Changelog

Track every update, new feature, and improvement as Queen Mama evolves.

October 2026

New

Features

  • Try Without an Account — New users on macOS can start right away with "Continue without an account" and turn it into a full account later without losing their sessions
  • Code, Math and Diagrams in Answers — The macOS overlay now shows syntax-highlighted code, formatted math and diagrams, handy for technical interviews
  • Smarter Audio During Silences — macOS now detects silence locally and stops streaming audio until someone speaks again, with no change to transcription

Improvements

  • Newer AI Model for Assist — Real-time suggestions now run on GPT-6 Luna, with more up-to-date knowledge and the same response speed
  • Claude Sonnet 5.5 for Smart and Recap — Smart mode and meeting recaps now run on Claude Sonnet 5.5 for sharper analysis and summaries
  • What to Say Appears Instantly — On macOS, What to Say and Recap now write their answer as it is generated instead of showing it all at the end
  • Button Hints in the Overlay — Hover any overlay button on macOS to see what it does and when to use it
  • Decode Button Now Optional — The Decode button is hidden by default on macOS to keep the overlay compact; turn it on in Settings

Bug Fixes

  • Recap Uses the Right Model — Meeting recaps on macOS now always run on the dedicated recap model instead of the faster Assist one

June 2026

Features

  • iOS Native Subscriptions — The iOS app now supports Apple in-app purchases — subscribe or change plans directly without being redirected to a browser
  • Speaker Identification on iOS — The iOS app now tells you and the other party apart in the transcript, using the same dual-stream audio analysis as macOS
  • Stays Active Offline (24h) — The overlay reuses its last valid proxy configuration for up to 24 hours when offline — coaching keeps working even when connectivity drops mid-call
  • Enterprise Plan — A new Enterprise tier unlocks premium features including live DeepL translation and higher device limits — available on iOS and macOS
  • Device Management — New dashboard page lets you see all connected devices and revoke access from anywhere
  • Per-Plan Device Limits — Each subscription plan now enforces a device cap with automatic cleanup of stale sessions
  • Live Translation on iOS — Enterprise subscribers can now translate conversations in real time with DeepL directly in the iOS app

Improvements

  • System Audio Health Warning (macOS) — The macOS overlay now detects when the other party's system audio isn't being captured and clearly flags it (showing "Mic only"), so a dropped audio stream never goes unnoticed
  • Smarter Context Awareness — Context intelligence is re-enabled with improved relevance scoring — responses now track topic shifts and recover instantly after a connectivity drop
  • Coaching for Your Colleague — Default Assist now detects when a colleague is struggling and suggests how to help them recover — not just advice for your own performance
  • Richer Screen-Only Responses — When there's no active conversation, AI responses now proactively surface the most useful insights from what's visible on screen

Bug Fixes

  • Dictation Stops on Submit — Dictation now automatically stops when you trigger an AI request, preventing the mic from staying on unintentionally
  • AI Fallback for All Tabs — What to Say, Follow-Up, and Recap now properly cascade through backup AI providers — no more silent failures when the primary is unavailable
  • Stable Login on Startup (Windows) — Fixed a race condition where multiple login windows appeared at startup, sometimes causing the session to be lost
  • Windows Updates via GitHub — App updates are now served from GitHub releases, bypassing corporate proxies that were blocking the update server
  • iOS Login Screen Fix — The registration and login screens on iOS are now scrollable with a working back button
  • Diarization Reliability (iOS) — Switched to a more stable transcript stream on iOS, reducing speaker detection errors during calls
  • Transcription Language Display (iOS) — The iOS app now correctly shows the transcription language as Automatic when no language is manually selected

May 2026

Features

  • Live Translation on Windows — Real-time DeepL translation is now available on the Windows app with the same dedicated Translate tab as macOS
  • Live Translation Tab — Translate the other party's speech in real time directly inside the overlay — a dedicated Translate tab keeps a full history with source and target text side by side
  • Smarter Speaker Detection — Transcript now distinguishes you from the other party more reliably during meetings, with cleaner labels and fewer mix-ups
  • Choose Your AI Model — Pick which AI model powers your standard responses — select directly from Settings to match your speed and reasoning preferences
  • Live Transcript Toggle — Show or hide the real-time transcript from Settings — useful when you want the overlay focused on AI responses only
  • In-App Pricing Modal — Browse plans and upgrade directly from the app without leaving your workflow
  • Windows Auto-Updates — The Windows app now updates itself automatically with clear notifications when a new version is ready
  • Refined Windows UI — The Windows app gets a deeper visual identity with a more cohesive and polished interface

Improvements

  • Live Translation On by Default — Live Translation and the Translate tab are now enabled out of the box — open Settings → Translation only if you want to turn them off
  • Smoother Translation Across Chunks — DeepL now receives rolling conversation context, keeping terminology and phrasing consistent across consecutive transcript chunks
  • Stronger Response Language Lock — AI responses now stay locked to the detected meeting language even on long sessions, preventing mid-conversation language drift
  • Localized Overlay Speaker Labels — Speaker labels in the overlay (You / Other) now follow your app language instead of being hardcoded in English
  • Live Settings Sync (Windows) — Changing a setting in any Windows window now propagates instantly across the dashboard, overlay, and other open windows
  • Manual Language Override — Force AI responses into a specific language using the LANG_OVERRIDE directive in your mode prompt, bypassing automatic detection
  • Stabler French Transcription — Transcription now stays locked on French when set as primary language, preventing drift to English mid-meeting
  • AI Cascade Fallback — If the primary AI model fails, Queen Mama now automatically cascades through backup providers so you never lose a response
  • Unified Notifications (Windows) — Notifications now route through a single central toast system for a more consistent and less noisy experience

Bug Fixes

  • Sharper Tab Specialization — What to Say and Follow-up now have distinct prompts from Assist, giving each tab a clearer purpose and more targeted output
  • No More Duplicate Knowledge — Knowledge extraction now deduplicates atoms, enforces a cap, and throttles re-extraction to prevent the same fact from being saved twice
  • Fresh Manual Triggers — Manual tab triggers now always generate a fresh response instead of returning stale cached output
  • Fewer macOS Sign-Outs — Transient auth refreshes are now retried with backoff and only force a sign-in screen when the session is genuinely revoked — no more random disconnects after the laptop wakes from sleep
  • More Reliable Session Sync — Session sync now retries Neon's idle-disconnect drops, so syncing right after the server has been quiet no longer fails on the first attempt

April 2026

Features

  • Smarter Speaker Detection — Transcript now distinguishes you from the other party more reliably during meetings, with cleaner labels and fewer mix-ups
  • Coaching-Style Assist — The Assist mode now gives direct, actionable coaching — concrete points to mention instead of generic advice
  • Smart Mode for All Tabs — What to Say and Follow-Up now work correctly in Smart Mode with deep reasoning from Sonnet 4.6
  • 10 Daily AI Requests (Free) — Free plan now includes 10 AI requests per day instead of 1, making it much more useful for getting started
  • Speaker Identification (Windows) — Speaker identification with dual-stream audio analysis is now available on Windows
  • Single-Stream Fallback — Automatic fallback to single-stream transcription when system audio capture is unavailable
  • Meeting Cost Tracking — See real-time AI spend per meeting and configure custom cost thresholds from Settings

Improvements

  • Faster AI Responses — GPT-5.4-mini as primary model with prompt caching reduces response time by up to 44%
  • Real-Time Transcript Context — Assist now captures the very latest speech including words being spoken right now for more relevant responses
  • Laser-Focused Context — Assist responses now focus strictly on the current topic being discussed, ignoring older conversation
  • Improved Language Detection — Responses now consistently match the transcript language — French transcript always gets a French response
  • Better Response Formatting — Bullet points now render correctly on separate lines regardless of the AI model used
  • Clean Overlay Between Sessions — The overlay now automatically clears previous session's AI responses when starting a new session
  • Responsive Pricing Modal — The pricing modal now adapts properly to the overlay widget size
  • Windows Streaming Performance — Windows app now matches macOS streaming performance and AI response quality
  • Speaker-Aware Coaching — Assist no longer puts your interlocutor's words in your mouth, and shifts to insight mode when you're silently listening instead of forcing a reply
  • Mode File Attachments — Drag and drop files directly into a mode, remove attachments, and keep them across sessions
  • Windows AI Coaching Parity — The Windows app now applies the same language lock and speaker perspective guards as macOS for more reliable coaching

Bug Fixes

  • Smart Mode Error Fixed — Fixed an issue where Smart Mode returned empty responses due to incompatible API parameters
  • Token Refresh Stability — Fixed a race condition in token refresh that could cause authentication failures
  • Windows License Sync Fix — Fixed license status mismatch between dashboard and overlay on Windows

March 2026

Features

  • Windows App Available — Queen Mama is now available for Windows 10+ with automatic updates and full feature parity
  • Cross-Platform Downloads — The download page now auto-detects your OS and offers the right installer for macOS or Windows
  • Full French Interface — The macOS app is now fully available in French with a comprehensive String Catalog covering all screens
  • In-App Language Switching — Choose your preferred app language (French, English, or system default) directly from Settings
  • Quick Language Toggle — Switch between French and English with a single click from the dashboard sidebar
  • Limitless Mode — A new AI mode inspired by NZT for enhanced cognitive pattern recognition, anticipation, and deep knowledge synthesis
  • Screen Recording Protection (Windows) — The overlay widget is now excluded from screen captures and recordings when content protection is enabled on Windows
  • Bulk Session Deletion (Windows) — Select and delete multiple sessions at once from the Sessions view in the Windows app
  • Electron macOS App — A new macOS version is now available as a universal Electron DMG on the download page
  • Live Transcript in Overlay (macOS) — See your meeting transcript in real time directly in the overlay widget — recent spoken text scrolls automatically as you speak
  • Save Transcript to File (Windows) — Export your session transcript as a text file directly from the Windows app
  • Instant Update Installation (Windows) — A Restart & Update button now appears in the overlay as soon as a new version is downloaded, so you can update without leaving your session

Improvements

  • Refined AI Coaching Prompts — Rewritten system prompts across all AI modes for more natural, concise, and actionable coaching responses
  • Upgraded to Claude Sonnet 4.6 — All AI modes now use Anthropic's latest Sonnet 4.6 model for faster, higher-quality responses
  • Adaptive Smart Mode — Smart Mode now uses adaptive thinking — the AI decides when deep reasoning is needed, improving speed on simple requests
  • Comprehensive Meeting Recaps — Recaps now capture up to 4x more conversation context for more complete and detailed meeting summaries
  • Consistent Language Detection — AI responses now match the transcript language from the very first word, eliminating mixed-language outputs
  • Sign-in Method Badge — Account settings now display whether you signed in with Google or Email
  • App Icons Update — Updated app icons and asset configurations for a refreshed visual identity
  • Auto-Move to Applications — The app now detects when it's running from Downloads or a disk image and offers to move itself to Applications for seamless updates
  • Faster App Startup — Keychain credentials, audio hardware, and update checker now initialize in background, eliminating startup freezes
  • Enhanced Error Diagnostics — Improved error handling and diagnostic logging across authentication, networking, and session management
  • Adaptive Response Depth — AI responses now dynamically adjust their depth and detail based on conversation complexity and context
  • Streamlined Assist Responses — Assist responses during live sessions now use a concise bullet-point format for quick scanning
  • Smarter Transcript Context — AI now prioritizes recent conversation over older background, ensuring responses reference what's being discussed right now
  • Accurate Speaker References — AI no longer incorrectly attributes statements to named individuals — responses now use generic references unless attribution is certain
  • Date-Aware AI Responses — AI now knows the current date, enabling accurate temporal reasoning for scheduling and deadline discussions
  • Accurate Plan Features — All three subscription plans now accurately reflect the features available in the apps, including screenshot capture for Pro and proactive suggestions for Enterprise
  • Built-in Mode Descriptions (Windows) — Built-in mode prompts are now shown as clear capability descriptions instead of raw system prompts
  • Plan Badge Design (Windows) — Pro and Enterprise plan badges now use gold/amber colors to visually distinguish paid tiers, matching the macOS design
  • Active Mode on Launch (Windows) — The Default mode is now auto-selected and shown as active when the Windows app first loads
  • Smart Mode Context (Windows) — Smart Mode now displays an explanatory tooltip clarifying its slower-but-more-accurate advanced reasoning behavior
  • FREE Plan Upgrade Prompts (Windows) — Clear upgrade prompts with pricing comparison when FREE plan limits are reached, including dashboard banner and overlay counter
  • Honest, Accurate AI Coaching — Built-in modes now prioritize factually correct answers over confirmation — if you're wrong, the AI says so clearly and explains why
  • Built-in Mode Descriptions (macOS) — Built-in mode prompts on macOS are now shown as clear capability descriptions instead of raw system prompts
  • AI Processing Indicator — A visual indicator now shows when the AI is actively generating a response
  • Proactive Pattern Recognition — AI modes now automatically identify psychological and organizational patterns in conversations and connect them to concrete actions — without waiting for you to introduce the concept
  • Auto-Detect Transcription Language — The transcription engine now automatically detects your spoken language — no manual language selection required
  • Cleaner Transcript Display (Windows) — The overlay transcript panel now shows only the most recent lines for a less cluttered, easier-to-read experience

Bug Fixes

  • Windows Auto-Update Fix — Fixed an issue where the Windows app always reported 'up to date' even when newer versions were available
  • Privacy Protection Messaging — Clear, user-friendly explanations when AI safety filters activate instead of generic refusal messages
  • Recap Reliability Fix — Recaps no longer send screenshots, preventing safety filter false positives during video calls
  • Reliable App Startup — The app now waits up to 10 seconds for authentication instead of 2, preventing false 'Session Expired' on slow networks
  • Secure Token Persistence — Token storage uses a safer update-or-add pattern, preventing accidental token loss after app restarts
  • Magic Link Session Fix — Opening the web dashboard from the app now correctly signs into the right account, even if another user was previously logged in
  • Onboarding Button Visibility — Navigation buttons are now always visible during the onboarding flow without needing to scroll
  • Stale Session Auto-Recovery — The app now detects and clears expired authentication states instead of silently failing
  • Overlay Mode Selection Fix — Fixed duplicate selection highlight when custom and built-in modes share the same name
  • Built-in Mode Defaults Fix — Fixed built-in mode default flags and filtering to correctly separate built-in and custom modes
  • Single Instance Protection — Prevents duplicate app instances when switching the UI language between French and English
  • Automatic Server Error Recovery — Server errors and token refresh failures now retry automatically before failing, reducing forced logouts
  • Screen Capture Auto-Recovery — Automatically recovers screen capture when an external display is disconnected or reconfigured
  • Screen Capture Refresh Fix (Windows) — Fixed a deduplication issue preventing fresh screenshots from being analyzed in Screen-Only AI mode
  • Manual AI Request Fix (Windows) — Manual AI requests are no longer blocked by the auto-response cooldown timer
  • Windows Auto-Updater Reliability — Resolved an issue where the Windows updater sometimes failed to detect available updates due to incorrect file path handling after code signing
  • Reduced Error Noise (Windows) — Expected transient errors like WebSocket reconnections and auth validation are no longer reported as crashes, improving issue visibility
  • Auto-Updater Path Fix (Windows) — Fixed an issue where the auto-updater could fail with a file not found error due to incorrect URL-to-path conversion on Windows
  • AI Language Matching Fix (Windows) — AI responses now correctly match the transcript language — previously, French examples in system prompts biased all responses toward French regardless of the meeting language
  • Taskbar Icon Visibility (Windows) — The Queen Mama icon now stays visible in the Windows taskbar when the dashboard is closed, making it easy to see the app is running

Under the Hood

  • Streamlined Sign-In Flow — Simplified the email authentication UI with clearer navigation between sign-in and sign-up
  • Regression Test Suite — Added comprehensive tests for authentication error handling, session lifecycle, and service availability checks
  • Windows CI/CD Pipeline — Automated build, test, and release pipeline for the Windows app with platform-specific versioning
  • Windows Build Environment Fix — Google Sign-In and environment detection now work correctly in production Windows builds
  • Secure Transcription Relay — WebSocket proxy that keeps Deepgram API keys on the server, improving security for desktop app audio streaming
  • Desktop App Sync Support — CORS and session synchronization infrastructure enabling desktop apps to sync data with the web dashboard
  • CI/CD Pipeline Optimization — Eliminated duplicate test runs and added concurrency controls across all build and deploy workflows

February 2026

Features

  • Voice Dictation Mode — Standalone voice dictation for quick AI queries without starting a full session
  • Meeting Detection & Reminders — Automatic detection of active meetings with smart reminders to start recording
  • Contact Management — Full contact management with notes, sync, and relationship tracking in Memory Palace
  • Bulk Session Deletion — Select and delete multiple sessions at once from the dashboard
  • Memory Palace — AI-powered mini-CRM that automatically extracts and manages contacts from your conversations
  • Display Selection — Choose which display to use for screen capture in multi-monitor setups
  • Internationalized Landing Site — Full French and English support across the entire website with locale-based routing
  • Custom Modes in Overlay — Switch between custom AI modes directly from the overlay widget popup menu
  • Advanced Markdown Rendering — Full Markdown support in overlay responses including code blocks, bullet lists, and ordered lists

Improvements

  • Improved Transcription Reliability — Enhanced WebSocket reconnection logic and more robust error recovery for uninterrupted transcription
  • Refined Overlay Appearance — More transparent and polished overlay widget for a less intrusive experience
  • Enhanced AI Response Quality — Improved prompt engineering and context handling for clearer, more actionable AI suggestions
  • Proactive Health Monitoring — Background service health checks to detect and recover from issues before they impact your session
  • Polished Overlay Widget — Refined tab buttons, status indicators, and popup menu for a smoother overlay experience
  • Smarter AI Suggestions — Improved prompt handling and screenshot integration for more relevant AI responses
  • Energy Optimization — Reduced energy consumption by optimizing animations to only run when needed
  • Enhanced AI Streaming — Improved AI response streaming with better Markdown table support via remark-gfm
  • Refined Authentication Flow — Streamlined sign-in experience with improved session management and smoother transitions
  • Polished Dashboard Animations — Smoother transitions and refined visual feedback throughout the dashboard interface

Bug Fixes

  • WebSocket Connection Recovery — Fixed infinite reconnection loop that could leave transcription stuck on 'Reconnecting' during WiFi disruptions
  • Authentication Persistence Fix — Resolved an issue where authentication state could be lost between app restarts
  • Web Dashboard URL Fix — Fixed the Web Dashboard button redirecting to the wrong domain

Under the Hood

  • Schema Versioning — Versioned data migration system for safe and seamless SwiftData schema updates
  • Orphaned Session Cleanup — Automatic detection and cleanup of incomplete sessions left behind after crashes
  • Deployment Pipeline — Separate staging and production environments with test-gated deployments

January 2026

Features

  • 4-Tier Subscription System — Introduced FREE, PRO, and ENTERPRISE plans with tailored features for every use case
  • Advanced License Management — Feature gating and license status badges throughout the application
  • Device Code Authentication — Seamless authentication flow with automatic session synchronization between macOS app and web dashboard
  • Admin Dashboard — Comprehensive user management, role assignment, and usage statistics for administrators
  • Server-Side Usage Tracking — Real-time monitoring of AI requests and feature usage for better insights

Improvements

  • Enhanced Stripe Integration — Dynamic checkout supporting multiple subscription tiers with automated plan assignment
  • Webhook Automation — Automatic subscription plan association and license updates via Stripe webhooks

December 2025

Features

  • License Status Badges — Visual indicators in the dashboard showing your current subscription tier
  • Markdown Support — Rich text rendering for AI responses with formatting, lists, and code blocks

Improvements

  • Authentication Error Handling — Improved error messages and recovery flows for authentication issues
  • Stripe Integration Refinements — More reliable payment processing and subscription management

November 2025

Features

  • Multi-Provider AI Support — Choose from OpenAI GPT-4o, Anthropic Claude Sonnet 4, Google Gemini 2.0, and more
  • Smart Mode — Automatic AI model selection based on context and task complexity
  • Intelligent Caching System — 30-50% cost reduction through smart response caching and optimization
  • Token Usage Tracking — Real-time monitoring of API usage with cost estimates

Under the Hood

  • Screenshot Deduplication — Optimized visual context handling to reduce redundant API calls
  • Dynamic Token Limits — Adaptive token management for cost-efficient AI interactions

October 2025

Features

  • Multi-Language Transcription — Automatic language detection supporting French, English, and more
  • Multiple Transcription Providers — Support for Deepgram and AssemblyAI with automatic fallback
  • Response History — View and navigate through AI responses directly in the overlay

Improvements

  • Enhanced Visual Context — Improved screenshot integration in AI prompts for better understanding
  • Auto-Scroll History — Automatic scrolling to latest responses in the overlay

September 2025

Features

  • Real-Time Audio Transcription — Live speech-to-text powered by Deepgram Nova-3
  • Contextual AI Assistance — Intelligent suggestions based on conversation context and screen content
  • Screen Context Analysis — AI understands what's on your screen to provide relevant advice
  • Undetectable Overlay — Privacy-focused overlay that stays invisible to screen sharing and recording tools
  • Secure API Key Storage — Military-grade encryption using macOS Keychain for your API keys
  • Multiple AI Modes — Pre-configured modes for Professional meetings, Interviews, Sales calls, and more