Cover Story

Why
we built
noisgate.

01 · Philosophy

What makes this different

Most tools start from a label and explain it. We start from what an attacker actually does and reverse-engineer the rating from there.

  1. Practical exploitability over theoretical severity
  2. Context from public research and field-relevant references
  3. Peer review & validation data to challenge weak assumptions
  4. Output meant for analysts, defenders, and leadership conversations
02 · Audience
For security teams drowning in scanner output, backlog, and pressure to justify prioritization decisions. If you need a cleaner explanation of what matters now and what can wait — that's the point.
03 · Operating Practice

Turn assessment data into operating practice

To get real value from the platform, your organization should define clear timeframes for both mitigation and remediation.

A good pattern is to introduce mitigation first. That creates space for operations to breathe while still reducing risk in a controlled way. The sweet spot is a tight mitigation target with a looser remediation target.

ReassessedMitigationRemediation
CRITICAL≤ 3 days≤ 90 days
HIGH≤ 30 days≤ 180 days
MEDIUM—≤ 365 days
LOW——
04 · Changelog

Release notes

v1.4.1 · Latest
Pipeline calibration
  • Reasoning effort recalibrated for steadier throughput.
  • Controlled A/B test confirmed verdicts unchanged.
  • Corpus passed 1,600 assessments.
v1.4
Verdict voting
  • Agree / disagree signal on every verdict band.
  • One vote per reader per CVE, counted live.
  • Report timestamps now read from real assessment data.
v1.3
Engine migration
  • Hot path moved to the Claude Agent SDK.
  • Built-in web search replaces the bespoke tool.
  • Turn loop managed and capped per assessment.
v1.2
Blast-radius scoring
  • Deployment-role rule floors verdicts by real blast radius.
  • Domain, CI/CD, hypervisor, and edge roles enumerated.
  • Friction can raise a verdict — never pull it below the floor.