Skip to content

Changelog

What we shipped.

Public log of feature releases, reliability events, and methodology changes. New entries every 2–3 weeks.

RSS coming soon · methodology revisions tracked in /methodology

  1. v1.6.0 · Feature · May 5, 2026

    Public AI Visibility Index — free for any brand

    Anyone can now check how often a brand is cited in ChatGPT, Claude, Perplexity, and Google AI Overviews — no signup required. We pre-indexed 20 popular B2B SaaS brands and added per-engine improvement playbooks. Try it on your brand.

    • → 20 brand pages live, refreshed weekly
    • → 6 category leaderboards (CRM, project mgmt, AI coding, data warehouse, observability, support)
    • → Per-engine 'how to improve' actions tied to current score
  2. v1.5.2 · Reliability · April 28, 2026

    Canary panel rebaseline after Sonnet 4.6 release

    Anthropic shipped Sonnet 4.6 on April 22. Our canary detected the shift within 14 hours; we rebaselined customer trends so no spurious drift alerts fired. Model event logged on /methodology.

    • → Detected via cosine drop to 0.87 (threshold 0.92)
    • → All Pilot accounts auto-rebaselined
    • → Methodology v1.2 published with the post-mortem
  3. v1.5.0 · Feature · April 14, 2026

    Slack drift alerts (Pilot+) and webhook routing (Growth+)

    Drift alerts now fire to a Slack channel of your choice — only when |Δp|/SE > 2.0 (no noise). Growth tier gets raw webhook payloads to plug into your own routing.

    • → Native Slack OAuth install — no copy-paste of webhook URLs
    • → Per-channel routing (e.g. CRITICAL → #pr-alerts, MEDIUM → #marketing)
    • → Webhook signed with HMAC SHA-256
  4. v1.4.0 · Feature · April 7, 2026

    Sentiment vector v2 — 89% accuracy on hand-labeled benchmark

    Replaced our v1 sentiment classifier with a fine-tuned model trained on 2,000 hand-labeled descriptor sentences from real BrandMirror reports. New head accuracy: 89% (up from 78%). Ships transparently — old reports re-scored with both versions side-by-side.

    • → F1 macro: 0.84 across positive/neutral/negative
    • → Reduces drift-alert false positives by ~40%
    • → Benchmark dataset available under NDA for peer review
  5. v1.3.5 · Compliance · March 31, 2026

    SOC 2 Type I audit kicked off

    Engaged a Big-4 advisory firm for SOC 2 Type I. Audit window April–July 2026. Type II observation period starts immediately after. Expect Type II report Q4 2026.

    • → Auditor disclosed under NDA on enterprise calls
    • → DPA with Standard Contractual Clauses live for EU customers
    • → Penetration test report (Cure53) available on request
  6. v1.3.0 · Feature · March 17, 2026

    Persona-conditioned sampling rolled out cluster-wide

    Every prompt now runs across 3 buyer personas per cycle. Internal benchmarking showed 73% of comparison prompts return different leaderboards across personas — single-persona monitoring was missing two-thirds of reality.

    • → Default personas: founder, RevOps lead, enterprise IT
    • → Customer-defined personas in Pilot+ tiers
    • → Per-persona delta visible in dashboard

Older entries archived. Email hello@brandmirror.dev for the full history.