--- created: 2026-03-12 source: Rivet tags: [agent-archive, rivet] --- # Voice Note Guidelines — Michael's Daily Briefs *Based on feedback from 2026-02-16: Voice during work hours (6am-6pm), multiple short notes under 1400 chars, current data only* --- ## The Voice-First Constraint ### Why Voice Notes Matter - **Michael can't read text while on site or driving** (safety issue) - **Single play-through requirement** — must be clear on first listen - **Time pressure** — gets briefs while preparing for site work - **Context switching** — from personal to business mindset quickly - **Safety critical** — bad info leads to wrong decisions ### Character vs Time Conversion - **1400 characters ≈ 90 seconds** at normal speech speed - **1200 characters ≈ 75 seconds** (sweet spot for attention) - **Edge TTS timing:** ~15-16 characters per second - **Attention span:** 60-90 seconds per note is maximum --- ## Multi-Note Structure — Morning Brief ### The 4-Note Formula **Why 4 notes?** Breaks complex business state into digestible chunks while driving. #### Note 1: IMMEDIATE STATUS (Critical Systems) **Purpose:** Can we work today? Any fires to fight? **Content:** - App status (LIVE/DOWN) - Growth Engine status (LIVE/DOWN) - Builder status (ACTIVE/IDLE/BLOCKED) - Weather for site work - Critical overnight alerts **Tone:** Authoritative, factual, reassuring **Hook:** "Good morning Michael, it's 5:15 AM..." **End:** "This is note 1 of 4." #### Note 2: BUSINESS STATE (Pipeline & Decisions) **Purpose:** What's the business doing? What needs my input? **Content:** - Pipeline metrics (leads, conversions, hot prospects) - Overnight business activity - Decisions waiting for Michael's approval - Revenue/financial updates **Tone:** Business-focused, specific numbers **Transition:** Continue energy from Note 1 **End:** "This is note 2 of 4." #### Note 3: TODAY'S EXECUTION (Plans & Priorities) **Purpose:** What's the plan? Where should I focus? **Content:** - Michael's schedule (site, crew, timing) - Top 3 priorities ranked by impact - Evening window plan (QA, coding, meetings) - Blockers that could derail the day **Tone:** Forward-looking, actionable **Transition:** "Here's today's plan..." **End:** "This is note 3 of 4." #### Note 4: STRATEGIC CONTEXT (Big Picture) **Purpose:** How does today fit the larger goal? **Content:** - Week's primary goal - Progress since yesterday - Upcoming events/decisions (next 48h) - Key metric to track **Tone:** Strategic, confident, motivating **Transition:** "Strategic update..." **End:** "Morning brief complete. Drive safe, talk tonight." --- ## Multi-Note Structure — Evening Brief ### The 3-Note Handoff **Why 3 notes?** End-of-day summary, overnight delegation, tomorrow prep. #### Note 1: TODAY'S RESULTS (What Actually Happened) **Purpose:** Close the loop on today's plan **Content:** - What shipped/deployed (specific) - Business progress (leads, conversations, deals) - Plan vs reality (what hit timeline, what didn't) - Critical issues discovered **Tone:** Results-focused, honest assessment **Hook:** "Evening handoff! Here's what got done today..." **End:** "This is note 1 of 3." #### Note 2: OVERNIGHT DELEGATION (Autonomous Work Plan) **Purpose:** What happens while Michael sleeps **Content:** - Builder's specific tasks - Rivet's research/monitoring priorities - Susan's lead gen/content work (when live) - Alert thresholds (when to wake Michael) **Tone:** Delegation-focused, confidence in team **Transition:** "Overnight autonomous work..." **End:** "This is note 2 of 3." #### Note 3: TOMORROW'S PRIORITIES (First Thing Focus) **Purpose:** Hit the ground running tomorrow morning **Content:** - Critical decisions needed at 6am - QA testing priorities - External communications to make - Strategic advancement opportunities **Tone:** Future-focused, prioritized **Transition:** "Tomorrow priorities..." **End:** "Handoff complete. Team's got the overnight shift." --- ## Language Optimization for Voice ### Spoken English vs Written English #### ✅ Voice-Optimized Language ``` "Builder finished three bugs overnight. The profile sync is working, browse workers went live, and signup flow captures names properly. That's all sixteen QA items complete." ``` #### ❌ Text-Converted Language ``` "Builder completed Bug #3 (profile synchronization), Bug #10 (browse workers functionality), and Bug #11 (signup flow name capture). QA completion status: 16/16 items achieved." ``` ### Voice-Friendly Patterns #### Numbers and Quantities - **Say:** "forty-two leads" not "42 leads" - **Say:** "three bugs fixed" not "3 bugs fixed" - **Say:** "sixteen of sixteen" not "16/16" - **Exception:** Dollar amounts "twelve thousand dollars" or "$12,000" #### Status and States - **Good:** "App is live and stable" - **Bad:** "Application operational status: nominal" - **Good:** "Builder's working on PWA icons" - **Bad:** "Builder agent executing PWA icon task" #### Transitions and Flow - **Between topics:** "Next," "Also," "Important," "Finally" - **Between notes:** "This is note 2 of 4" - **Time references:** "overnight," "this morning," "right now" - **Emphasis:** "Critical," "Urgent," "Complete," "Ready" #### Natural Speech Rhythms - **Use contractions:** "It's live," "We're ready," "Here's the plan" - **Use pauses:** Commas and periods for natural breaks - **Vary sentence length:** Mix short punchy statements with longer explanations - **Conversational tone:** Like talking to Michael in person --- ## Content Selection — What to Include vs Exclude ### ✅ ALWAYS Include (High Priority) #### System Status - App LIVE/DOWN with specific URL - Growth Engine status with any issues - Builder session state and current work - Any critical alerts or outages #### Business Metrics - Total leads with change from yesterday (+/-X) - Hot leads under 48 hours (names if <5) - Overdue callbacks (names and timeframes) - Revenue pipeline value and conversions #### Immediate Actions - Decisions waiting for Michael's approval - Blockers that stop progress today - Critical tasks with completion times - External dependencies (calls to make, approvals needed) #### Context for Safety - Weather conditions for site work - Schedule conflicts or timing issues - Equipment/crew status affecting site work - Any safety-critical information ### 🟡 Include If Space Allows (Medium Priority) #### Progress Updates - Features shipped with git commit references - Research completed with outcomes - Agent spawns and completion status - Documentation updates or process improvements #### Strategic Information - Weekly goal progress assessment - Competitive intelligence discoveries - Market opportunities or threats - Long-term project status updates #### Team Coordination - Cross-agent handoffs completed - Communication protocol successes/failures - Resource allocation changes - Process optimization results ### ❌ NEVER Include (Low Priority/Noise) #### Technical Internals - API endpoint details or response codes - Database query optimization - Server resource usage (unless critical) - Code architecture discussions #### Historical Context - "As we discussed last week..." - "Following up on yesterday's conversation..." - "Per our previous decision..." - Old decisions unless they're changing #### Vague Status Updates - "Things are going well" - "Progress continues" - "As expected" - "No significant changes" #### Administrative Overhead - Cron job execution logs - File system housekeeping - Agent session management details - Token usage statistics (unless over budget) --- ## Timing and Delivery Optimization ### Character Count Discipline - **Target:** 1200-1400 characters per note - **Minimum:** 1000 characters (too short = not comprehensive) - **Maximum:** 1450 characters (too long = attention loss) - **Buffer:** Allow 1380 characters as max to ensure <1400 ### Speech Timing Estimates - **Normal speed:** ~15 characters per second - **1200 chars:** ~80 seconds - **1400 chars:** ~93 seconds - **Optimal range:** 75-90 seconds per note ### Delivery Sequence - **Gap between notes:** 5-10 seconds for processing - **Total morning brief:** ~6-7 minutes (4 notes + gaps) - **Total evening brief:** ~4-5 minutes (3 notes + gaps) - **Michael's available time:** Usually 8-10 minutes max ### Voice Settings (Edge TTS) - **Voice:** en-AU-WilliamNeural (Australian male, professional) - **Speed:** Normal (not 1.2x - that's for phone calls) - **Tone:** Conversational but authoritative - **Emphasis:** Let content drive emphasis, not artificial vocal stress --- ## Quality Checklist — Before Sending ### Content Quality - [ ] All data timestamped within last 30 minutes - [ ] No stale references to completed tasks - [ ] Specific numbers, not vague estimates - [ ] Names included for people and projects - [ ] Current status verified via API/file checks ### Voice Optimization - [ ] Natural speech patterns (contractions, conversational tone) - [ ] Numbers spelled out ("three" not "3") - [ ] Clear transitions between topics - [ ] Note numbering included ("This is note 2 of 4") - [ ] Appropriate ending for each sequence ### Structure Compliance - [ ] Each note 1200-1400 characters - [ ] Content fits designated purpose (status/plans/results) - [ ] Logical flow between notes - [ ] No duplication across notes - [ ] Critical information in early notes ### Safety & Usability - [ ] Clear on single play-through - [ ] No complex explanations requiring re-listen - [ ] Action items clearly identified - [ ] Urgent vs non-urgent distinction clear - [ ] Safe for driving attention levels ### Freshness Verification - [ ] Builder status from latest RIVET-INBOX.md - [ ] System status from live API checks - [ ] Pipeline data from Growth Engine API - [ ] Weather from current API call - [ ] Task completion cross-verified with git/files --- ## Error Handling — When Data Is Stale ### Stale Data Detection - **>30 minutes old:** Flag with "last confirmed at TIME" - **API failures:** Use "checking" or "verifying" language - **Inconsistent sources:** Acknowledge uncertainty - **Builder claims unverified:** Note "pending git confirmation" ### Stale Data Language Patterns #### ✅ Good Stale Data Handling ``` "Builder reported finishing PWA icons as of 8:40 AM - checking git for confirmation. Pipeline shows forty-two leads, but that's from last night's scan, so current count may be higher." ``` #### ❌ Bad Stale Data Handling ``` "Builder completed PWA icons. Pipeline has forty-two leads." (Presents uncertain data as fact) ``` ### Uncertainty Qualifiers - **Time-based:** "as of TIME," "last confirmed TIME," "reported at TIME" - **Source-based:** "according to Builder," "pipeline indicates," "estimated" - **Verification-based:** "checking," "verifying," "pending confirmation" - **Confidence-based:** "likely," "probably," "appears to be" ### When in Doubt - **Better to be accurate than confident** - **Flag uncertainty rather than guess** - **Provide context for why data might be stale** - **Promise follow-up when fresh data available** --- ## Examples — Full Voice Brief Sequences ### Example Morning Brief (Good) #### Note 1: System Status ``` "Good morning Michael! It's 5:15 AM. App is live and stable at rivet dot rateright dot com dot au, Growth Engine came back online at 4:30 after the Railway issue. Builder's actively working on PWA icons and profile auto-research, expects to finish in two to three hours. Weather's seventeen degrees and clear - perfect conditions for concrete work. No critical alerts overnight. This is note 1 of 4." (Characters: 487) ``` #### Note 2: Business State ``` "Pipeline update: we've got forty-two leads total, eight hot leads under forty-eight hours, three overdue callbacks - Ardi, Steve, and the new Parramatta contractor. Growth Engine processed four fresh leads overnight from Indeed and Gumtree. Pipeline value sitting at twelve thousand dollars if everyone converts. Two decisions waiting for you - approve Susan's agent setup and review the launch readiness checklist. This is note 2 of 4." (Characters: 476) ``` #### Note 3: Today's Plan ``` "Today's priorities: Builder finishing PWA work and profile research should unlock launch readiness testing. Your site schedule has you at Homebush with the concrete crew, expected wrap by 4 PM. Evening window focus: QA test the new PWA features and review Susan's setup for deployment. Main blocker to watch: if Builder hits technical issues with the profile auto-research, that could delay launch testing until tomorrow. This is note 3 of 4." (Characters: 481) ``` #### Note 4: Strategic Context ``` "Strategic update: this week's goal is completing launch preparation, and we're on track. Yesterday's milestone - all sixteen QA bugs complete - unlocked the launch phase. Key metric today: test PWA features work smoothly across different devices. Looking ahead, tomorrow's priority is the launch decision meeting if testing goes well. Next forty-eight hours could see us moving from launch-ready to actually launched. Morning brief complete, drive safe, talk tonight." (Characters: 535) ``` ### Example Evening Brief (Good) #### Note 1: Today's Results ``` "Evening handoff! Here's what shipped today. Builder deployed PWA icons and profile auto-research - both live in production, commit hash 7f3a9e2. Pipeline processed six leads, advanced two warm conversations with contractors in Bankstown and Penrith. No new conversions but Ardi finally responded and wants to meet tomorrow. Plan versus reality: PWA work came in ahead of schedule, profile research took longer but it's solid. No critical issues to flag. This is note 1 of 3." (Characters: 543) ``` #### Note 2: Overnight Work ``` "Overnight autonomous work plan: Builder's tackling the launch readiness checklist - testing PWA offline functionality, verifying payment flows work end-to-end, and running the full user journey audit. Rivet's monitoring all systems, processing any new leads from the overnight job board scan, and preparing tomorrow's brief with fresh data. Alert threshold: wake you if the app goes down for more than five minutes or Builder gets blocked for more than two hours. Confidence level on overnight work: high. This is note 2 of 3." (Characters: 567) ``` #### Note 3: Tomorrow's Focus ``` "Tomorrow priorities: first thing, review Builder's launch readiness results and make the go/no-go decision. QA testing focus: PWA works offline, payment flow handles edge cases, user onboarding is smooth. External communications: call Ardi to set up that contractor meeting, and contact Steve about his overdue callback. Strategic opportunity: if launch testing passes, we could announce soft launch to existing contacts. Handoff complete, team's got the overnight shift." (Characters: 535) ``` --- ## Continuous Improvement ### Feedback Integration Process 1. **Monitor Michael's responses** to voice briefs 2. **Track questions** that indicate missing information 3. **Note timing issues** (too long, too short, too fast) 4. **Adjust language patterns** based on comprehension 5. **Update character limits** based on speech timing ### Monthly Review Checklist - [ ] Voice brief completion rates (does Michael listen to all notes?) - [ ] Data freshness failures (how often is information stale?) - [ ] Question patterns (what information is consistently missing?) - [ ] Timing optimization (are notes the right length?) - [ ] Language effectiveness (is the tone working?) ### Success Metrics - **Comprehension:** Michael asks fewer clarifying questions - **Action:** Decisions get made faster based on brief content - **Safety:** No accidents or issues due to distracted listening - **Efficiency:** Brief covers all necessary information without waste - **Trust:** Michael relies on brief accuracy for daily decisions *This format solves Michael's core feedback: voice-first design, current data, optimal length, safety-conscious structure.*