docs: fold the never-published 2.8.0-2.8.6 work into the 2.8.7 notes

v2.7.0 is the last release on GitHub, so 2.8.7 is the next official
update and everything in between belongs in one note — including the
2.8.0-2.8.6 builds that were tagged but never published.

Added across the release note and README: Hindsight long-term memory
(Pro, opt-in), signed macOS builds for both architectures, the macOS 12
compatibility guard for local speech, and the move of live transcription
to low-latency regional hosts.

Deliberately NOT re-added as headline features, despite appearing in the
old v2.8.0 copy:
- Sandboxed Code Verification — verificationEnabled.ts documents itself
  as "TEMPORARILY DISABLED across the app"; default OFF, explicit opt-in
  only, and the orchestrator early-returns when it is off.
- Spoken Answer Humanizer — gated on `answerDiversityGuard`, which is
  `default: false` in intelligenceFlags.ts.
Announcing either as shipped in 2.8.7 would be untrue for every user who
does not set an env var.

AboutSection's What's New is now the five most useful across the whole
2.7.0 -> 2.8.7 range: Auto Answer, the new answer engine, rebuilt meeting
notes, the Chrome companion, and Hindsight LTM. NVIDIA speech and the two
new modes drop to the release note only.

Validation: `npm run typecheck:ts7` clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JJS2xjF8uDPMa4EpXey5FL

Author: Hardening v2.7

README.md

@@ -418,18 +418,20 @@ We've launched the official **$NAT token** on Printr! Holders who maintain a spe
 
 ### What's New in v2.8.7 (Latest Release)
 
-Three months of work since v2.7.0 — a smarter answer engine, hands-free Auto Answer, rebuilt meeting notes, a Chrome companion extension, and NVIDIA speech.
+Everything since v2.7.0 — a smarter answer engine, hands-free Auto Answer, long-term memory, rebuilt meeting notes, a Chrome companion extension, and NVIDIA speech. The 2.8.0–2.8.6 builds were never published here, so their work ships in this release too.
 
 - **Auto Answer (Beta)**: Answers appear on their own the moment the other person finishes a question. Natively judges whether an answer is actually wanted and stays silent when it isn't — no card to dismiss, no button to press. Off by default; enable it in **Settings → General**.
 - **A New Answer Engine**: Answers now pick their source properly — your résumé, an uploaded document, the live conversation, or general knowledge — instead of blending them together. Follow-up questions keep the thread even after the topic changes.
 - **Meeting Notes, Rebuilt**: Summaries adapt to the kind of meeting you're in, get a real title instead of the first line of the transcript, and scale in length with the meeting. Plus editable speaker labels and a ready-to-send follow-up email draft.
 - **Browser Companion for Chrome**: Press **⌘/Ctrl+Y** to pull the page you're looking at into your next answer — a job description, a coding problem, documentation. One click enables it everywhere.
 - **Two New Built-in Modes**: **Seminar** for strict file-grounded Q&A (thesis defenses, paper walkthroughs), and **Call Center** for support calls framed as issue → resolution → escalation.
 - **NVIDIA NIM & NVIDIA Speech**: A new provider for both text and real-time transcription, with nine streaming models to choose from.
+- **Hindsight Long-Term Memory (Pro)**: A local vector database that indexes your past meetings, profiles and documents, so Natively can recall what was said weeks ago instead of only what's in the current session. Opt-in from Settings.
 - **Role Insight (Pro)**: Paste a job description and see every requirement pulled out, checked against your profile, and scored for coverage.
 - **Windows Stealth Typing**: Type into the meeting overlay without your call ever losing focus. Handles Alt+Tab, clicking away, and steps aside for CJK/IME input.
 - **Skills Upload & Delete**: Add your own skill as a file or folder with a preview before it installs, and remove ones you no longer want — no more editing folders by hand.
 - **Redesigned Meeting Overlay**: Panel, pill and resize control now move as one, answers stream in word by word, and scrolling back no longer fights you.
+- **Signed macOS Builds & Faster Transcription**: Signed builds for Apple Silicon and Intel, a macOS 12 (Monterey) compatibility guard for local speech, and live transcription moved to low-latency regional hosts.
 
 > Read the [full v2.8.7 release notes](docs/releases/v2.8.7.md) for every change, including fixes.
 

docs/releases/v2.8.7.md

@@ -1,6 +1,6 @@
 ## Summary
 
-Three months of work since v2.7.0 — a smarter answer engine, Auto Answer for meetings, rebuilt meeting notes, a Chrome companion extension, NVIDIA speech, and Windows stealth typing.
+Everything since v2.7.0 — a smarter answer engine, Auto Answer for meetings, long-term memory, rebuilt meeting notes, a Chrome companion extension, NVIDIA speech, and Windows stealth typing. The 2.8.0–2.8.6 builds were never published here, so their work is included below.
 
 ## What's New
 
@@ -11,16 +11,19 @@ Three months of work since v2.7.0 — a smarter answer engine, Auto Answer for m
 - **Built-in Modes**, including a new **Call Center** mode.
 - **Skills: upload and delete.** Add your own skill as a file or folder with a preview before it installs, and remove ones you no longer want.
 - **NVIDIA NIM** joins the provider list, with **NVIDIA speech-to-text** and nine streaming models to choose from.
+- **Hindsight Long-Term Memory (Pro)** — a local vector database that indexes your past meetings, profiles and documents, so Natively can recall what was said weeks ago instead of only what's in the current session. Opt-in from Settings.
 - **Role Insight (Pro)** — paste a job description and see every requirement pulled out, checked against your profile, and scored for coverage.
 - **Windows: type into the meeting overlay without losing focus** on your call. Works with Alt+Tab and clicking away, and steps aside for CJK/IME input.
 - **Redesigned meeting overlay** — the panel, pill and resize control now move as one, answers stream in word by word, and scrolling back no longer fights you (with a jump-to-latest button).
 - **Parakeet CTC 0.6B** added to the local speech models.
 - **LiteLLM** gets a model picker and a configurable default.
+- **Signed macOS builds** for both Apple Silicon and Intel, and a compatibility guard that stops local speech crashing on macOS 12 (Monterey).
 
 ## Improvements
 
 - New default models: **Gemini 3.1 Flash Lite** and **Gemini 3.7 Flash**. The full Claude catalogue is now listed and correctly ordered, and the model picker shows real provider logos.
 - Auto Answer and speech recognition are noticeably faster — answers are prepared ahead of time, and local speech no longer stalls when a meeting starts.
+- Live transcription moved to low-latency regional hosts, cutting round-trip time and ending the double-billing that could happen on a dropped connection.
 - Refreshed look throughout: Apple-style toggles, smoother section transitions, a new Soft Orchid accent, and a unified Plans & Billing page.
 - Code blocks redesigned — clearer syntax colours, and the language label and copy button combined into one control.
 - Calmer onboarding: guided steps instead of a pile of notifications.

src/components/AboutSection.tsx

@@ -128,16 +128,16 @@ export const AboutSection: React.FC<AboutSectionProps> = () => {
                         </div>
                     </div>
 
-                    {/* 5. NVIDIA Speech & New Modes */}
+                    {/* 5. Hindsight Long-Term Memory */}
                     <div className="p-3 bg-bg-card/50">
                         <div className="flex items-start gap-4">
                             <div className="w-10 h-10 rounded-lg bg-emerald-500/10 flex items-center justify-center text-emerald-400 shrink-0">
-                                <Volume2 size={20} />
+                                <Database size={20} />
                             </div>
                             <div>
-                                <h5 className="text-sm font-bold text-text-primary mb-1">NVIDIA Speech & Two New Modes</h5>
+                                <h5 className="text-sm font-bold text-text-primary mb-1">Hindsight Long-Term Memory</h5>
                                 <p className="text-xs text-text-secondary leading-relaxed">
-                                    NVIDIA NIM joins the provider list for both text and real-time transcription, with nine streaming models. Plus Seminar mode for strict file-grounded Q&A and Call Center mode for support calls.
+                                    A local vector database that indexes your past meetings, profiles and documents, so Natively can recall what was said weeks ago — not just in this session. Pro, and opt-in from Settings.
                                 </p>
                             </div>
                         </div>