Initial commit — design docs + Phase 0 validation harness

Sets up the Socrata project repo with:

docs/ — strategy and design documents
  - idea.md: full product vision
  - implementation-plan.md: Phase 0 + Phase 1 MVP plan
  - phase-0-validation.md: 2-week validation experiment strategy
  - phase-0-plan.md: concrete Phase 0 build plan
  - phase-0-results.md: Phase 0 gate outcome — GO for MVP
  - sysml-modeling.md: metamodel + SE discipline + validation rules
  - socrates.md: agent character, surfaces, modes, prompts, lifecycle
  - sync.md: bidirectional text↔diagram sync engineering
  - design-source/: HTML/CSS/JS handoff bundle from Claude Design

phase-0/ — validated harness (CLI, no UI, no DB)
  - LM Studio (local OpenAI-compatible) generation + detection + judge
  - PlantUML rendering for SysML model visualization
  - 10-seed corpus (8 working + 2 holdouts)
  - 5 corpus runs with iteration history in reports/
  - Final gate: 10/10 pass, mean 4.32/5, holdouts validated

Phase 1 MVP scope and milestones documented in implementation-plan.md.

phase-0/seeds/01-aristotle.json

@@ -0,0 +1,18 @@
+{
+  "id": "01-aristotle",
+  "title": "Aristotle",
+  "problem": "First-year STEM students at large public universities frequently disengage from coursework not because the material is intractable, but because they lack a low-stakes thinking partner during the long tail between lectures and office hours.",
+  "targetUser": "Undergraduates at large public universities, weeks 3-10 of an intro course.",
+  "desiredOutcome": "Students re-engage with material via a low-stakes thinking partner — without producing solutions.",
+  "initialHypothesis": "Students will adopt a tool that explicitly refuses to solve their homework, because the market is saturated with answer-givers.",
+  "constraints": [
+    "Must never output a complete solution to a graded problem",
+    "Response latency under 1.2s P50 to preserve flow",
+    "FERPA tenancy — coursework never leaves institutional boundary"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "easy",
+    "testsFor": ["baseline", "consumer-edu", "explicit-constraints"]
+  }
+}

phase-0/seeds/02-habit-coach.json

@@ -0,0 +1,18 @@
+{
+  "id": "02-habit-coach",
+  "title": "Habit Coach",
+  "problem": "Parents of school-age kids want to limit screen time but the existing tooling (OS-level limits, single-app blockers) is either too coarse or too easily bypassed. Worse, parents have no visibility into whether limits are working — they only see the friction.",
+  "targetUser": "Parents of kids aged 8–14 who already use one or more screen-time tools and find them inadequate.",
+  "desiredOutcome": "A weekly conversation between parent and child about screen time becomes routine, grounded in shared data instead of conflict.",
+  "initialHypothesis": "Parents will adopt a tool that's child-visible and child-collaborative if it lets them stop being the enforcer.",
+  "constraints": [
+    "Must work alongside iOS Screen Time and Google Family Link, not replace them",
+    "Child data must be parent-controlled per COPPA (US) and GDPR-K (EU)",
+    "No social features for kids — no leaderboards, no public sharing"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "medium",
+    "testsFor": ["consumer-saas", "multi-actor", "regulatory"]
+  }
+}

phase-0/seeds/03-redline-ai.json

@@ -0,0 +1,19 @@
+{
+  "id": "03-redline-ai",
+  "title": "Redline AI",
+  "problem": "In-house legal teams spend hours per contract on first-pass redlining — tedious mechanical work that doesn't require senior judgment but does require consistent application of company-specific clause libraries. Junior associates burn out doing it; outside counsel charges $400/hr for it.",
+  "targetUser": "General Counsel and Legal Operations leads at companies with 50–500 employees that handle 100+ contracts per quarter (SaaS, marketplaces, services).",
+  "desiredOutcome": "First-pass redlines on standard contracts complete in minutes, with the GC reviewing AI suggestions instead of authoring redlines from scratch.",
+  "initialHypothesis": "Legal teams will trust AI for first-pass mechanical redlining but reserve final judgment for humans — i.e., the bar is 'better than a tired junior associate', not 'better than the GC'.",
+  "constraints": [
+    "Must integrate with Microsoft Word — that's where the work happens",
+    "Contract content must never train cross-customer models (privilege concerns)",
+    "Audit trail of every change suggestion must be exportable",
+    "SOC 2 Type II compliance from day one"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "medium",
+    "testsFor": ["b2b", "domain-heavy", "regulatory", "integration"]
+  }
+}

phase-0/seeds/04-skillswap.json

@@ -0,0 +1,18 @@
+{
+  "id": "04-skillswap",
+  "title": "SkillSwap",
+  "problem": "Remote workers have skills they could exchange (a designer who needs a quick legal opinion; a lawyer who wants design help with a deck) but no good marketplace for non-cash skill trades. Cash marketplaces (Fiverr, Upwork) introduce a money layer that kills casual exchange.",
+  "targetUser": "Remote knowledge workers, 25–45, who value reciprocal exchange over hiring out and have at least one marketable skill outside their day job.",
+  "desiredOutcome": "Two strangers complete a skill swap end-to-end (matched, scoped, exchanged, both rated positively) within one week of joining.",
+  "initialHypothesis": "A two-sided marketplace works for skill swap if matching is high-trust (reputation portable from LinkedIn) and scope is bounded (max 2-hour exchanges, no open-ended retainers).",
+  "constraints": [
+    "No money changes hands inside the platform",
+    "Identity verification via LinkedIn or similar at signup",
+    "Disputes capped to mediator review — no chargeback equivalent"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "medium",
+    "testsFor": ["marketplace", "two-sided", "trust-mechanisms"]
+  }
+}

phase-0/seeds/05-cyclist-thing.json

@@ -0,0 +1,12 @@
+{
+  "id": "05-cyclist-thing",
+  "title": "Cyclist Thing",
+  "problem": "I want to build something for cyclists. Maybe an app, maybe a tool, not sure yet.",
+  "targetUser": "Cyclists.",
+  "desiredOutcome": "Something useful for them.",
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "failure-prone",
+    "testsFor": ["vague-seed", "underspecified", "socrates-pushback"]
+  }
+}

phase-0/seeds/06-eventstream.json

@@ -0,0 +1,18 @@
+{
+  "id": "06-eventstream",
+  "title": "EventStream",
+  "problem": "Existing event-streaming systems (Kafka, Pulsar, Redpanda) treat backpressure as an afterthought — slow consumers either drop messages, lag indefinitely, or trigger broker-level fan-out failures. Modern application workloads (especially LLM-pipelined ones) need first-class backpressure semantics that the producer can observe in-band.",
+  "targetUser": "Platform engineers at companies running event-driven infrastructure at moderate scale (100k–10M events/day) where consumer lag matters and Kafka's operational complexity is unwelcome.",
+  "desiredOutcome": "A Go-native streaming system where producers receive backpressure signals as a first-class part of the protocol, and consumer-driven flow control is the default.",
+  "initialHypothesis": "Teams will adopt a Kafka alternative if (a) operations are 10× simpler — single binary, no ZooKeeper, no Schema Registry — and (b) backpressure is a first-class protocol feature, not a config flag.",
+  "constraints": [
+    "Wire-compatible with Kafka producer protocol for migration safety",
+    "Single binary, single config file, runs on a single node up to 100k events/sec",
+    "Apache 2.0 license"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "hard",
+    "testsFor": ["technical", "infra", "metamodel-strain"]
+  }
+}

phase-0/seeds/07-pet-translator.json

@@ -0,0 +1,17 @@
+{
+  "id": "07-pet-translator",
+  "title": "Pet Translator",
+  "problem": "Pet owners want to know what their dogs and cats are 'saying'. Existing apps are gimmicks that produce random translations.",
+  "targetUser": "Pet owners with dogs or cats, particularly first-time owners struggling to decode their pet's behavior.",
+  "desiredOutcome": "Owners get a believable, useful interpretation of their pet's vocalization and body language — and use that interpretation to bond with their pet.",
+  "initialHypothesis": "Recent advances in audio ML and computer vision make it possible to infer pet emotional state from vocalization + body language with sufficient accuracy that users perceive the output as 'translation' even if scientifically it's classification.",
+  "constraints": [
+    "On-device inference for privacy (no cloud upload of household audio/video)",
+    "Honest about the difference between 'emotion classification' and 'translation' in onboarding"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "failure-prone",
+    "testsFor": ["feasibility-suspect", "wishful-tech", "honesty-bar"]
+  }
+}

phase-0/seeds/08-quiet-hours.json

@@ -0,0 +1,18 @@
+{
+  "id": "08-quiet-hours",
+  "title": "Quiet Hours",
+  "problem": "Knowledge workers lose 2–3 hours per day to fragmented calendars — meetings scheduled mid-deep-work, no time long enough to make progress on anything substantial. Existing focus-time tools are either too rigid (block 9am–noon every day) or too easily overridden by colleagues.",
+  "targetUser": "Individual contributors and engineering managers at 50+ person companies that use Google Calendar or Outlook and have collaborative cultures (lots of ad-hoc meetings).",
+  "desiredOutcome": "Each user reliably gets at least 2 contiguous deep-work blocks per week without it costing them political capital.",
+  "initialHypothesis": "If the tool can negotiate the protected blocks invisibly (proposing alternative meeting times that work for both parties), users get focus without the social cost of saying no.",
+  "constraints": [
+    "Read-write Google Calendar and Outlook (Microsoft Graph) integration",
+    "Must respect timezone differences explicitly in proposals",
+    "No automatic blocking without user confirmation for the first 30 days (build trust)"
+  ],
+  "meta": {
+    "isHoldout": false,
+    "expectedDifficulty": "medium",
+    "testsFor": ["productivity", "calendar-integration", "constraint-heavy"]
+  }
+}

phase-0/seeds/09-carbon-coach.json

@@ -0,0 +1,18 @@
+{
+  "id": "09-carbon-coach",
+  "title": "Carbon Coach",
+  "problem": "Climate-aware consumers want to lower their personal carbon footprint but can't tell which lifestyle changes meaningfully matter and which are theatrical (e.g., reusable straws vs. flying less). Existing carbon calculators give one-shot estimates without sustained behavior change support.",
+  "targetUser": "Climate-aware consumers in high-income countries, 25–55, already willing to change habits but unsure where to invest effort.",
+  "desiredOutcome": "Sustained, measurable per-user reduction in monthly footprint over a 6-month period, driven by 2–3 high-impact behavior changes per user.",
+  "initialHypothesis": "Social comparison via leaderboards (anonymous opt-in) drives behavior change more than absolute numbers — people respond to relative status more than abstract climate goals.",
+  "constraints": [
+    "Footprint estimates must cite their methodology (no black-box numbers)",
+    "Anonymous social features only — no real-name sharing",
+    "EU compliance with the upcoming Green Claims Directive"
+  ],
+  "meta": {
+    "isHoldout": true,
+    "expectedDifficulty": "medium",
+    "testsFor": ["behavioral", "social-mechanics", "regulatory", "sustained-engagement"]
+  }
+}

phase-0/seeds/10-telemetry-lite.json

@@ -0,0 +1,18 @@
+{
+  "id": "10-telemetry-lite",
+  "title": "Telemetry Lite",
+  "problem": "PostHog has won the product-analytics-with-session-replay category but its self-hosted footprint is too heavy for small teams (Postgres, Kafka, ClickHouse, Redis, multiple workers). Privacy-conscious teams who want to self-host are stuck either running PostHog at $200/month of infra OR using nothing and getting blind product decisions.",
+  "targetUser": "Engineering teams at 5–30 person companies (or solo founders) who want product analytics + session replay but refuse to send user data to a third-party SaaS, and don't have an infra engineer to run PostHog.",
+  "desiredOutcome": "A single Docker container, single SQLite file, that delivers 80% of PostHog's value — events, funnels, session replay — at 5% of the operational complexity.",
+  "initialHypothesis": "There's a self-host segment that won't pay for SaaS analytics OR run a PostHog cluster. They'll accept a smaller feature set in exchange for trivial ops.",
+  "constraints": [
+    "Single binary or single Docker container, SQLite-only storage option",
+    "Apache 2.0 license, hosted version is a separate commercial play",
+    "GDPR-compliant data export and deletion built in"
+  ],
+  "meta": {
+    "isHoldout": true,
+    "expectedDifficulty": "hard",
+    "testsFor": ["technical", "oss", "competitive-positioning"]
+  }
+}