neotoma 0.3.0 → 0.3.2

package/docs/releases/archived/aspirational/v1.0.0/value_discovery_plan.md

@@ -1,279 +0,0 @@
-## Release v1.0.0 — Value Discovery Plan
-### Purpose
-This document defines the value discovery activities for v1.0.0, focused on validating that AI-Native Individual Operators need a way to give AI tools (Claude/ChatGPT) access to their personal data context via MCP.
-**Related Documents:**
-- `discovery_plan.md` — Discovery overview and coordination
-- `usability_discovery_plan.md` — Usability discovery activities
-- `business_viability_discovery_plan.md` — Business viability discovery activities
-- `participant_recruitment_log.md` — Participant tracking
-### 1. Hypothesis and Assumptions
-**Hypothesis:** "AI-Native Individual Operators need a way to give AI tools (Claude/ChatGPT) access to their personal data context via MCP"
-**Assumptions:**
-- Users want to query their personal data via AI interfaces (not just search documents)
-- Current workarounds (copy-paste, manual preparation) are painful
-- Users value deterministic, structured data access for AI over semantic search
-- MCP (Model Context Protocol) is the right interface for AI-native workflows
-- Problem exists: AI tools lack access to personal data context, forcing manual workarounds
-### 2. Discovery Method
-**Approach**: Hybrid (async screening survey + live interviews + optional async validation)
-**Why Hybrid:**
-- Async screening survey: Quickly identify promising participants from large pool (40-50 responses)
-- Live interviews: Deep discovery using Mom Test to validate problem and solution (high-fidelity)
-- Optional async validation: Validate findings at scale after hypothesis formed
-### 3. Async Screening Survey
-#### 3.1 Purpose and Timing
-- **Purpose**: Identify promising participants from large pool quickly
-- **Timing**: Week -8 (before interviews start)
-- **Target**: 40-50 responses, 20-25 willing to interview
-- **Selection**: Choose 17-18 to schedule (accounting for no-shows)
-#### 3.1a When Survey is Used vs. Direct Outreach
-**Survey Path (Community Posts, Broad Outreach):**
-- Indie Hackers community posts
-- Reddit/Discord community posts
-- Hacker News posts
-- Twitter/X public posts (not DMs)
-- **Workflow**: Post → Survey link → Responses → Select → Schedule interviews
-**Direct Outreach Path (Personalized Outreach):**
-- LinkedIn personalized messages
-- Twitter/X DMs (personalized)
-- Email (personalized)
-- Existing network
-- **Workflow**: Personalized message → Calendar link → Schedule interview directly
-**Decision Criteria:**
-- **Use Survey**: Broad community posts, public outreach, when you can't personalize
-- **Use Direct Outreach**: When you can personalize, have specific lead, or from existing network
-**Important**: For direct outreach, subscription detection relies on:
-1. Bio/post mentions (detected by lead sourcing tools)
-2. Proxy indicators (usage frequency, feature mentions)
-3. Direct ask in interview (if unclear)
-For survey path, subscription detection uses self-reported survey data (most reliable).
-#### 3.2 Survey Questions
-**ICP Qualification:**
-- "How often do you use Claude or ChatGPT?" (Daily/Weekly/Monthly/Never)
-- "What type of work do you do?" (Open text)
-- "How long have you been using AI tools regularly?" (Months/years)
-**Enhanced ICP Qualification (For AI-Native Operators):**
-- "Which AI tools do you currently pay for?" (Multiple choice: ChatGPT Plus/Pro, Claude Pro, Cursor, Raycast, Other, None) — **Primary subscription detection method**
-- "How much do you spend per month on AI tools?" ($0/$10-20/$20-50/$50+)
-- "How often do you use Cursor or Raycast?" (Daily/Weekly/Monthly/Never)
-- "Do you participate in AI tool communities?" (Reddit, Discord, Twitter/X, None)
-- "Do you follow AI tool creators on social media?" (Yes/No)
-- "Have you set up any integrations with AI tools?" (MCP servers, API integrations, Other, None)
-**Note on Subscription Detection**:
-**For Survey Path (Community Posts):**
-- **Primary method**: Self-reported in screening survey (most reliable)
-- Survey responses provide subscription data before scheduling
-**For Direct Outreach Path (Personalized Messages):**
-- **Primary method**: Bio/post mentions detected by lead sourcing tools (high confidence if detected)
-- **Secondary method**: Proxy indicators (high usage frequency, feature mentions) — inferred but not definitive
-- **Fallback**: Ask directly in interview if subscription status unclear
-**Recommendation**: For direct outreach leads with unclear subscription status, include a brief screening question in calendar confirmation: "Quick question: Which AI tools do you currently pay for? (ChatGPT Plus/Pro, Claude Pro, Cursor, None) — helps us prepare for our chat."
-**Enhanced ICP Qualification (For Knowledge Workers):**
-- "What is your exact job title?" (Open text)
-- "What type of organization do you work for?" (Consulting firm, Research institution, Law firm, Corporate, Other)
-- "What tools do you use for research/document management?" (Zotero, Mendeley, Legal research platforms, Other, None)
-- "How many documents do you typically work with per week?" (1-10/11-50/51-100/100+)
-- "Do you need to synthesize information across multiple documents?" (Yes/No)
-- "How long are your typical research cycles?" (Days/Weeks/Months)
-**Past Behavior Facts:**
-- "In the last 30 days, how many times did you copy-paste a document into Claude/ChatGPT?" (0/1-5/6-10/10+)
-- "Do you currently have a process for giving AI tools access to your personal data? If yes, describe briefly." (Open text)
-- "What tools do you currently use to manage your documents/data?" (Multiple choice + other)
-**Commitment Signals:**
-- "How much time per week do you spend preparing data to give to AI tools?" (0 min/1-15 min/15-30 min/30-60 min/1+ hours)
-- "Have you paid for any tools that help you work with AI or access your data? If yes, what and how much?" (Open text)
-**Recruitment:**
-- "Would you be open to a 30-45 minute conversation about your AI workflows? We're not selling anything, just researching." (Yes/No)
-- "If yes, what's your email or preferred contact method?" (Open text)
-#### 3.3 Selection Criteria
-Prioritize participants who:
-**For AI-Native Operators:**
-- ✅ Paid AI tool subscription (ChatGPT Plus/Pro, Claude Pro, Cursor) — **High Priority**
-- ✅ Daily AI tool usage
-- ✅ Spend 15+ min/week preparing data
-- ✅ Community engagement (posts, follows AI creators)
-- ✅ Has integrations or mentions MCP
-**For Knowledge Workers:**
-- ✅ Specific job title match (Analyst, Researcher, Consultant, Lawyer, Strategist) — **High Priority**
-- ✅ Works at relevant company type (consulting, research, law firm)
-- ✅ Uses research/document management tools (Zotero, Mendeley, legal research)
-- ✅ Needs cross-document synthesis
-- ✅ Works with multiple documents per week (11+)
-**Minimum Threshold:**
-- Must meet at least 3 criteria from above
-- Must have at least one "high priority" signal (paid subscription OR specific job title)
-- Must have commitment signal (paid tools OR time spent preparing data)
-**See `discovery_filtering_criteria.md` for complete filtering criteria and scoring system.**
-#### 3.4 Success Criteria
-- 40-50 survey responses
-- 20-25 express willingness to interview
-- Select 17-18 to schedule (accounting for no-shows)
-### 4. Live Interviews
-#### 4.1 Purpose and Timing
-- **Purpose**: Deep discovery using Mom Test to validate problem and solution
-- **Timing**: Week -8 to Week -6
-- **Why Live Required**:
-  - Must probe past behavior: "Tell me about the last time..." requires follow-up questions
-  - Detect commitment signals: Time spent, money spent, emotional intensity need conversation to validate
-  - Avoid false positives: People are polite on surveys, need conversation to filter vague enthusiasm
-  - Dynamic question flow: Pivot based on responses, test solution only if problem validated
-#### 4.2 Participants
-- **AI-Native Individual Operators**: 8 participants
-  - Source: Selected from screening survey based on commitment signals
-- **High-Context Knowledge Workers**: 5 participants
-  - Source: Selected from screening survey based on commitment signals
-#### 4.3 Interview Structure
-**Duration**: 30-45 minutes per interview
-**Format**: Structured interview with recording permission
-**Sections:**
-##### Section 1: Opening (5 min)
-- Introduce yourself and Neotoma
-- Explain purpose: Understand how you use AI tools with your personal data
-- Request permission to record
-- **Important**: Do NOT mention solution yet. Focus on current AI workflow.
-##### Section 2: Problem Discovery (20 min) — Mom Test Questions Only
-**Goal**: Validate problem exists: AI tools lack access to personal data context, forcing painful workarounds
-**Questions:**
-1. "Tell me about the last time you asked Claude/ChatGPT about your personal data — like a contract, receipt, financial record, or travel booking. Walk me through exactly what happened step-by-step."
-   - Follow-ups:
-     - What question were you trying to answer?
-     - What data did you need the AI to know?
-     - How did you give the AI access to that data? (copy-paste? upload file? describe manually?)
-     - How long did it take to prepare the data?
-     - What was frustrating about the process?
-2. "Show me or walk me through the last time you wanted an AI to answer a question about your personal information but it couldn't because it didn't have access. What did you do?"
-   - Follow-ups:
-     - What question were you trying to ask?
-     - What did you do instead? (manually search, copy-paste, give up?)
-     - What was the impact of not being able to ask AI directly?
-     - How often does this happen?
-3. "Tell me about a specific time when you needed Claude/ChatGPT to understand something about your personal data — maybe a contract detail, financial transaction, or travel booking. What happened?"
-   - Follow-ups:
-     - What information did the AI need to know?
-     - How did you try to communicate it to the AI?
-     - Did it work? What was the problem?
-     - What did you do instead?
-4. "Walk me through how you currently give AI tools access to your personal data when you need them to answer questions about it."
-   - Follow-ups:
-     - Do you copy-paste documents into chat?
-     - Do you upload files? What format?
-     - Do you manually type out the information?
-     - What's the most frustrating part of this process?
-     - How much time does it take to prepare data before asking AI?
-5. "Have you tried any tools or solutions that let AI access your personal data automatically? What did you try?"
-   - Follow-ups:
-     - What tools or approaches did you experiment with?
-     - What happened when you tried them?
-     - Why did you stop using them? (or continue using them?)
-     - What was missing or frustrating?
-6. "Tell me about your experience with MCP (Model Context Protocol) servers, if any. Have you tried connecting Claude/ChatGPT to external data sources?"
-   - Follow-ups:
-     - Have you heard of MCP?
-     - Have you used any MCP servers? Which ones?
-     - What worked well? What didn't?
-     - What kind of data would you want to connect via MCP?
-**Commitment Signal Questions:**
-- How much time do you spend per week preparing data to give to AI tools? (copy-pasting, uploading, describing)
-- Have you built any custom solutions or workarounds for giving AI access to your data?
-- What happens if you need AI to answer a question about your data but you can't access the document easily? (reputation risk, time lost)
-- What have you paid for tools that help AI access personal data? (subscriptions, APIs, custom solutions)
-- How often do you wish Claude/ChatGPT could just 'know' your personal data without you having to paste it?
-**Look For:**
-- Time spent (preparing data, copy-pasting, manual work)
-- Money spent (tools, APIs, custom solutions)
-- Reputation risk (can't answer questions quickly, wrong answers from incomplete context)
-- Emotional intensity (frustration with manual workarounds)
-- Workarounds built (custom scripts, manual processes)
-##### Section 3: Value Proposition Testing (15 min) — Only If Problem Validated
-**Precondition**: Only proceed if problem exists (AI lacks personal data access, workarounds painful), and commitment signals present
-**Questions:**
-1. "We're thinking about solving this by giving Claude/ChatGPT access to your personal documents via MCP — so you could ask questions like 'What's in my contract with Acme Corp?' or 'Show me all my travel bookings next month' without copy-pasting. Does this address what you described?"
-   - Follow-ups:
-     - Does this match the problem you just told me about?
-     - What questions do you have?
-     - What's missing?
-     - Would this change how you use AI with your personal data?
-2. "Here's how we're thinking about it: You upload your documents (contracts, receipts, travel bookings), we extract structured data deterministically (no AI guessing), and then Claude/ChatGPT can query that data via MCP. Walk me through how you'd use this."
-   - Follow-ups:
-     - What would you do first after uploading documents?
-     - What questions would you ask the AI?
-     - Can you see yourself using this regularly?
-     - What would make you stop using it?
-3. "Tell me about the last time you noticed inconsistent or wrong answers from AI because it lacked your personal data context. What happened?"
-   - Follow-ups:
-     - What question were you asking?
-     - What went wrong?
-     - How did you discover the AI was wrong or missing context?
-     - Would deterministic, structured data have helped?
-4. "Walk me through a typical workflow: When you need to ask AI about your personal data, what do you do today? How would this change with MCP access?"
-   - Follow-ups:
-     - Current workflow: [what they described]
-     - New workflow: Upload once, query forever
-     - What would you gain? (time, accuracy, convenience)
-     - What concerns do you have?
-**Commitment Signal Questions:**
-- If this existed today, would you switch from your current workaround (copy-paste, manual prep)?
-- Would you recommend this to a colleague? What would you say?
-- What would need to be true for you to use this regularly?
-- What personal data questions do you wish you could ask AI but currently can't?
-##### Section 4: Closing (5 min)
-- "Anything else you'd like to share about how you use AI with your personal data?"
-- "Can we follow up in 2 weeks to show you a prototype?"
-- **Early Access Offer** (optional, after problem validated): "As a thank you for your time, I'd love to give you early access when we launch. Would that be helpful?"
-### 5. Mom Test Principles
-**Methodology**: The Mom Test (https://www.momtestbook.com)
-**Question Framing**: Focus on past behavior and concrete examples of AI interactions with personal data, not hypotheticals
-**Bias Avoidance:**
-- Separate problem discovery from solution validation
-- Ask about last time, specific examples, concrete AI workflows
-- Look for commitment signals: time spent preparing data, money spent on workarounds, tools built
-- Avoid leading questions and solution pitches during problem discovery
-- Don't ask "Would you use this?" — ask about past AI interactions instead
-**Commitment Signals:**
-- Time spent preparing data for AI ("I spend 30 min per week copy-pasting...")
-- Money spent on workarounds ("I pay for API access...", "I hired someone to...")
-- Custom solutions built ("I wrote a script to...", "I built a workflow to...")
-- Reputation risk ("If I can't answer quickly, client gets frustrated")
-- Emotional intensity ("This drives me crazy every time I need to paste documents")
-**Red Flags:**
-- Vague enthusiasm without commitment signals ("Sounds interesting")
-- Hypothetical interest without past behavior evidence ("I would definitely use this")
-- Solution-focused responses before problem validated ("Yes, I need this")
-- Agreement without evidence ("This could be useful")
-- No actual AI usage patterns described
-### 6. Post-Interview Notes
-Capture:
-- Problem validated? (yes/no, with evidence - actual AI usage patterns described)
-- Commitment signals found? (time preparing data, money spent, workarounds built)
-- Solution addresses problem? (if tested - does MCP access solve their workflow?)
-- Interest in solution? (if tested - would they use this?)
-- Red flags? (vague interest, no commitment signals, no actual AI usage)
-### 7. Optional Async Validation
-**Purpose**: Validate interview findings at scale (100-200 responses)
-**Timing**: Week -6 to Week -5 (after interviews complete)
-**When to Use**: After hypothesis formed from live interviews
-**Questions Focus**: Test validated hypothesis, gather quantitative data
-**Note**: Do NOT use to replace interview insights — only to validate that findings apply at scale
-### 8. Success Criteria
-- Screening survey: 40-50 responses, 20-25 willing to interview
-- Live interviews: ≥70% validate problem exists (struggle to give AI tools access to personal data context)
-- Live interviews: ≥60% express interest in MCP-based solution (only after problem validated)
-- Live interviews: ≥50% care about deterministic, structured data access (vs semantic search)
-- Live interviews: ≥60% see value in "upload once, query forever" workflow
-- Live interviews: Commitment signals present in ≥70% of interviews (time preparing data, money spent, workarounds built)
-- Optional validation: Survey confirms interview findings at scale (if conducted)
-### 9. Deliverable
-**Deliverable**: `value_discovery_report.md`
-Includes:
-- Synthesis of all interview findings
-- Problem validation results
-- Solution interest results
-- Commitment signal analysis
-- Red flags identified
-- Comparison against success criteria
-### 10. Related Documents
-- `discovery_plan.md` — Discovery overview and coordination
-- `usability_discovery_plan.md` — Usability discovery activities
-- `business_viability_discovery_plan.md` — Business viability discovery activities
-- `participant_recruitment_log.md` — Participant tracking
-- `docs/feature_units/standards/discovery_process.md` — Discovery process standard

package/docs/releases/archived/aspirational/v1.0.0/visibility_timing_analysis.md

@@ -1,402 +0,0 @@
-# Visibility Timing Risk Analysis: Dogfooding-First Approach
-
-**Date**: 2026-01-15  
-**Purpose**: Assess risk of missing visibility opportunities during build phase and how to capture visibility without premature commitments
-
-## Executive Summary
-
-**Risk Assessment**: **Medium-High**. Missing visibility windows during 12-week build phase is a real risk.
-
-**Solution**: **Strategic build-in-public content** can capture more visibility during industry announcements than traditional pre-launch marketing, because it is:
-- Authentic and timely (responds to industry news)
-- Thought leadership (positions you as expert, not just product)
-- No premature commitments (can pivot based on learnings)
-**Timing Principle**: Prioritize first credible articulation over first-mover speed. Developer preview during dogfooding lets you enter conversations early without overstating guarantees.
-
----
-
-## 1. Visibility Risk Scenarios
-
-### 1.1 Competitive Announcements (High Risk)
-
-**Scenario**: Major competitors announce new memory features during your 12-week build phase.
-
-**Examples:**
-- OpenAI announces structured memory for ChatGPT (beyond conversation-only)
-- Anthropic announces entity resolution or timeline features for Claude
-- Google Gemini announces cross-platform memory access
-- Supermemory.ai or other MCP memory tools launch major updates
-
-**Risk:**
-- ❌ **Mindshare capture**: Competitors capture attention, you're invisible
-- ❌ **Positioning difficulty**: Harder to differentiate after competitors announce
-- ❌ **Market timing**: You miss the conversation window
-- ❌ **First-mover advantage loss**: Competitors establish positioning first
-
-**Timing Windows:**
-- Industry announcements create 1-2 week "conversation windows"
-- During these windows, relevant content gets 3-5x more engagement
-- Missing these windows means lower visibility when you do launch
-
-### 1.2 MCP Ecosystem Growth (Medium Risk)
-
-**Scenario**: MCP ecosystem announcements during your build phase.
-
-**Examples:**
-- MCP protocol updates or new capabilities
-- Major MCP server launches (memory-related or otherwise)
-- MCP directory launches or major updates
-- MCP community events or conferences
-
-**Risk:**
-- ❌ **Ecosystem positioning**: Miss opportunity to position Neotoma in MCP ecosystem
-- ❌ **Community engagement**: Miss opportunity to engage with MCP community
-- ❌ **Directory visibility**: Miss opportunity to be included in ecosystem discussions
-
-### 1.3 Industry Conversations (Medium Risk)
-
-**Scenario**: Industry conversations about AI memory, privacy, determinism during your build phase.
-
-**Examples:**
-- Privacy debates about AI memory (provider-controlled vs. user-controlled)
-- Discussions about deterministic vs. ML-based extraction
-- Conversations about cross-platform AI memory
-- Technical discussions about MCP integration patterns
-
-**Risk:**
-- ❌ **Thought leadership**: Miss opportunity to establish thought leadership
-- ❌ **Community engagement**: Miss opportunity to engage with relevant communities
-- ❌ **Credibility building**: Miss opportunity to build credibility before launch
-
----
-
-## 2. Why Build-in-Public Captures MORE Visibility
-
-### 2.1 Authentic and Timely Response
-
-**Traditional Pre-Launch Marketing:**
-- Scheduled content (Week -4, Week -3, etc.)
-- Generic messaging (not tied to industry events)
-- Product-focused (waitlist CTAs)
-- **Problem**: Can't respond to industry announcements in real-time
-
-**Build-in-Public Content:**
-- ✅ **Real-time response**: Can respond to industry announcements immediately
-- ✅ **Authentic positioning**: "I'm building this because [industry announcement] doesn't solve [your problem]"
-- ✅ **Thought leadership**: Positions you as expert responding to industry news
-- ✅ **Higher engagement**: Timely content gets 3-5x more engagement
-
-**Example:**
-- **Day 1**: OpenAI announces structured memory
-- **Day 2**: Your build-in-public thread: "OpenAI's new memory is provider-controlled. I'm building Neotoma for user-controlled, deterministic memory. Here's why it matters..."
-- **Result**: 10x more engagement than scheduled pre-launch content
-
-### 2.2 Thought Leadership vs. Product Marketing
-
-**Traditional Pre-Launch Marketing:**
-- Focus: "Join waitlist for Neotoma"
-- Message: Product-focused
-- Engagement: Lower (people ignore marketing)
-- **Problem**: Marketing content gets less engagement than thought leadership
-
-**Build-in-Public Content:**
-- ✅ **Focus**: Technical insights, architectural decisions, industry analysis
-- ✅ **Message**: Thought leadership, not product marketing
-- ✅ **Engagement**: Higher (people engage with valuable content)
-- ✅ **Credibility**: Establishes you as expert before product launch
-
-**Example:**
-- **Pre-Launch Marketing**: "Join waitlist for Neotoma - privacy-first AI memory" (low engagement)
-- **Build-in-Public**: "Why deterministic extraction matters for AI memory (thread on OpenAI's announcement)" (high engagement)
-
-### 2.3 No Premature Commitments
-
-**Traditional Pre-Launch Marketing:**
-- Creates expectations (waitlist, beta, launch date)
-- Commits to specific positioning
-- **Problem**: Can't pivot based on learnings or industry changes
-
-**Build-in-Public Content:**
-- ✅ **No expectations**: Just sharing learnings, not promising product
-- ✅ **Flexible positioning**: Can evolve positioning based on learnings
-- ✅ **Pivot-friendly**: Can change direction without breaking promises
-
----
-
-## 3. Strategic Build-in-Public Content Strategy
-
-### 3.1 Reactive Content (High Visibility)
-
-**When Industry Announcements Happen:**
-- **Immediate response** (within 24-48 hours)
-- **Technical analysis**: "Here's what [competitor] announced, here's why it doesn't solve [problem]"
-- **Positioning**: "I'm building Neotoma to solve [problem] differently"
-- **No CTA**: Just thought leadership, not marketing
-
-**Examples:**
-- OpenAI announces structured memory → Your thread: "Why provider-controlled memory doesn't solve the privacy problem"
-- Anthropic announces entity resolution → Your thread: "Why deterministic entity resolution matters (building this in Neotoma)"
-- MCP ecosystem announcement → Your thread: "MCP integration patterns I'm discovering while building Neotoma"
-
-**Time Investment**: ~2-3 hours per industry announcement (high ROI)
-
-### 3.2 Proactive Content (Ongoing Visibility)
-
-**Regular Build-in-Public Content:**
-- **Weekly progress updates**: Screenshots, demos, learnings
-- **Technical deep-dives**: Determinism, MCP integration, entity resolution
-- **Architecture decisions**: Why you're making specific technical choices
-- **No CTA**: Just sharing, not marketing
-
-**Examples:**
-- "Week 3 of building Neotoma: Here's what I learned about deterministic OCR"
-- "MCP integration patterns: Lessons from replacing parquet MCP in ateles"
-- "Why I'm building a truth layer instead of using semantic search"
-
-**Time Investment**: ~5-10 hours/month (ongoing)
-
-### 3.3 Community Engagement (Ecosystem Visibility)
-
-**Engage with Industry Conversations:**
-- **Comment on competitor announcements**: Value-first, not promotional
-- **Engage with MCP community**: Share insights, answer questions
-- **Participate in relevant discussions**: Privacy, determinism, MCP integration
-- **No CTA**: Just helpful contributions, not marketing
-
-**Examples:**
-- Comment on OpenAI memory announcement: "Interesting approach. I'm exploring user-controlled memory because..."
-- Answer MCP questions: "I'm building Neotoma and discovered this MCP pattern..."
-- Engage with privacy discussions: "Privacy-first architecture is why I'm building Neotoma differently..."
-
-**Time Investment**: ~2-3 hours/week (ongoing)
-
----
-
-## 4. Visibility Capture Strategy
-
-### 4.1 Industry Announcement Response Framework
-
-**When Competitor/Industry Announcement Happens:**
-
-1. **Immediate Analysis** (Day 1):
-   - What did they announce?
-   - What problem does it solve?
-   - What problem does it NOT solve? (your opportunity)
-   - How does Neotoma approach it differently?
-
-2. **Content Creation** (Day 1-2):
-   - Technical thread on Twitter (5-7 tweets)
-   - Blog post (if significant announcement)
-   - Community engagement (comments, discussions)
-
-3. **Positioning**:
-   - "I'm building Neotoma to solve [problem they didn't solve]"
-   - "Here's why [their approach] doesn't work for [use case]"
-   - "Here's how Neotoma approaches it differently"
-
-4. **No CTA**:
-   - Just thought leadership, not marketing
-   - No waitlist, no beta, no product pitch
-   - Just valuable insights
-
-**Example Response Timeline:**
-- **Day 1, 10am**: OpenAI announces structured memory
-- **Day 1, 2pm**: Your Twitter thread: "OpenAI's memory is provider-controlled. I'm building Neotoma for user-controlled, deterministic memory. Here's why it matters..."
-- **Day 2**: Blog post: "Why Provider-Controlled Memory Doesn't Solve the Privacy Problem"
-- **Day 3**: Community engagement: Comments on relevant discussions, answer questions
-
-**Result**: High visibility during industry announcement window, establishes thought leadership
-
-### 4.2 MCP Ecosystem Engagement
-
-**Ongoing Engagement:**
-- **Monitor MCP ecosystem**: GitHub, Discord, Twitter, directories
-- **Engage with MCP community**: Share insights, answer questions
-- **Position Neotoma**: "I'm building Neotoma, here's what I'm learning about MCP integration"
-- **No CTA**: Just helpful contributions, not marketing
-
-**Examples:**
-- MCP Discord: "I'm building Neotoma and discovered this MCP pattern for memory..."
-- MCP GitHub discussions: "Here's how I'm handling entity resolution in my MCP server..."
-- MCP Twitter: "MCP integration patterns I'm learning while building Neotoma..."
-
-**Result**: Ecosystem visibility, community engagement, credibility building
-
-### 4.3 Technical Thought Leadership
-
-**Regular Technical Content:**
-- **Weekly threads**: Technical insights, architectural decisions
-- **Blog posts**: Deep-dives on determinism, MCP integration, entity resolution
-- **Community contributions**: Answer questions, share learnings
-
-**Examples:**
-- "Why deterministic extraction matters for AI memory (technical deep-dive)"
-- "MCP integration patterns: Lessons from building Neotoma"
-- "Entity resolution in practice: What I'm learning building Neotoma"
-
-**Result**: Technical credibility, thought leadership, visibility
-
----
-
-## 5. Risk Mitigation: Missing Visibility Windows
-
-### 5.1 Risk: Missing Industry Announcement Windows
-
-**Mitigation:**
-- ✅ **Monitor industry news**: Set up alerts for OpenAI, Anthropic, Google, MCP ecosystem
-- ✅ **Rapid response capability**: Have content templates ready for common announcement types
-- ✅ **Build-in-public content**: Can respond immediately (no waitlist/beta overhead)
-- ✅ **Thought leadership positioning**: Establishes you as expert, not just product
-
-**Advantage Over Pre-Launch Marketing:**
-- Pre-launch marketing: Scheduled content, can't respond to announcements
-- Build-in-public: Can respond immediately, higher engagement
-
-### 5.2 Risk: Missing MCP Ecosystem Visibility
-
-**Mitigation:**
-- ✅ **Ongoing engagement**: Regular participation in MCP community
-- ✅ **Technical contributions**: Share insights, answer questions
-- ✅ **Ecosystem positioning**: Position Neotoma in MCP ecosystem discussions
-- ✅ **Directory preparation**: Prepare MCP directory listings (submit after dogfooding)
-
-**Advantage Over Pre-Launch Marketing:**
-- Pre-launch marketing: Generic messaging, not ecosystem-specific
-- Build-in-public: Ecosystem-specific engagement, higher relevance
-
-### 5.3 Risk: Missing Thought Leadership Opportunities
-
-**Mitigation:**
-- ✅ **Regular technical content**: Weekly threads, blog posts
-- ✅ **Industry analysis**: Respond to industry announcements with technical insights
-- ✅ **Community engagement**: Participate in relevant discussions
-- ✅ **Credibility building**: Establish expertise before product launch
-
-**Advantage Over Pre-Launch Marketing:**
-- Pre-launch marketing: Product-focused, lower engagement
-- Build-in-public: Thought leadership-focused, higher engagement
-
----
-
-## 6. Comparison: Pre-Launch Marketing vs. Strategic Build-in-Public
-
-| Aspect | Pre-Launch Marketing | Strategic Build-in-Public |
-|--------|---------------------|---------------------------|
-| **Visibility During Industry Announcements** | Low (scheduled content, can't respond) | High (can respond immediately) |
-| **Engagement** | Low (marketing content) | High (thought leadership content) |
-| **Thought Leadership** | Low (product-focused) | High (technical insights) |
-| **Ecosystem Engagement** | Low (generic messaging) | High (ecosystem-specific) |
-| **Timing Flexibility** | Low (scheduled) | High (reactive) |
-| **Credibility Building** | Medium (product marketing) | High (technical expertise) |
-| **Risk of Missing Windows** | High (can't respond to announcements) | Low (can respond immediately) |
-
-**Winner**: **Strategic Build-in-Public** captures MORE visibility during industry announcements
-
----
-
-## 7. Recommended Strategy: Strategic Build-in-Public
-
-### 7.1 During Build Phase (Week 0-12)
-
-**Ongoing Activities:**
-1. **Monitor Industry News** (1 hour/week):
-   - Set up alerts for OpenAI, Anthropic, Google, MCP ecosystem
-   - Monitor relevant communities (Twitter, Discord, GitHub)
-
-2. **Regular Build-in-Public Content** (5-10 hours/month):
-   - Weekly progress updates (Twitter, Indie Hackers)
-   - Technical deep-dives (blog posts, Twitter threads)
-   - Architecture decisions (why you're making specific choices)
-
-3. **Reactive Content** (2-3 hours per industry announcement):
-   - Immediate response to industry announcements (within 24-48 hours)
-   - Technical analysis, positioning, thought leadership
-   - No CTA, just valuable insights
-
-4. **Community Engagement** (2-3 hours/week):
-   - Engage with MCP community
-   - Participate in relevant discussions
-   - Answer questions, share insights
-
-**Total Time Investment**: ~10-15 hours/month (vs. 35 hours for pre-launch marketing)
-
-### 7.2 After Dogfooding (Week 12+)
-
-**If Product Works:**
-- Proceed to discovery + marketing
-- Build-in-public content continues (now with working product to show)
-- Launch waitlist/beta/announcement after discovery validation
-
-**If Product Doesn't Work:**
-- Continue building in public
-- Iterate based on learnings
-- Build-in-public content continues (authentic about challenges)
-
----
-
-## 8. Example: Industry Announcement Response
-
-### Scenario: OpenAI Announces Structured Memory (Week 6 of Build)
-
-**Day 1, 10am**: OpenAI announces structured memory for ChatGPT
-
-**Day 1, 2pm**: Your Twitter Thread (5-7 tweets):
-```
-1. OpenAI just announced structured memory for ChatGPT. 
-   It's provider-controlled and platform-locked.
-
-2. I'm building Neotoma for user-controlled, deterministic memory 
-   that works across all AI tools via MCP.
-
-3. Here's why provider-controlled memory doesn't solve the privacy problem:
-   [Technical explanation]
-
-4. Here's why deterministic extraction matters:
-   [Technical explanation]
-
-5. Here's how Neotoma approaches it differently:
-   [Architecture explanation]
-
-6. Building for my own use (ateles integration), but here's what 
-   I'm learning about AI memory architecture.
-
-7. What do you think? Are you building anything in this space?
-```
-
-**Day 2**: Blog Post:
-- "Why Provider-Controlled Memory Doesn't Solve the Privacy Problem"
-- Technical deep-dive on Neotoma's approach
-- No CTA, just thought leadership
-
-**Day 3**: Community Engagement:
-- Comment on OpenAI announcement discussions
-- Answer questions about user-controlled memory
-- Engage with privacy-focused communities
-
-**Result:**
-- ✅ High visibility during industry announcement window
-- ✅ Thought leadership positioning
-- ✅ Technical credibility
-- ✅ No premature commitments (just sharing learnings)
-
----
-
-## 9. Conclusion
-
-**Visibility Risk Assessment:**
-- **Risk Level**: Medium-High (missing visibility windows is a real risk)
-- **Mitigation**: Strategic build-in-public content actually captures MORE visibility than pre-launch marketing
-
-**Key Insights:**
-1. **Build-in-public captures MORE visibility** during industry announcements (can respond immediately)
-2. **Thought leadership > Product marketing** (higher engagement, better credibility)
-3. **Reactive content > Scheduled content** (responds to industry news, higher relevance)
-4. **No premature commitments** (can pivot based on learnings, industry changes)
-
-**Recommended Strategy:**
-- **Strategic build-in-public content** (10-15 hours/month)
-- **Reactive content** (2-3 hours per industry announcement)
-- **Community engagement** (2-3 hours/week)
-- **No waitlist/beta** (just thought leadership, not marketing)
-
-**Result**: Higher visibility, better credibility, no premature commitments, can respond to industry announcements in real-time.