0. Document Purpose

This PRD is for the Sproutopia product team and the downstream BMad workflows that build from it — UX design, technical architecture, and epic/story breakdown. It builds on the Product Brief: Sproutopia (_bmad-output/planning-artifacts/briefs/brief-homesteading-2026-07-02/brief.md) and its addendum (the complete concept capture — persona, seven engines, closed loop, Winter War-Room, Corvid Corps, monetization, copyright posture); it does not duplicate them. Where this PRD and the brief conflict, this PRD wins.

Structure: vocabulary is anchored in the Glossary (§3) and used verbatim throughout. Features are grouped (§4) with globally-numbered Functional Requirements (FR-N) nested under them; FRs carry stable IDs even if features are reorganized. Requirements are stated as capabilities, not implementation — technology choices (Expo, RAG grounding, AI vendors, storage) live in the PRD addendum (addendum.md), not here. Inferences the team has not confirmed are tagged Assumption ... inline and indexed in §16. Callouts for the PM are tagged Note for PM .

Priority within v1. Doug's committed decision is that the full closed-loop world ships in v1 with a staged reveal — this is not a phased-MVP product. To give the architect and epic breakdown a build sequence, this PRD tags a P1 · SPINE priority on the requirements that prove the core loop and must land first (the Digital Twin, real-signal intake, the AI Brain, the weather-aware Calendar, the Beds flagship, the Knowledge Base rules engine, and the guided micro-win path). Everything untagged is P2: still committed to v1, sequenced behind the spine.

1. Vision

Sproutopia turns a person's real backyard into a living digital twin — a gamified, AI-guided companion that walks an overwhelmed beginner through a complete plan → plant → tend → harvest loop to a real first harvest, then keeps working through winter to turn this year's results into next year's smarter plan. Its core promise is subtraction, not more content: "here is exactly what to do this month, in your space, with your sunlight, budget, and skill level — ignore everything else."

Underneath the beginner's front door sits a full closed-loop homestead: seven interconnected engines — Beds, Livestock, Orchard, Bees, Weather, Cellar/Pantry, and a Knowledge Base — that reveal themselves progressively as the user's real homestead grows. The whole system is one core mechanic reskinned seven times: track a living thing's real life-cycle through evolutions, fed by real-world signals (harvests, photos, check-ins), where each thing buffs its neighbors. It is a real-life homestead-management tool wearing a game's clothes; a "win" is always a real, flourishing homestead — never points for their own sake.

It matters because the enemy is not ignorance but overwhelm and abandonment. The aspiring homesteader has started and quietly failed several times; every dead seedling feels like proof she's "not that kind of person." Sproutopia is the first product that optimizes for a first successful season and the restored confidence that comes with it — measured three ways at once: a flourishing homestead over time (the feeling), "days of food on hand" in the larder (security), and a rising Self-Sufficiency Score (%). If it works, it keeps the dream from ever quietly getting shelved again.

2. Target User

2.1 Jobs To Be Done

Primary user — "Megan," the aspiring beginner homesteader (38, half-acre, two kids, overloaded not lazy; full persona in the brief addendum):

• Functional: "Tell me exactly what to do this month in my actual yard, scoped to my sunlight, budget, schedule, and beginner skill — and let me ignore everything else." Get to a real first harvest.
• Functional: "When something looks wrong with this plant in my bed, tell me what it is and what to do — not a forum's generic answer."
• Emotional: "Prove to me I can do this, so the dream stops shrinking." Restore confidence and follow-through; fight the abandonment spiral.
• Emotional / identity: "Make me the kind of person who comes back better next season."
• Social: "Let my family feel we're building something real with our hands" (and, later, play the real farm together).
• Contextual: "Work with muddy hands, in the yard, in thirty seconds — and don't make winter a dead zone where I drift away."

Secondary user — the active homesteader (already running beds/animals): "Give me better tracking, real AI diagnosis, pedigree/rotation memory, and year-round planning." They enter the staged reveal further along and adopt engines faster.

2.2 Non-Users (v1)

• The 40-acre / commercial farmer needing herd-management ERP, compliance records, or equipment/field-scale tooling.
• The pure-content consumer who wants a video library or social feed with no real plot to model. Sproutopia's value requires a real property to twin.
• The off-grid prepper seeking energy/water/structural infrastructure planning (Sproutopia's Self-Sufficiency Score touches this but v1 is food-first).
• Children as independent account holders. Kids participate only through a parent's household in the roadmap Family Co-op (not v1), which keeps COPPA out of v1 scope. Assumption no standalone child accounts in v1.

2.3 Key User Journeys

Named-persona narratives, numbered globally. FRs reference these by ID inline ("Realizes UJ-N").

• UJ-1. Megan onboards and meets her yard as a living map. Megan, burned by five fizzled starts, installs the app on her phone and is asked only for her location and a quick sketch/walk of her property. The app seeds a Digital Twin — her zone, frost dates, sun exposure, and a tailored starter plan appear on a map of her yard. Entry state: unauthenticated, skeptical. Path: grant location → confirm/adjust property boundary and sun → pick an ambition level → see the twin populate. Climax: she sees her side yard on screen with "start here" beds and a plan scoped to her — not a 40-acre stranger's. Resolution: she has one clear next action, not 147 videos. Edge case: GPS/boundary is wrong → she nudges pins by hand; the app never blocks on perfect data.

• UJ-2. Megan gets a guaranteed first win, then pays after proof. Free tier. The guided micro-win path routes her to a fast, near-guaranteed first crop (radish/lettuce/herbs). Path: plant the recommended micro-win crop → the Calendar gives 3–4 dead-simple timed tasks → she logs a harvest with a photo. Climax: she logs her first real harvest and the app celebrates "it actually worked." At that peak-motivation moment — never on faith, never in the dead season — the Pay-after-proof upgrade to the Gardening tier fires. Resolution: confidence restored; she unlocks the core gardening experience. Edge case: the micro-win crop fails → Failure Autopsy reframes it as data ("here's what your yard told us"), offers a fast retry, and does not trigger the paywall.

• UJ-3. Megan's 30-second weekly tending loop. Biometric-authenticated, opens the app between chores. Path: the weather-aware Task Calendar shows today's short list scoped to her skill → she walks to bed 3, does a one-tap check-in (thriving/struggling), snaps a bed photo pinned to the tile, taps a harvest weight. Climax: the tile's time-lapse grows another ring; the task list re-sorts against live weather ("rain coming — skip watering"). Resolution: two minutes, done, the twin is more real. Edge case: muddy hands → she uses a voice note instead of typing; AI transcribes it into the tile's history.

• UJ-4. Megan asks the AI what's wrong with this plant. Something looks off. Path: she photographs the leaf → Vision AI reads it (early blight) → the AI Agronomist, grounded in her twin (this bed, this variety, her zone, her rotation history), tells her what it is, what to do this week, and what not to buy. Climax: a specific, correct, plot-aware answer in seconds. Resolution: she acts before it spreads. Edge case: low-confidence diagnosis → the AI says so, offers the two most likely causes and a safe next step, and never fabricates certainty.

• UJ-5. A frost villain threatens her specific beds and animals. The Weather engine ingests an incoming cold night. Path: a push arrives — "frost tonight → protect the peppers in bed 3, move the two-week chicks, insulate the hive" — flagging the exact real things at risk. She completes the real prep tasks before dark. Climax: "the power is yours" — she beat the weather; next morning her plants live. Resolution: a concrete save she can feel, and a lesson in why the whole homestead moves together. Edge case: she's away → the app gives the lowest-effort protective action and logs what's at risk for triage.

• UJ-6. Winter isn't dead — Megan re-buys the dream. November, off-season. Path: the Winter War-Room opens with her Homestead Report Card (attempted vs. harvested, best performer, biggest failure, money spent vs. food value, top 3 changes) → December she picks next year's ambition and gets a Next-Season Blueprint → January a seed & supply list that tells her what she already owns and what not to buy → February seed-starting dates and reminders. Climax: "you didn't fail — you learned what your yard was telling you," and a realistic, cheaper, exciting spring plan. Resolution: she's already gardening in January instead of waiting; the dream didn't get shelved. Edge case: thin first-season data → the Report Card leans on check-ins and photos, not just harvest weights, so it still has something true to say.

• UJ-7. The homestead asks to grow — a closed-loop upsell. Megan is throwing kitchen scraps in the trash. Path: a friendly gamified CTA appears — "your scraps are wasted 🗑️ → add Livestock: scraps → feed → manure → richer beds. Close the loop." → she adds the à-la-carte engine (or all-access). Climax: she watches manure charge her soil tiles and next season's trails run hotter. Resolution: expansion feels like the game rewarding her, not an upsell. Edge case: she declines → the CTA backs off, never nags, and the core loop is unaffected.

• UJ-8. The active homesteader enters further along. Dana already runs four beds and six hens. Path: onboarding detects existing scope → the staged reveal starts deeper (Beds + Livestock live immediately) → Dana imports her flock into the Pokédex registry with pedigree, logs weights, and uses Vision AI on a frame photo for varroa. Climax: real diagnosis and pedigree memory a spreadsheet can't give. Resolution: faster engine adoption, quicker path to all-access. Edge case: Dana wants bulk entry → batch logging avoids one-at-a-time friction.

• UJ-9. Megan discovers the Corvid Corps and collects what they bring. Never advertised. Path: Megan leaves scraps on the same fence post for days (habituation) and Vision AI notices crows recurring in her bed photos → a rare event fires: "a crow has been watching you. It left something on the post." → a trust ladder begins (curious loner → trusting regular → a small murder (flock) → a trained Corvid Corps), earned through consistency, never cash. Climax: at her first successful harvest a crow gifts a magic feather — the confidence she thought she lacked was hers all along; and as trust grows the crows start leaving real shiny trinkets on the post, which she photographs one by one into a found-object Pokédex (FR-64). Resolution: a screenshot-worthy, ever-growing collection and a living pest-defense aura across her engines. Edge case (grounded): crows also raid seedlings → "training" is really redirecting via a decoy station away from the beds — a real gardener's trade-off, not a cheat.

3. Glossary

Downstream workflows and readers must use these terms exactly. FRs, UJs, and Success Metrics use these terms verbatim; synonyms anywhere in the PRD are a discipline violation.

• Digital Twin (or Twin) — the living map/model of the user's real property: boundary, zones, sun, and every Bed, animal, tree, hive, and structure pinned to a real location. All engines render on it. Includes the Soil-Life View.
• Property — the user's real parcel that the Twin models. One user has one Property in v1. Assumption single active property per account in v1.
• Bed — a discrete growing area pinned on the Twin, the unit of the Beds engine and rotation.
• Rotation Trail (or Trail) — the glowing, plant-family-colored path a planting lays across Beds over time; the spatial representation of crop rotation.
• Self-Collision — the blocked state when a new planting would cross a recent same-family Rotation Trail (a rotation guardrail caught before the mistake).
• Ghost-Line — a faded past-season Rotation Trail, toggleable, forming the season-over-season rotation history ("tree-rings").
• Engine — one of the seven interconnected subsystems (Beds, Livestock, Orchard, Bees, Weather, Cellar/Pantry, Knowledge Base), each a costume over the shared core mechanic.
• Real-World Signal (or Signal) — a user-supplied truth input about the real homestead: a logged harvest, a photo, a check-in, or a voice note.
• Check-In — a one-tap status Signal on a living thing: thriving / struggling / dead, plus optional note.
• Harvest Log — a Signal recording a harvest event, optionally with weight and photo, attributed to a Bed/plant/animal.
• Vision AI — the capability that reads user photos to diagnose disease, pests, readiness/growth stage, and (livestock/bees) laying pattern, varroa, queen cells.
• AI Agronomist — the grounded advisor that has "read" the user's entire Twin and answers questions personalized to their plot; never generic, never fabricated.
• AI Brain — Vision AI and AI Agronomist together.
• Knowledge Base — the reference-data engine that is also the hidden rules engine grounding every other engine and the AI Brain. Includes the Almanac Layer.
• Task Calendar (or Calendar) — the weather-aware, skill-scoped list of what to do and when; the engine of the guided loop.
• Weather Quest — a gamified elemental event the Weather engine raises from real incoming conditions, flagging the exact real Beds/animals/hives at risk and the prep tasks to win.
• Elemental Balance Meter — the five-element (Earth, Fire, Wind, Water, Heart) state-of-the-world display of the Weather engine.
• Evolution — a real life-cycle milestone that advances a living thing to its next stage and unlocks that stage's real husbandry tasks (Livestock animals, Orchard trees, Bee queen).
• Pokédex — a registry of individuals in an engine (flock/herd registry with pedigree; hive; saved-seed library; found-trinket collection). Used generically across engines.
• Pollination Aura — the visible, measurable yield buff Bees project onto nearby Orchard and Beds.
• Cellar — the Cellar/Pantry engine's transformation view where preserved goods age on real timers.
• Pantry / Larder — the store of preserved and standing food; the tangible win-state, surfaced as days of food on hand.
• Closed Byproduct Loop (or Closed Loop) — the rule that nothing leaves the farm: outputs of one engine feed another (scraps → animals → manure → soil → trails).
• Staged Reveal — progressive disclosure of engines and features as the real homestead grows and as purchases unlock them.
• Micro-Win — the free, fast, near-guaranteed first-crop path that produces a first Harvest Log.
• Guided Harvest Success — the composite activation state defined in SM-1.
• Winter War-Room — the off-season planning mode (Report Card → Blueprint → Seed & Supply → Seed-Starting).
• Homestead Report Card — the November season-autopsy artifact opening the Winter War-Room.
• Failure Autopsy — the one-tap "why did this die?" flow that turns a death into a Signal the AI learns from.
• Local Zone Early-Warning Network (or Early-Warning Network) — anonymized microclimate pest/disease radar built from nearby users' Signals. Distinct from the deferred public marketplace.
• Self-Sufficiency Score — the live % of food/energy produced vs. bought (a resource ledger).
• Biodiversity Score — the ecosystem-health measure rewarding beneficials, pollinator strips, and hedgerows.
• Corvid Corps — the hidden, unlockable, free crow-ally feature; original characters only.
• Magic Feather — the Corvid Corps milestone gift symbolizing self-belief.
• Skill Level — the user's declared/derived beginner→advanced setting that scopes task detail and content.
• Ambition Level — the user's chosen scope target (e.g., keep it easy / expand / save money / kid-friendly / add chickens) that shapes the plan.

4. Features

Each subsection is a coherent feature: behavioral description, FRs nested and globally numbered, optional feature-specific NFRs/notes. P1 · SPINE marks the priority-1 build spine; untagged feature groups are P2 (still v1). Engines are costumes over one core mechanic — the shared core (§4.1–§4.3) is built once and reused.

4.0 Gamification Lineage (at a glance)

Every engine is the same core mechanic — "track a living thing's real life-cycle through evolutions, fed by real-world signals, where each thing buffs its neighbors" — wearing a different game/film costume. The costume contributes tone and mechanics only; never characters, names, art, or quotes (§11). This table is the quick reference; each engine's own section carries the detail and its copyright note.

Borrowed formats outside the engines: the hidden Corvid Corps (§4.19) reuses the Livestock/Bees/Weather patterns with an original musical-corvid corps + public-domain "magic feather" motif; Harvest Wrapped (FR-52) borrows the Spotify-Wrapped recap format; Strava-style segments (FR-53) borrow the Strava self/peer-comparison format. All tone/format only.

4.1 Digital Twin & Onboarding P1 · SPINE

Description The Twin is the foundation every engine renders on. Onboarding is deliberately minimal: location plus a light property capture seeds a personalized map with zone, frost dates, sun exposure, and a tailored starter plan — so the first thing the user sees is their yard, not a stranger's. The design bias is never block on perfect data: infer, show, let the user nudge. Realizes UJ-1.

Functional Requirements

FR-1 Seed the Digital Twin from location P1 · SPINE

A user can grant/enter a location and receive a Digital Twin pre-populated with USDA-equivalent zone, average first/last frost dates, day-length, and climate baseline. Realizes UJ-1. Consequences (testable) - Given a valid location, the Twin displays a zone, a last-frost and first-frost date, and a sun/climate baseline within one onboarding session. - Zone and frost dates are treated as distinct and both surfaced (zone = winter survival / what perennials live; frost dates = when to plant), using current USDA-equivalent data (2023 hardiness map + frost-date norms). Planting/timing keys off frost dates, not zone alone. - The user can override any inferred value (zone, frost dates) and the override persists. - Onboarding never hard-blocks on missing/denied GPS; manual location entry produces an equivalent Twin.

FR-2 Capture and edit property layout P1 · SPINE

A user can define their Property boundary and place/adjust Beds, zones, and structures on the Twin map. Realizes UJ-1, UJ-3. Consequences (testable) - The user can draw or adjust a Property boundary and drop at least one Bed pinned to a real location. - Every Bed/zone/structure persists with its location and is re-rendered on return. - Sun exposure can be set per Bed (full/partial/shade) and is used by planning (FR-6).

FR-3 Tailored starter plan on first run P1 · SPINE

A new user receives a starter plan scoped to their zone, sun, chosen Ambition Level, and Skill Level. Realizes UJ-1, UJ-2. Consequences (testable) - After onboarding, the user sees a concrete "start here" recommendation (≥1 Bed and specific crops) without further input. - The plan reflects the selected Ambition Level and Skill Level (a "keep it easy" beginner sees fewer, simpler crops than "expand"). - The plan routes a first-time free user into the Micro-Win path (FR-30).

FR-4 Soil-Life (Underground) View

A user can toggle a below-grid view modeling roots, worms, mycorrhizae, and compost maturity from what is planted/added. Realizes UJ-3. Consequences (testable) - The Soil-Life View renders modeled state per Bed derived from plantings and amendments (no hardware required). - Adding manure/compost (from the Closed Loop, FR-45) visibly changes the modeled soil state.

Feature-specific NFRs Onboarding to a populated Twin should complete in a single sitting on a mobile device with intermittent connectivity. Assumption target < 5 minutes to first populated Twin.

4.2 Real-World Signal Intake P1 · SPINE

Description The Signals are what make the Twin real and power the data-engine moat. Every input is designed for one muddy-thumbed tap. Photos are the hero input; over seasons each tile becomes a scrubbable time-lapse. Realizes UJ-3.

Functional Requirements

FR-5 One-tap Harvest Log P1 · SPINE

A user can log a harvest event (crop, optional weight, optional photo) attributed to a Bed/plant in one primary action. Realizes UJ-2, UJ-3. Consequences (testable) - A Harvest Log persists yield data (weight when given) against the specific Bed/plant and feeds the Report Card (FR-49) and Self-Sufficiency Score (FR-53). - Logging is completable in ≤3 taps from the Twin or Calendar.

FR-6 Photo Signal pinned to a tile P1 · SPINE

A user can attach a photo to any Bed/animal/tree/hive tile; photos accumulate into a scrubbable time-lapse. Realizes UJ-3, UJ-4. Consequences (testable) - A photo persists pinned to the correct tile and appears in that tile's chronological time-lapse. - A tile with ≥2 dated photos exposes a scrub/time-lapse control. - Photos are available to Vision AI (FR-8) without re-upload.

FR-7 One-tap Check-In and Voice-Note journaling P1 · SPINE

A user can record a Check-In (thriving/struggling/dead + note) and can journal by voice, transcribed into the tile's history. Realizes UJ-3. Consequences (testable) - A Check-In persists a status and timestamp on the target tile and contributes to the AI's history (FR-9) and Failure Autopsy (FR-51). - A voice note is transcribed to text, threaded into the tile history, and treated as a Signal. - A "dead" Check-In offers the Failure Autopsy flow (FR-51).

Feature-specific NFRs All Signal capture must function with degraded connectivity and sync when reconnected (see NFR-3). Voice/photo capture must be usable one-handed outdoors.

4.3 AI Brain — Vision AI & Agronomist P1 · SPINE

Description Two capabilities that turn Signals into personalized guidance. Vision AI reads photos; the AI Agronomist is grounded in the user's entire Twin so answers are about their plot. Copyright shield: the AI synthesizes functional concepts fresh and never parrots protected source prose. Realizes UJ-4, UJ-5.

Functional Requirements

FR-8 Vision AI diagnosis from photos P1 · SPINE

A user can submit a plant/animal/frame photo and receive a diagnosis (disease, pest, readiness/growth stage; livestock/bee-specific reads where applicable). Realizes UJ-4, UJ-8. Consequences (testable) - A submitted photo returns a structured diagnosis with a confidence indicator. - When confidence is low, the system states uncertainty, offers the top likely causes, and never presents a fabricated certain answer. - Diagnoses attach to the tile's history and can seed a task or Failure Autopsy.

FR-9 Grounded AI Agronomist P1 · SPINE

A user can ask free-form questions and receive answers grounded in their Twin (Beds, rotation history, harvest weights, zone, budget, Skill Level). Realizes UJ-4. Consequences (testable) - Answers reference the user's actual Twin context (e.g., the specific Bed/variety/zone) rather than generic advice when that context exists. - The Agronomist does not reproduce protected verbatim source text (copyright guardrail, §11). Its grounding/RAG corpus contains only our own-authored Knowledge Base content and functional data — never Seymour's (or any third party's) copyrighted prose (NFR-17), so it cannot retrieve and echo protected text. - The Agronomist can incorporate a prior Vision AI diagnosis and current weather into its recommendation.

FR-10 Detect slow multi-week drift P1 · SPINE

The AI Brain can surface gradual changes across a tile's photo/Check-In history that a user would miss week-to-week. Realizes UJ-3, UJ-4. Consequences (testable) - Given a series of tile photos/Check-ins trending negative, the system raises a proactive flag with a recommended action.

Feature-specific NFRs See NFR-4 (AI latency), NFR-5 (per-user AI cost ceiling), NFR-6 (AI safety/uncertainty), NFR-11 (photo/data privacy).

4.4 Weather-Aware Task Calendar P1 · SPINE

Description The Calendar is the engine of the guided loop — it tells the user what to do and when, scoped to Skill Level, and reorders itself against live weather. It is the surface Megan opens every visit. Realizes UJ-3, UJ-5.

Functional Requirements

FR-11 Skill-scoped task generation P1 · SPINE

The Calendar generates timed tasks (plant, thin, water, scout, harvest, protect) from the Twin, Knowledge Base timings, and the user's plan, scoped to Skill Level. Realizes UJ-2, UJ-3. Consequences (testable) - Tasks are attributed to specific Beds/plants/animals and carry a due window derived from zone/frost/life-cycle data. - A beginner Skill Level yields fewer, simpler, more-explained tasks than an advanced one for the same plan. - Tasks the user marks critical vs. optional are visually distinguished (critical tasks drive SM-1).

FR-12 Live weather ingestion and Task Shuffling P1 · SPINE

The Calendar ingests live weather and reorders/adjusts the day's tasks accordingly. Realizes UJ-3, UJ-5. Consequences (testable) - Incoming rain suppresses/defers watering tasks; a favorable window promotes weather-sensitive tasks (transplant). - Reordering is explained to the user in plain language ("rain coming — skip watering").

FR-13 Task completion feeds activation and history P1 · SPINE

A user can complete/snooze/skip tasks; completion of critical tasks feeds Guided Harvest Success (SM-1) and the Report Card. Realizes UJ-2, UJ-6. Consequences (testable) - Task state persists and on-time critical-task completion is measurable per user (input to SM-1). - Skipped/snoozed tasks are captured for the Report Card's "most-missed tasks" (FR-49).

Feature-specific NFRs Notification volume is governed by NFR-7 (anti-nag) and counter-metric SM-C2.

4.5 Beds Engine — the Flagship P1 · SPINE

Description Planting a Bed lays a glowing, plant-family-colored Rotation Trail; you cannot cross your own recent same-family trail (Self-Collision) because replanting a family in the same soil means disease buildup and nutrient crash. This is crop rotation made spatial and legible — a real guardrail caught with a seed packet in hand. Legume trails charge tiles with nitrogen; past trails fade to Ghost-Lines that layer into rotation tree-rings. Beds alone proves the Twin + rotation + real-signal loop, is the most universal engine, and is the most copyright-safe — hence the flagship. Metaphor costume: Tron light-cycle — the glowing, family-colored trails you lay across the plot and can't cross without a self-collision (tone and mechanic only; no Tron characters, names, or art). Realizes UJ-2, UJ-3.

Functional Requirements

FR-14 Rotation Trail on planting P1 · SPINE

Planting a crop in a Bed lays a plant-family-colored Rotation Trail on the Twin. Realizes UJ-3. Consequences (testable) - Each planting renders a Trail colored by plant family (e.g., nightshade/brassica/legume) using Knowledge Base family data (FR-27). - Trails persist across the season and are attributed to the Bed and planting date.

FR-15 Self-Collision guardrail P1 · SPINE

The system blocks or warns when a new planting would cross a recent same-family Rotation Trail. Realizes UJ-3. Consequences (testable) - Attempting a same-family planting within the rotation-risk window on a crossed tile triggers a Self-Collision warning explaining the rotation reason. - The user can override with an informed confirmation; the override is recorded. - The risk window is derived from Knowledge Base rotation data, not hard-coded per crop in the UI.

FR-16 Nitrogen charge from legumes P1 · SPINE

Legume Trails charge the tiles they cross with nitrogen, cueing heavy feeders next season. Realizes UJ-3, UJ-6. Consequences (testable) - Tiles crossed by a legume Trail show elevated nitrogen state in the Soil-Life View (FR-4) and influence next-season recommendations (FR-3/FR-11).

FR-17 Ghost-Lines and rotation tree-rings P1 · SPINE

Past-season Trails fade to toggleable Ghost-Lines, forming a layered rotation history. Realizes UJ-6. Consequences (testable) - The user can toggle prior-season Ghost-Lines on the Twin. - A Bed with ≥2 seasons shows a legible layered rotation history.

4.6 Knowledge Base — the Hidden Rules Engine P1 · SPINE

Description Reference data for every plant and animal, delivered just-in-time and in-context (never a manual to go read) — and secretly the rules engine that powers every other engine and grounds the AI. Self-Collision only works because it knows plant families; Evolution timings, frost-tenderness flags, spacing, gestation, and preservation times all live here. The substance is John Seymour's teaching — the methods and self-sufficiency system of The Self-Sufficient Life and How to Live It (crop rotation, planning by acreage, the productive garden, food from animals, preserving, the seasonal calendar) — re-expressed entirely in our own words. We teach what Seymour taught; we never use his exact wording, prose, or illustrations (§11). Built as our own database of these functional concepts (facts and methods — not protectable). Metaphor costume: none by design — the Knowledge Base is the un-gamified substrate every other engine's costume is painted over (its "game" is the almanac/rules that make the others' mechanics real). Realizes UJ-3, UJ-4, UJ-5.

Functional Requirements

FR-18 In-context reference data P1 · SPINE

A user can tap any crop/animal/tree to get its guide, and relevant lessons surface at the moment the season/task calls for them. Realizes UJ-3. Consequences (testable) - Every plantable/keepable entity exposes a reference card (needs, timing, spacing, hazards). - Lessons surface contextually (tied to a task/season/diagnosis), not as a standalone library the user must seek out. - Content faithfully conveys Seymour's methods and teaching (his self-sufficiency system and techniques) but is authored in original wording — a content-review step verifies no verbatim Seymour prose or reproduced illustrations ship (§11, NFR-16). - Seymour is the foundation, not the ceiling: the KB pairs his methods with modern researched content he lacks or gets dated on — beginner-friendly intensive/square-foot spacing, no-dig bed prep, Integrated Pest Management, and current food-safety standards — sourced from Cooperative Extension / NCHFP-equivalent authorities (see the KB Coverage Map, kb-coverage-map.md, §8–§9).

FR-19 Rules engine grounding for all engines P1 · SPINE

The Knowledge Base supplies the physics/rules other engines and the AI Brain consume (plant families, life-cycle timings, frost tolerance, spacing, gestation, preservation times). Realizes UJ-3, UJ-4, UJ-5. Consequences (testable) - Beds Self-Collision, Livestock Evolution timing, Weather frost-risk flagging, and Cellar food-safety times all resolve from Knowledge Base data (single source of truth). - The AI Agronomist retrieves from the Knowledge Base for grounded answers (FR-9).

FR-20 Almanac Layer

The Knowledge Base provides moon phase, day-length, and historical local weather woven into timing guidance. Realizes UJ-6. Consequences (testable) - Planting/timing guidance can reference almanac data (day-length/frost history) where relevant.

Feature-specific NFRs Content teaches Seymour's methods faithfully but is authored in original wording — no verbatim source text or reproduced illustrations. Produced clean-room (extract a functional coverage map first, then author fresh) and enforced by the content-provenance review (NFR-16). Seymour's copyrighted text is never placed in the AI grounding corpus (NFR-17); any source PDF is a planning reference only, never deployed. (§11)

4.7 Livestock Engine P2 — v1

Description Each animal evolves through real life-cycle milestones (age/weight/first egg/weaning); each Evolution unlocks that stage's real husbandry tasks. Branching evolutions are husbandry decisions (layer vs. meat; dairy vs. breeding). The Pokédex is the flock/herd registry with a real pedigree/genetics breeding engine. Metaphor costume: Pokémon GO — real life-cycle evolutions and a Pokédex registry (tone and mechanic only; our own creatures/characters, no Pokémon IP). Realizes UJ-7, UJ-8.

Functional Requirements

FR-21 Animal Evolution through life-cycle milestones

An animal advances through Evolution stages driven by real milestones, unlocking the real husbandry tasks for each stage. Realizes UJ-8. Consequences (testable) - Reaching a milestone (age/weight/first egg/weaning) advances the stage and generates the corresponding real tasks (vaccination, feed switch, breeding window, butcher date) via the Calendar. - Evolution timings resolve from the Knowledge Base (FR-19).

FR-22 Branching husbandry decisions

A user can choose an evolution branch (e.g., layer vs. meat; dairy vs. breeding) that sets the real task timeline and end-state. Realizes UJ-8. Consequences (testable) - Selecting a branch changes the generated task timeline and target end-state for that animal.

FR-23 Pokédex registry with pedigree/genetics

A user can maintain a flock/herd registry and plan pairings with an inbreeding-aware pedigree engine. Realizes UJ-8. Consequences (testable) - Each animal has a registry card; pairings surface relatedness/inbreeding warnings and track productive lines. - Weight/egg/milk logs and health Check-ins attach to the animal and feed Evolution and Vision AI.

4.8 Orchard Engine — the Slow Legacy Engine P2 — v1

Description Trees are decades-long characters. A harvest mini-loop fills an on-screen basket; real hazards (birds, late frost, coddling moth, squirrels) — fed by the Weather engine — knock fruit off the count. Each tree's mood is its real health and it speaks in character (AI). Slow multi-year Evolutions; grafting is a fusion mechanic; pollination-partner placement is a real puzzle on the Twin (and a Bee hook). Metaphor costume: Hi Ho Cherry-O (the basket-fill harvest mini-loop) + Wizard of Oz-style talking trees whose mood is their real health (tone and mechanic only — all tree characters are our own original creations; no Oz characters, names, or art). Realizes UJ-7.

Functional Requirements

FR-24 Tree Evolution and character health

A tree advances through multi-year Evolution stages (sapling → establishing → first fruit → mature → veteran); its displayed mood/character reflects its real health and needs. Realizes UJ-7. Consequences (testable) - Stage advances on real elapsed time/growth; the tree's character state maps to health signals (needs pruning/water/pest care vs. thriving) and can voice status via AI (in an original character, §11).

FR-25 Harvest mini-loop with weather-fed hazards

Picking real fruit fills a basket count; Weather-engine hazard events reduce it. Realizes UJ-5, UJ-7. Consequences (testable) - Logged fruit harvests fill the basket; active Weather Quests (frost/pest) apply the corresponding hazard to the count/yield.

FR-26 Grafting and pollination-partner placement

A user can graft varieties onto a rootstock and must place compatible pollination partners near a tree for it to fruit. Realizes UJ-7. Consequences (testable) - Grafting records multiple varieties on one tree; a tree lacking a required pollination partner is flagged with a placement recommendation on the Twin (hooks Bees, FR-30-series aura).

4.9 Bees Engine P2 — v1

Description The hive is a superorganism managed like a colony-sim on one dashboard (population, brood, honey stores, queen health) — not 60,000 cards. The queen's lifecycle is an Evolution; swarming is reproduction (capture a swarm into a new hive). Bees project a visible Pollination Aura that measurably buffs nearby Orchard and Beds; pull the hive and yields drop. Honey is a golden loot economy. A wise-cracking original bee-narrator reports status. Winter survival is a Weather quest. Metaphor costume: colony-sim management + Bee Movie tone only — the wise-cracking narrator's attitude and the "no bees → collapse" consequence arc (our own original bee character; no Bee Movie characters, names, or quotes). Realizes UJ-5, UJ-7.

Functional Requirements

FR-27 Colony dashboard and queen Evolution

A user manages each hive as one superorganism dashboard and the queen advances through an Evolution lifecycle. Realizes UJ-7. Consequences (testable) - The hive shows population, brood pattern, honey stores, and queen health as aggregate state (not per-bee entities). - Hive inspections, hive-weight logs, and frame photos (Vision AI: queen cells, varroa, laying pattern) attach to the hive. - Varroa mite management is first-class (it post-dates Seymour entirely): the app supports mite-count logging, an infestation-threshold alert (~3%), and surfaces IPM guidance (monitoring, drone-brood removal, treatment windows) — the single most important modern beekeeping task.

FR-28 Swarming reproduction

When a colony swarms, the user can capture the swarm into a new hive on the Twin. Realizes UJ-7. Consequences (testable) - A swarm event lets the user create a new hive pinned on the Twin; the apiary count grows.

FR-29 Pollination Aura buffing neighbors

Active hives project a measurable Pollination Aura that increases fruit set/yields for nearby Orchard and Beds; removing the hive removes the buff. Realizes UJ-7. Consequences (testable) - Beds/trees within a hive's range show a quantified yield/fruit-set buff driven by real nectar-flow/weather; removing or losing the hive visibly removes the buff (a real consequence lesson).

4.10 Weather & Seasons Engine P2 — v1

Description The gamified elemental layer over the Calendar's live-weather core (§4.4). A five-element Balance Meter (Earth, Fire, Wind, Water, Heart) turns weather into a state of the world; incoming conditions become Weather Quests ("villains") that flag the exact real Beds/animals/hives at risk and reward completing the real prep before the weather hits. Seasons are elemental chapters (winter = Heart = the strategy season → Winter War-Room). Metaphor costume: Captain Planet — the five-element balance and "the power is yours" villain-quest framing (tone and mechanic only; our own characters, no Captain Planet IP). Realizes UJ-5.

Functional Requirements

FR-30 Elemental Balance Meter

The Weather engine displays a five-element Balance Meter reflecting the homestead's current environmental state. Realizes UJ-5. Consequences (testable) - The meter reflects real conditions (soil moisture proxy, heat, wind, rain, steward activity) and updates with weather and user action.

FR-31 Weather Quests targeting real assets

Incoming weather raises a Weather Quest that names the specific real Beds/animals/hives at risk and the prep tasks to protect them. Realizes UJ-5. Consequences (testable) - A frost event flags exactly the frost-tender Beds (via Knowledge Base flags, FR-19) and vulnerable animals/hives, and generates protective tasks with a pre-weather deadline. - Completing the prep before the event resolves the quest as a "win"; missing it records the outcome for the Report Card.

FR-32 Seasonal chapters

The app frames the year as elemental seasonal chapters that shift the dominant activity (sow → heat management → harvest/storm prep → winter planning). Realizes UJ-6. Consequences (testable) - The active season changes the Calendar's emphasis and routes winter into the Winter War-Room (FR-48-series).

4.11 Cellar & Pantry Engine — Closes the Circle P2 — v1

Description Every raw output (Beds, Orchard, Livestock, Bees) flows into a crafting tree of processed goods (sauce, jam, pickles, cheese, cured meat, cider, mead, kraut). The Cellar runs real timers (mead months, kraut ~2 weeks, cheese cave-ages); the Pantry/Larder is the tangible win-state (a "Pokédex of preserves"). AI suggests what to make and enforces food-safety times — genuine, life-protecting value. Includes an Apothecary branch (functional herbal remedies). Metaphor costume: Stardew Valley crafting — the raw-output → processed-goods crafting tree and maturing cellar (tone and mechanic only; no Stardew art, names, or assets). Realizes UJ-6.

Functional Requirements

FR-33 Crafting tree from raw outputs

A user can transform raw harvest/animal/bee outputs into processed goods along a crafting tree. Realizes UJ-6. Consequences (testable) - Available recipes reflect the user's actual on-hand raw outputs (from Harvest Logs and engine yields).

FR-34 Real-time aging with timers

Ferments/brews/aging carry real-world timers; the Cellar view shows goods maturing over actual time. Realizes UJ-6. Consequences (testable) - A started ferment/brew/age tracks real elapsed time and notifies at readiness; the Cellar renders in-progress vs. ready.

FR-35 AI food-safety enforcement

The AI recommends what to make from a surplus and enforces food-safety method/time (water-bath vs. pressure-canning). Realizes UJ-6. Consequences (testable) - Given a surplus, the AI proposes recipes and jar counts and states the required safe method/time; unsafe combinations are flagged, not silently allowed (safety guardrail, §10). - All preservation times and temperatures come from current tested standards (USDA/NCHFP-equivalent), never from Seymour's dated 1970s figures — his methods may inform what to make, never how long/how hot (NFR-11).

FR-36 Pantry/Larder as win-state and Apothecary branch

The Pantry fills shelf-by-shelf and feeds the days-of-food Larder metric; an Apothecary branch supports functional herbal remedies. Realizes UJ-6. Consequences (testable) - Preserved goods accrue to the Pantry and contribute to the Larder days-of-food (FR-52) and Self-Sufficiency Score (FR-53). - The Apothecary offers functional remedy recipes (copyright-safe herbalism, §11).

4.12 Closed Byproduct Loop P2 — v1; cross-engine

Description The engines are not separate mini-games — they feed and buff each other. This is the product's soul, teaching method, and growth engine. It is grounded directly in Seymour's "High Farming" first principle — plants feed animals, animals' manure feeds the soil, the soil feeds the plants, and species are rotated so each gives back what it takes (the honest lineage of the closed loop). Realizes UJ-7.

Functional Requirements

FR-37 Cross-engine byproduct flows

Outputs of one engine become inputs to another (Livestock manure → soil/Beds; Cellar byproducts → animal feed/compost; kitchen scraps → Livestock feed). Realizes UJ-7. Consequences (testable) - Applying an animal/compost byproduct to a Bed changes that Bed's Soil-Life state (FR-4) and next-season recommendation. - The system surfaces available byproduct flows the user is not yet using (basis for the CTA, FR-58). - Kitchen scraps route to modern composting paths appropriate to the user's setup — hot composting, vermicomposting, or bokashi — even before the user has Livestock, so the scraps→soil loop works for a beginner without animals.

FR-38 Full-circle visualization

The user can see the closed loop (seed → plant → harvest → preserve → pantry → survive winter → byproducts → soil → next season). Realizes UJ-7. Consequences (testable) - A view represents how the user's active engines connect and where the loop is currently broken/incomplete.

4.13 Learning Memory & Retention P2 — v1

Description Three features that form one system that learns: Failure Autopsy (from your mistakes), season-over-season tree-rings (from your history), and the Early-Warning Network (from your neighbors). Core of off-season retention and the data-engine moat. Realizes UJ-4, UJ-6.

Functional Requirements

FR-39 Failure Autopsy

When a plant/animal dies, a one-tap "why?" flow captures cause and feeds the AI to warn the user next season. Realizes UJ-2, UJ-6. Consequences (testable) - A "dead" Check-In (FR-7) offers an Autopsy; the recorded cause becomes a Signal that influences future warnings/recommendations and the Report Card. - A Micro-Win failure Autopsy reframes the loss as data and offers a fast retry (never triggers the paywall) (UJ-2 edge case).

FR-40 Local Zone Early-Warning Network

A user receives anonymized microclimate alerts derived from nearby users' Signals (e.g., a regional blight/pest radar). Realizes UJ-4. Consequences (testable) - When enough nearby anonymized Signals indicate an emerging threat, the user gets a scout-now alert scoped to their zone. - Contributed data is anonymized and governed by NFR-11/§12; participation is consent-based.

FR-41 Season-over-season memory (tree-rings)

The system retains multi-season history per tile/engine and exposes it as comparative memory. Realizes UJ-6. Consequences (testable) - Prior seasons are retrievable per Bed/tree/animal (photos, yields, outcomes) and feed the Report Card and next-season plan.

4.14 Winter War-Room — Off-Season Retention P2 — v1

Description Retention is only real if the app has an urgent winter job: make sure next spring isn't another chaotic false start. Powered by the user's own season data (generic content can't retain her). Monthly cadence Nov–Feb. Realizes UJ-6. This is a core moat, not a feature afterthought.

Functional Requirements

FR-42 Homestead Report Card (November)

The user receives a season-autopsy summarizing attempts vs. harvests, best performer, biggest failure, money spent vs. estimated food value, most-missed tasks, and top-3 changes. Realizes UJ-6. Consequences (testable) - The Report Card is generated from the season's Signals and task history and works even with thin first-season data (leans on Check-ins/photos). - Framing is confidence-restoring ("you learned what your yard was telling you").

FR-43 Next-Season Blueprint (December)

The user picks an Ambition Level and receives a realistic spring plan. Realizes UJ-6. Consequences (testable) - Selecting an ambition (keep it easy / expand / save money / kid-friendly / add chickens / one more bed) generates a concrete plan carried into the Calendar for next season.

FR-44 Seed & Supply Decisions (January) and Seed-Starting (February)

The user gets an exact seed/supply list distinguishing what they already own, what to buy, and what not to buy — then a seed-starting plan scoped to the equipment and budget they actually have, with dated tasks, a germination tracker, and a hardening-off timeline. Realizes UJ-2, UJ-6. Consequences (testable) - The list reflects inventory the app knows the user owns and flags "don't buy yet" items, producing a lower, itemized cost. - Seed-starting never assumes special equipment. For each crop the app recommends the lowest-friction viable path for that user: buy transplants/starts, direct-sow outdoors, winter-sow (no-electricity milk-jug method), or a bright windowsill for suitable crops — and only recommends indoor grow-lights when the user already owns them or explicitly opts in. Assumption onboarding/plan captures the user's available gear + budget so this scoping is possible. - When indoor light-starting is chosen, the guidance includes the discipline that prevents the classic beginner failure — lights kept 2–4 in above seedlings, 14–16 h/day, airflow — plus the hardening-off timeline. (Directly addresses Megan's leggy-seedling failure without requiring her to buy a light.) - A backup plan (a direct-sow or buy-a-start alternative) is offered if seeds fail or the window is missed, so a setback never ends the season.

4.15 Larder, Self-Sufficiency & Biodiversity — the Win-State Dashboards P2 — v1

Description The north-star win-state, measured three ways: flourishing homestead (feeling), Larder days-of-food (security), Self-Sufficiency Score (%). Plus a Biodiversity Score rewarding whole-system health. Realizes UJ-6.

Functional Requirements

FR-45 Larder / Food-Security Dashboard

The user sees "days/weeks of food on hand" from Pantry stores plus standing crops. Realizes UJ-6. Consequences (testable) - The Larder computes a days-of-food figure from Pantry (FR-36) and standing/harvestable crops and updates as stores change.

FR-46 Self-Sufficiency Score

The user sees a live % of food/energy produced vs. bought via a resource ledger. Realizes UJ-6. Consequences (testable) - The Score aggregates produced outputs vs. logged inputs (water/feed/energy where available) into a single percentage that trends over time.

FR-47 Biodiversity / Ecosystem-Health Score

The user sees a score rewarding beneficials, pollinator strips, hedgerows, and diversity. Realizes UJ-9. Consequences (testable) - Adding qualifying beneficial habitat/diversity raises the Biodiversity Score; it reflects whole-system health, not just yield.

4.16 AR / GPS Layer P2 — v1; mobile-only

Description Walk the land; geo-fenced zones trigger Check-ins (stand at the coop → flock cards; at bed 4 → that tile). AR overlays trails/animal cards onto the real dirt. Mobile-only by hardware nature; the web companion degrades gracefully without it. Realizes UJ-3.

Functional Requirements

FR-48 Geo-fenced check-in triggers

Standing in a real zone surfaces that zone's tiles/cards for quick Check-ins. Realizes UJ-3. Consequences (testable) - On mobile with location enabled, entering a Bed/coop/hive geo-fence surfaces the relevant tiles for one-tap Check-in. - The feature is absent-but-non-blocking on web and when location is denied.

FR-49 AR overlay of the Twin

A user can view Rotation Trails/animal cards overlaid on the real property through the camera. Realizes UJ-3. Consequences (testable) - AR mode renders the relevant Twin elements aligned to the real location on a supported mobile device.

4.17 Additional Homestead Features P2 — v1

Description The remaining committed v1 features that deepen the living system and the sharable delight. Grouped here for brevity; each is a real v1 FR.

Functional Requirements

FR-50 Seed-Saving Engine and seed library

A user can track what to let go to seed and build a personal saved-seed Pokédex. Realizes UJ-6. Consequences (testable) A saved variety persists in the seed library and can seed next season's plan; harvesting-for-seed generates the relevant tasks.

FR-51 Foraging / Wild Layer

A user can map wild edibles/medicinals on the Property with seasonal foraging prompts. Realizes UJ-9. Consequences (testable) Foraged items pin to the Twin; seasonal prompts surface when a mapped wild edible is in season. Foraging content carries positive-identification guidance and explicit "never eat an uncertain ID / avoid toxic look-alikes" safety framing (a safety requirement, §9.1); the app never asserts an edible ID with unwarranted certainty.

FR-52 Harvest Wrapped

At first frost, the user gets a shareable end-of-season recap of their year — a Spotify-Wrapped-style retrospective (borrowed format only). Realizes UJ-6. Consequences (testable) A recap is generated from the season's Signals and is shareable as an image/card.

FR-53 Strava-style segments

A user can compare their current season against their own past seasons and, anonymized, against nearby zone peers. Realizes UJ-9. Consequences (testable) Segment comparisons render against the user's history and anonymized zone cohort; peer comparison is opt-in and anonymized (§12).

FR-54 Legacy / Mentor Export

A user can package their proven, plot-specific playbook to hand to another person as a starting template. Realizes UJ-6. Consequences (testable) Export produces a shareable playbook derived from the user's Twin; import seeds a new user's plan (does not expose raw private data beyond what the owner chooses).

4.18 Onboarding Staged Reveal, Micro-Win & Monetization P1 · SPINE

Description The full seven-engine world is present but disclosed gradually so it never overwhelms Megan on day one. The beginner's front door is Beds → Calendar → first harvest via the Micro-Win. Engines unlock as the real homestead grows and as purchases gate them, surfaced through closed-loop CTAs. Monetization converts a burned skeptic: free micro-win → pay-after-proof at the first-harvest peak → à-la-carte engines → all-access no-brainer. Realizes UJ-2, UJ-7, UJ-8.

Functional Requirements

FR-55 Guided Micro-Win path P1 · SPINE

A free user is routed to a fast, near-guaranteed first crop that produces a first Harvest Log. Realizes UJ-2. Consequences (testable) - The free tier delivers the Micro-Win crop path plus a "Homestead Snapshot" preview of the Twin. - The path is completable by a beginner and culminates in a first Harvest Log (SM-1 trigger).

FR-56 Staged Reveal / progressive disclosure P1 · SPINE

Engines and features disclose progressively based on the user's real homestead state and entry point. Realizes UJ-1, UJ-8. Consequences (testable) - A brand-new beginner sees only Beds + Calendar + first-harvest surfaces initially; other engines are not shown week one. - An active homesteader (secondary persona) enters deeper (e.g., Beds + Livestock live).

FR-57 Pay-after-proof conversion at first harvest

The Gardening-tier upgrade prompt fires at the first-harvest moment, not on faith or in the off-season. Realizes UJ-2. Consequences (testable) - The conversion prompt is triggered by the first real Harvest Log event; it does not fire during the Micro-Win failure/autopsy path or the low-motivation off-season.

FR-58 Closed-loop upsell CTAs

Locked engines appear as friendly gamified "your homestead wants to grow" prompts showing how the missing engine interconnects with what the user has. Realizes UJ-7. Consequences (testable) - A CTA is contextually triggered by a real unused byproduct flow (e.g., unlogged kitchen scraps) and frames the upsell as "complete your living system." - Declining a CTA suppresses further nagging for that engine (anti-nag, NFR-7).

FR-59 Pricing tiers and à-la-carte engine gating

The system supports Free, Gardening tier, à-la-carte engines, and all-access, gating engine access accordingly. Realizes UJ-7. Consequences (testable) - Purchasing an à-la-carte engine unlocks exactly that engine; all-access unlocks all engines. - Tier/entitlement state governs which engines are usable; the Beds spine and Micro-Win remain in the free/gardening experience. Assumption exact price points from the brief ($24 Gardening / $12 per engine / $60 all-access / $39 founding) are indicative and validated post-PRD (OQ-2).

FR-60 Trust-first affiliate supply lists

Seed/soil/supply recommendations are framed as "the smallest list you need — don't buy the rest," never as a marketplace. Realizes UJ-6. Consequences (testable) - Affiliate suggestions are minimal-by-default and appear within planning/seed-list flows; they never gate core functionality and are counterbalanced by SM-C3.

4.19 The Corvid Corps — Hidden Easter Egg v1, in scope; free / never paywalled

Description A discoverable, unlockable secret — found, not sold; not advertised in the UI or store listing. The user can secretly befriend and "train" backyard crows into a homestead ally corps via a real-corvid-behavior trust ladder, unlocking a living pest-defense aura, a found-object Pokédex of the real shiny trinkets the crows leave (photographed by the user), and a Magic Feather self-belief moment that maps onto Megan's arc. Architecturally it is not an eighth engine — it reuses the Livestock evolution/Pokédex pattern, the Bees passive-aura pattern, the Weather/hazard defense pattern, and the animal-narrator character pattern, built on the shared core. Realizes UJ-9.

Functional Requirements

FR-61 Hidden discovery triggers

The Corvid Corps is discoverable only through grounded real-behavior triggers, never advertised. Realizes UJ-9. Consequences (testable) - Repeated same-spot scrap offerings, placing a shiny object on a fence post, or Vision AI noticing recurring crows in bed photos can each fire the rare discovery event. - The feature does not appear in onboarding, menus-by-default, or store listing copy.

FR-62 Trust-ladder progression and defense aura

A user advances crows through a trust ladder earned by consistency (never cash), unlocking a defense aura across engines and gifted trinkets. Realizes UJ-9. Consequences (testable) - Progression advances only via consistent offerings/behavior; no purchase can advance it (never paywalled). - A sufficiently trained Corps applies a pest-defense benefit (mob hawks threatening Livestock, harass Orchard squirrels, eat Beds grubs, sentinel early-warning ping); as trust deepens, the crows begin leaving real trinkets the user collects (FR-64). - Each crow is an original character/card with a name/personality (§11); the honest "redirect via decoy station, not cheat" trade-off is represented.

FR-63 Magic Feather milestone

At a milestone (e.g., first successful harvest), a crow gifts the Magic Feather self-belief moment. Realizes UJ-9. Consequences (testable) - The Magic Feather event fires at the defined milestone and is shareable/screenshot-worthy; content uses only original characters and the public-domain self-belief motif (§11 hard guardrail).

FR-64 Found-Object Pokédex — photograph what the crows bring

Once the Corvid Corps is discovered, a user can photograph the real shiny trinkets the crows leave and collect them in a Pokédex-like gallery. Realizes UJ-9. Consequences (testable) - After discovery, a user can capture a photo of a real found object (via the Photo Signal, FR-6) and file it into a dedicated Corvid Corps found-object collection. - Each collected object persists as a named card in a screenshot-worthy, shareable gallery (the "Pokédex of found objects") that grows as trust deepens — this is the delight currency and the viral/word-of-mouth hook. - The collection is free / unlockable through engagement, never paywalled (§4.19 posture); it is not advertised — it exists only after the easter egg is found.

Notes Note for PM Corvid Corps is a retention/word-of-mouth/brand-love driver, explicitly not a revenue line. Guard its "found, not sold" nature against growth pressure to surface it in onboarding.

5. Non-Goals (Explicit)

• Not a content library or social feed. Sproutopia is a doing/planning system tied to a real plot; it does not compete on volume of videos/articles. Its promise is subtraction.
• Not commercial/farm-scale management. No field-scale equipment, compliance/traceability records, or ERP herd management.
• Not a public social marketplace in v1. No buying/selling between users, public listings, or take-rate marketplace. (Anonymized Early-Warning Network and Legacy Export are not a marketplace.)
• Not a hardware/sensor product in v1. No soil-sensor hardware integration; the Soil-Life View is modeled, not sensed.
• Not an off-grid energy/water/structural planner. Self-Sufficiency touches energy inputs but v1 is food-first.
• Not standalone child accounts. No independent minor accounts (keeps COPPA out of v1).
• Not a points-for-points game. Every mechanic must map to a real husbandry/gardening action or outcome; a "win" is a real flourishing homestead.

6. MVP Scope

Reminder: "MVP" here means the committed v1. v1 ships the full closed-loop world with a staged reveal; the P1 · SPINE tags mark what must land first, not what is "in vs. out."

6.1 In Scope (v1)

• P1 spine (ship first): Digital Twin + onboarding (§4.1), Real-World Signal intake (§4.2), AI Brain — Vision + Agronomist (§4.3), Weather-Aware Task Calendar (§4.4), Beds flagship (§4.5), Knowledge Base rules engine (§4.6), Guided Micro-Win + Staged Reveal path (§4.18 spine portions).
• P2 (still v1, sequenced behind the spine): Livestock, Orchard, Bees, Weather/Seasons elemental layer, Cellar/Pantry, Closed Byproduct Loop, Learning Memory & Retention, Winter War-Room, Larder/Self-Sufficiency/Biodiversity dashboards, AR/GPS, all additional features (§4.17), full monetization ladder (§4.18), and the Corvid Corps (§4.19).
• Platforms: mobile-first (iOS + Android) as the primary experience; web as the planning/Winter War-Room companion that degrades gracefully without camera/AR/GPS.

6.2 Out of Scope for MVP

• Soil-sensor hardware integration — deferred to a later version (Soil-Life View stays modeled). Note for PM emotionally load-bearing for the "living soil" promise; revisit for v2.
• Public / social marketplace — deferred. Local Family Co-op (household chore quests, shared Pokédex, assigning beds/animals to kids) is near-term roadmap, not v1 — its multi-user/minor-participation design pulls in COPPA and household-account complexity. Note for PM Family Co-op is a strong retention/virality lever; first fast-follow candidate after v1.
• Standalone child accounts / independent minor use — out (tied to the Co-op deferral).
• Non-food self-sufficiency planning (energy/water/structural) beyond the Self-Sufficiency ledger inputs.

7. Success Metrics

Each SM cross-references the FR(s) it validates. Targets are Assumption placeholders to validate (OQ-2/OQ-3). Task completion is a leading indicator only — never the win.

Primary - SM-1 — Guided Harvest Success Rate: % of first-season users who (1) create a personalized plan, (2) plant ≥3 recommended crops, (3) complete ≥70% of critical tasks on time, (4) log ≥3 harvest events, and (5) harvest from ≥2 different crops. Target ≥40% Assumption . Validates FR-1..FR-13, FR-55. This is the product promise, measured.

Secondary - SM-2 — Crop-type harvest rate (quality signal): ≥60% Assumption of planted crop types reach harvest stage. Validates FR-11, FR-14, FR-8/FR-9. - SM-3 — Off-season retention: % of first-season users who create next season's plan before spring (open ≥2×/month Nov–Feb, complete ≥1 critical off-season task/month, plan by January, seed/supply list by February). Target ≥30% Assumption . Validates FR-42..FR-44. - SM-4 — Identity-change (30-day): % of season-end users who within 30 days plan next season, add a new crop/bed/system, or state intent to grow again. Target ≥50% of harvesters Assumption . Validates FR-39, FR-42, FR-43. - SM-5 — Pay-after-proof conversion: free→Gardening conversion at the first-harvest moment. Target ≥8% of activated free users Assumption . Validates FR-55, FR-57. - SM-6 — Annual retention & engine attach: year-two retention (target ≥45% Assumption , validates FR-42..FR-44, FR-58) and à-la-carte/all-access attach among paying users (target ≥20% Assumption , validates FR-58, FR-59).

Counter-metrics (do not optimize) - SM-C1 — Task-completion-as-vanity: raw task-completion rate must not be optimized in place of SM-1; a rising task rate with flat harvest logging is a red flag. Counterbalances SM-1. - SM-C2 — Notification load / opt-out: push volume per active user and notification opt-out rate must stay low; do not drive engagement by nagging. Counterbalances SM-1/SM-6. (Enforced by NFR-7.) - SM-C3 — Affiliate revenue per user: must stay low/bounded; optimizing affiliate take erodes the trust-first promise (FR-60). Counterbalances any monetization metric. - SM-C4 — AI cost per active user: watch as a margin guard against over-calling the AI Brain; do not degrade AI quality to chase it below the trust threshold. Counterbalances SM-5/SM-6 (tension with NFR-5).

8. Cross-Cutting Non-Functional Requirements

System-wide quality attributes not tied to a single feature.

• NFR-1 — Mobile-first responsiveness: the in-yard loop (Twin, Signals, Calendar, Beds) must be fully usable one-handed on a phone outdoors. Web provides planning/Winter War-Room parity minus camera/AR/GPS.
• NFR-2 — One-tap capture: every Signal (harvest, photo, check-in, voice) must be completable in ≤3 taps from the primary surfaces.
• NFR-3 — Offline-tolerant capture & sync: Signal capture must work with degraded/no connectivity and reliably sync when reconnected without data loss. Assumption photos/check-ins queue locally and sync opportunistically.
• NFR-4 — AI latency: Vision AI diagnosis and Agronomist answers should return fast enough to keep the in-yard loop fluid. Assumption target < 10s typical for a diagnosis; validate.
• NFR-5 — AI cost ceiling per user: the AI Brain must operate within a per-active-user cost envelope that protects margin (paired with SM-C4). Assumption target envelope TBD in OQ-4.
• NFR-6 — AI honesty/uncertainty: the AI Brain must express confidence, refuse to fabricate certainty, and degrade to safe generic guidance when ungrounded (FR-8/FR-9).
• NFR-7 — Anti-nag notifications: notifications must be weather/task-justified, respect declined CTAs, and stay within a per-user volume ceiling (SM-C2).
• NFR-8 — Data durability / season memory: multi-season history (photos, yields, outcomes) must be durably retained — the moat depends on year-over-year memory (FR-41).
• NFR-9 — Accessibility: the app should meet mainstream mobile accessibility expectations (contrast, dynamic type, screen-reader labels on core flows). Assumption WCAG 2.1 AA as the target where applicable.
• NFR-10 — Graceful capability degradation: features requiring camera/AR/GPS must be absent-but-non-blocking where unavailable (web, denied permissions).

9. Constraints & Guardrails

9.1 Safety

• NFR-11 — Food-safety enforcement (life-critical): Cellar preservation guidance (FR-35) must enforce correct method/time (water-bath vs. pressure-canning); unsafe combinations are flagged and never silently permitted. Incorrect canning guidance can cause botulism — this is a hard safety requirement, not a nicety. All preservation times/temperatures derive from current tested standards (USDA/NCHFP-equivalent); Seymour's (or any legacy) preservation figures are never used — his teaching informs method and intent, never the safety-critical numbers.
• Apothecary caution: functional herbal remedy content (FR-36) must carry appropriate not-medical-advice framing. Note for PM legal review of Apothecary and food-safety copy before launch.

9.2 Privacy & Data Governance

• NFR-12 — Location & imagery privacy: the Twin holds precise home location and yard photos — sensitive data requiring encryption at rest/in transit, clear consent, and user deletion/export rights. Assumption GDPR/CCPA-aligned data handling.
• NFR-13 — Anonymized network integrity: the Early-Warning Network (FR-40) and Strava segments (FR-53) must be consent-based and genuinely anonymized/aggregated so no individual's plot/location is re-identifiable. Distinct from the deferred public marketplace.
• NFR-14 — No standalone minors: no independent child accounts in v1 (keeps COPPA out of scope; revisit with Family Co-op).

9.3 Cost

• NFR-15 — AI margin discipline: per-user AI cost (NFR-5/SM-C4) must be managed so the low entry price ($24 Gardening indicative) remains viable at scale.

10. Aesthetic, Tone & Characters

• Tone: a real-life homestead tool wearing a game's clothes — warm, encouraging, confidence-restoring; never childish-for-its-own-sake, never points-chasing. Every mechanic maps to a real action/outcome.
• Characters: all in-app characters (the bee narrator, the talking Orchard trees, the Corvid Corps and its sentinel) are our own original creations — original designs, names, and voices.
• Metaphor costumes (Tron, Pokémon, Captain Planet, Bee Movie, Wizard of Oz, Hi Ho Cherry-O, Stardew) contribute tone and mechanics only — never their characters, names, art, or quotes.
• Emotional through-line: subtraction over content; fight abandonment; celebrate the first win; make winter alive. The Magic Feather (self-belief) is the emotional spine.

11. Copyright Posture (applies to all mechanics — hard requirement)

• Seymour is the source of the teaching, not the wording. The Knowledge Base (§4.6) deliberately teaches what John Seymour taught — his self-sufficiency methods and system from The Self-Sufficient Life and How to Live It (crop rotation, planning by acreage, seasonal calendars, life-cycle timings, preservation methods — functional concepts, facts, not protectable). It must never reproduce his verbatim prose, exact wording, or illustrations; every piece is re-expressed in our own words. The AI Agronomist (FR-9) likewise synthesizes these functional concepts fresh.
• NFR-16 — Content-provenance review (hard requirement): before launch, all Knowledge Base and AI-authored content passes a review confirming it conveys Seymour's methods faithfully while containing no verbatim Seymour text or reproduced illustrations. Note for PM include legal/editorial sign-off on KB provenance in the pre-launch gate (with the food-safety and Apothecary reviews).
• NFR-17 — Grounding-corpus exclusion (hard requirement): the AI Brain's retrieval/RAG grounding corpus contains only our own-authored Knowledge Base content and functional data. Seymour's (or any third party's) copyrighted prose and illustrations are never ingested into the corpus, embeddings, prompts, or fine-tuning data — this is what structurally prevents the AI Agronomist (FR-9) from reproducing protected text. Any copyrighted source used to inform scope is handled as reference-only, outside the product corpus.
• Clean-room authoring (process requirement): Knowledge Base content is produced clean-room — a factual coverage map of functional concepts/methods (facts, not expression) is extracted first, then content is authored fresh in our own words from that map. Source materials (including any Seymour PDF used to build the coverage map) are planning references only: kept out of site/public/, out of the deployed product, and out of the grounding corpus (NFR-17).
• Borrowed game/film references contribute tone/mechanics only — no characters, names, art, or quotes.
• Corvid Corps hard guardrail: entirely original characters/designs/names/voices. Do NOT base them on Disney's Dumbo crows (Disney-owned and a recognized racist caricature — the lead was named "Jim Crow"). Borrow only the un-protectable generic archetype of a charismatic musical corvid corps and the public-domain self-belief / "magic feather" theme.

12. Information Architecture & Staged Reveal

• Primary surfaces (mobile): the Twin (home/map), the Calendar (what to do now), a Capture action (Signals), the AI (ask/diagnose), and Engine views revealed progressively.
• Web companion: planning-centric — Twin editing, Calendar, Winter War-Room, Report Card, seed/supply lists; camera/AR/GPS features absent-but-non-blocking.
• Staged Reveal (FR-56) governs which surfaces appear: beginners start at Twin + Calendar + Beds + Micro-Win; engines unlock by real-homestead state, entry point, and entitlement (FR-59).

13. Platform

• v1 primary: iOS + Android (Expo) — full experience incl. camera/Vision AI, AR/GPS, geo-fenced check-ins, in-yard tending loop.
• v1 companion: web — planning and Winter War-Room; graceful degradation of camera/AR/GPS (NFR-10).
• Technology specifics (Expo/Expo Router, AI vendor, RAG grounding, storage/sync) live in addendum.md, not this PRD.

14. Monetization Summary

(Detail in §4.18; prices indicative pending OQ-2.) Free micro-win + Homestead Snapshot → Pay-after-proof Gardening tier (~$24/yr) firing at the first-harvest peak → à-la-carte engines (~$12/yr each) surfaced by closed-loop CTAs → all-access "Whole Homestead" (~$60/yr; founding $39) as the third-engine no-brainer. Affiliate supply revenue is a secondary, trust-first stream (FR-60, SM-C3), never a marketplace take-rate.

15. Open Questions

1. OQ-1 — Name / trademark: sproutopia.com is taken; an indie farming game "Sproutopia" (~2024) exists in an adjacent category; sproutopia.app is available. Run a USPTO trademark search + app-store name check before branding spend. May force a rename (brainstorm candidates: Plot, The Steading, Croft, Verdance, Rooted, Larder).
2. OQ-2 — Pricing validation: all price points ($24 / $12 / $60 / $39 founding) and the SM-5/SM-6 targets are indicative and need real willingness-to-pay testing.
3. OQ-3 — Activation-metric targets: SM-1..SM-4 placeholder targets need calibration against beta cohort data.
4. OQ-4 — AI quality bar & cost: validate Vision-AI diagnosis accuracy (trust) and per-user AI cost/latency envelope (NFR-4/NFR-5) before scale.
5. OQ-5 — À-la-carte vs. connected-system tension: monitor whether single-engine buyers still feel the closed loop; the gamified CTAs (FR-58) are the mitigation.
6. OQ-6 — Early-Warning Network cold-start: the anonymized microclimate radar (FR-40) needs local user density to be useful; what's the minimum viable density and the pre-density fallback?
7. OQ-7 — Seven-engine execution risk: the full-v1 scope is ambitious; validate the P1-spine-first sequence delivers a shippable, valuable core even if P2 engines slip.

16. Assumptions Index

Every Assumption surfaced for confirmation: - §2.1/§2.3 — Secondary "active homesteader" persona enters the Staged Reveal deeper; no standalone child accounts in v1. - §3 / FR — Single active Property per account in v1. - §4.1 FR-4 — Soil-Life View is modeled from plantings/amendments (no hardware) in v1. - §4.1 NFR — Target < 5 minutes to first populated Twin. - §4.14 FR-44 — Onboarding/plan captures the user's available gear + budget so seed-starting can be scoped to what they actually have (never assumes grow lights). - §4.18 FR-59 — Brief price points are indicative, validated post-PRD (OQ-2). - §7 — All SM targets (SM-1 ≥40%, SM-2 ≥60%, SM-3 ≥30%, SM-4 ≥50%, SM-5 ≥8%, SM-6 ≥45%/≥20%) are placeholders (OQ-3). - §8 — NFR-3 offline queue/sync model; NFR-4 <10s AI latency; NFR-5 AI cost envelope TBD (OQ-4); NFR-9 WCAG 2.1 AA target. - §9.2 — GDPR/CCPA-aligned data handling for location/imagery.