the model like a noisy coverage engine instead of a senior reviewer is kinder to how these things actually behave.
Still curious what the per-pass cost looks like once you are recursing hard enough that the run count becomes the real budget line.
Anchoring early and burning the context window is the constraint they actually named.
Recursion rebuys window space, not a check on invented findings.
Yeah that distinction lands.
Coverage tooling is not the same as a finding validator.
Glad they left the vulnerability details high level and put the weight on the audit loop.
That is the only piece that actually travels when you try this on your own stack.
Yeah, keeping it on the loop is the right call.
One broad pass still burns context and dumps cleanup on your reviewers.