diff --git a/.gitignore b/.gitignore
index 47f7917..b5da942 100644
--- a/.gitignore
+++ b/.gitignore
@@ -16,6 +16,7 @@ Thumbs.db
 # IDE
 .idea/
 .vscode/
+.codex/
 *.swp
 *.swo
 
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 077e50a..261fa7f 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -1,6 +1,6 @@
 # Contributing to OpenPasture
 
-Thanks for helping improve OpenPasture! This project focuses on the Morning Farm Brief experience and adaptive grazing intelligence.
+Thanks for helping improve OpenPasture! See [docs/vision.md](docs/vision.md) for the product thesis before diving in.
 
 ## Table of Contents
 
@@ -42,13 +42,13 @@ npm install
 
 ### 2. Configure Environment
 
-Copy the example environment file:
+Copy the app example environment file:
 
 ```bash
-cp .env.example .env.local
+cp app/.env.example app/.env.local
 ```
 
-Edit `.env.local` with your values:
+Edit `app/.env.local` with your values:
 
 ```bash
 # Convex - Get from your Convex dashboard
@@ -62,6 +62,14 @@ VITE_CLERK_PUBLISHABLE_KEY=pk_test_your_key_here
 VITE_DEV_AUTH=true
 ```
 
+If you work on ingestion/pipeline code, also copy:
+
+```bash
+cp src/ingestion/.env.example src/ingestion/.env.local
+```
+
+For the complete env matrix (including Convex Dashboard server-side vars like `ANTHROPIC_API_KEY`), see [docs/environment.md](docs/environment.md).
+
 ### 3. Set Up Convex
 
 If you haven't already, create a Convex project:
diff --git a/README.md b/README.md
index 46ce78e..0301f86 100644
--- a/README.md
+++ b/README.md
@@ -1,26 +1,6 @@
 # OpenPasture
 
-An open-source intelligence and orchestration layer for adaptive grazing. This system translates satellite-derived land signals into clear daily grazing decisions.
-
-## Overview
-
-Grass is dynamic. Fences are static. Modern sensing allows us to understand land conditions daily - what's been missing is software that turns those signals into actionable guidance.
-
-This project provides the **decision-making brain** that tells farmers:
-
-> "Here is the current state of your land, and here is the best grazing action for today."
-
-## The Morning Farm Brief
-
-The core experience is a daily ritual:
-
-1. **Open the app** each morning
-2. **Receive a status report** on your farm's pasture conditions
-3. **Review the recommended grazing plan** for today
-4. **Approve or adjust** using natural language feedback
-5. **Get execution instructions** ready for your virtual fencing system or manual workflow
-
-This is decision support - you stay in control.
+An open-source intelligence layer for adaptive grazing. See [docs/vision.md](docs/vision.md) for the thesis and product philosophy.
 
 ## Features
 
@@ -57,8 +37,6 @@ We use publicly available satellite imagery (primarily Sentinel-2) to compute ve
 - Frequent revisits (daily to every few days)
 - Historical archive for trend analysis
 
-Satellite cadence aligns with biological reality - grazing decisions are daily or multi-day decisions, not minute-by-minute.
-
 ### Vegetation Indices
 
 | Index | Purpose |
@@ -77,17 +55,6 @@ Satellite cadence aligns with biological reality - grazing decisions are daily o
 6. Score pasture readiness using transparent rules
 7. Generate recommendation with confidence level
 
-## What This Is Not
-
-To maintain focus, this demo does not include:
-
-- Real-time livestock tracking
-- Custom hardware or collars
-- Automated fence actuation
-- Drone-based sensing
-
-These are future integration points, not current scope.
-
 ## Getting Started
 
 The web app lives in `app/`. Prerequisites: Node.js 18+ and npm.
@@ -107,11 +74,22 @@ Common scripts:
 
 ### Environment Variables
 
-The following environment variables are required for full functionality:
+Environment setup is documented in [docs/environment.md](docs/environment.md).
+
+Quick start:
+
+```bash
+cp app/.env.example app/.env.local
+cp src/ingestion/.env.example src/ingestion/.env.local
+```
 
-- `BRAINTRUST_API_KEY` - API key for Braintrust observability platform. Required for agent behavior tracking and debugging. Get your API key from [Braintrust](https://www.braintrust.dev). Set this in your Convex dashboard environment variables for production, or in `.env.local` for local development.
+Set server-side keys (for Convex backend actions) in Convex Dashboard environment variables:
 
-  Without this key, agent logging will be disabled but the application will continue to function.
+- `ANTHROPIC_API_KEY` (required for AI recommendations)
+- `BRAINTRUST_API_KEY` (optional observability)
+- `BRAINTRUST_PROJECT_NAME` (optional, defaults to `grazing-agent`)
+- `GITHUB_TOKEN` (optional, enables GitHub issue creation from reports)
+- `CONVEX_DEBUG` (optional backend debug logging)
 
 ## Tech Stack
 
@@ -132,6 +110,7 @@ The following environment variables are required for full functionality:
 │   ├── src/
 │   └── ...
 ├── docs/
+│   ├── vision.md          # Thesis and product philosophy
 │   ├── architecture.md    # Technical architecture details
 │   ├── domain.md          # Remote sensing domain knowledge
 │   └── phasing.md         # Development phases
@@ -140,9 +119,11 @@ The following environment variables are required for full functionality:
 
 ## Documentation
 
-- [Development Phasing](docs/phasing.md) - How we build this incrementally
+- [Vision](docs/vision.md) - Thesis and product philosophy
 - [Technical Architecture](docs/architecture.md) - System design and data flow
 - [Domain Knowledge](docs/domain.md) - Remote sensing and vegetation science primer
+- [Development Phasing](docs/phasing.md) - How we build this incrementally
+- [Environment Setup](docs/environment.md) - Required and optional variables by runtime
 
 ## Contributing
 
@@ -156,6 +137,3 @@ If you discover a vulnerability, see [SECURITY.md](SECURITY.md).
 
 Licensed under the Apache License, Version 2.0. See [LICENSE](LICENSE) and [NOTICE](NOTICE).
 
----
-
-*OpenPasture - Building the intelligence layer for regenerative grazing.*
diff --git a/agents.md b/agents.md
index cfd9660..8ae0a3d 100644
--- a/agents.md
+++ b/agents.md
@@ -3,23 +3,7 @@
 > This file is the primary reference for AI agents working on this codebase.
 > Read this first before making any changes.
 
-## What This Project Is
-
-A **daily decision intelligence system for adaptive grazing**, powered by satellite data and human-in-the-loop AI. The core product is called the **Morning Farm Brief** - a daily ritual where farmers receive land status and recommended grazing actions.
-
-**Core thesis:** Grass is dynamic, fences are static. Satellite sensing lets us understand land conditions daily. We build the software brain that translates land signals into clear daily grazing decisions.
-
-## What This Project Is NOT
-
-Do not build or propose:
-- Livestock collar integration or firmware
-- Real-time livestock tracking/movement
-- Custom hardware
-- Automated actuation systems
-- Drone-based sensing
-- Real-time autonomous herding
-
-This is a **decision support system**, not an automation or control system.
+For the product thesis and philosophy, read [docs/vision.md](docs/vision.md) first.
 
 ## Core User Flow
 
@@ -118,14 +102,6 @@ These rules are intentionally simple and explainable for farmer trust.
 - **Format:** Cloud-Optimized GeoTIFFs (COGs)
 - **API:** STAC (SpatioTemporal Asset Catalog)
 
-## Design Principles
-
-1. **Plain language over charts** - Briefs should read like a farmer talking to a farmer
-2. **Transparency over magic** - Show assumptions and confidence, not black box scores
-3. **Daily cadence is sufficient** - Satellite refresh aligns with biological reality
-4. **Cloud cover is manageable** - Use rolling composites and multiple sources
-5. **Collar-agnostic output** - Generate geometry and copy-ready instructions, no direct integration
-
 ## File Structure Convention
 
 ```
@@ -144,13 +120,6 @@ These rules are intentionally simple and explainable for farmer trust.
 └── ...
 ```
 
-## Success Metrics
-
-The demo succeeds when:
-- A farmer says: "This matches how I think about my land."
-- An investor says: "This has a clear wedge and expansion path."
-- A collar company says: "This would make our product more valuable."
-
 ## Questions to Ask Before Building
 
 1. Does this feature support the Morning Farm Brief flow?
diff --git a/app/.env.example b/app/.env.example
index 4202d89..fcf520b 100644
--- a/app/.env.example
+++ b/app/.env.example
@@ -1,17 +1,14 @@
 # =============================================================================
-# Convex Configuration
+# App Runtime (Vite, read from app/.env.local)
 # =============================================================================
-# Get your Convex URL from the Convex dashboard after setting up your project
+# REQUIRED
+# Where used: app/src/main.tsx
+# Acquire: Convex Dashboard -> Deployments -> Settings -> URL and Deploy Key
 VITE_CONVEX_URL=https://your-project.convex.cloud
 
-# Convex deploy key (for server-side API calls)
-# Generate from Convex dashboard → Settings → Deploy Keys
-CONVEX_DEPLOY_KEY=prod:your_deploy_key_here
-
-# =============================================================================
-# Clerk Authentication
-# =============================================================================
-# Get from Clerk dashboard → API Keys
+# OPTIONAL when using development auth bypass.
+# Where used: app/src/lib/auth/index.tsx
+# Acquire: Clerk Dashboard -> API Keys -> Publishable key (pk_test_* or pk_live_*)
 VITE_CLERK_PUBLISHABLE_KEY=pk_test_your_key_here
 
 # Clerk JWT template required by ConvexProviderWithClerk.
@@ -28,14 +25,14 @@ CLERK_WEBHOOK_SECRET=whsec_your_webhook_secret_here
 # Example with custom Clerk domain: https://clerk.themodernstrategy.com
 CLERK_JWT_ISSUER_DOMAIN=
 
-# =============================================================================
-# Development Mode
-# =============================================================================
-# Set to "true" to bypass Clerk auth during local development
+# OPTIONAL, defaults to false.
+# Where used: app/src/lib/auth/index.tsx and demo onboarding routes
+# true = bypass Clerk auth locally
 VITE_DEV_AUTH=false
 
-# Set to "true" to disable the subscription paywall (useful for dev)
-# Paywall is enabled by default in production
+# OPTIONAL, defaults to false.
+# Where used: app/src/routes/app.tsx and app/src/routes/subscribe.tsx
+# true = disable subscription paywall checks
 VITE_PAYWALL_DISABLED=false
 
 # =============================================================================
@@ -48,9 +45,29 @@ VITE_POSTHOG_KEY=phc_your_project_api_key
 VITE_POSTHOG_HOST=https://us.i.posthog.com
 
 # =============================================================================
-# Braintrust Observability (Optional)
+# Convex Server Runtime (set in Convex Dashboard env vars, not Vite-exposed)
 # =============================================================================
-# For agent tracing and observability
-# Get from Braintrust dashboard → Settings → API Keys
+# REQUIRED for AI recommendations in Convex actions.
+# Where used: app/convex/grazingAgentDirect.ts via Anthropic SDK
+# Acquire: Anthropic Console -> API Keys
+ANTHROPIC_API_KEY=sk-ant-api03-your-key-here
+
+# OPTIONAL observability.
+# Where used: app/lib/braintrust.ts
+# Acquire: Braintrust Settings -> Organization -> API keys
 BRAINTRUST_API_KEY=your_braintrust_api_key
+
+# OPTIONAL, defaults to "grazing-agent" if omitted.
+# Where used: app/lib/braintrust.ts
 BRAINTRUST_PROJECT_NAME=grazing-agent
+
+# OPTIONAL, enables automatic GitHub issue creation from bug/feature reports.
+# Where used: app/convex/bugReportsAction.ts and app/convex/featureRequestsAction.ts
+# Acquire: GitHub -> Settings -> Developer settings -> Personal access tokens
+# Recommended: fine-grained token with Issues: Read and write on this repository.
+GITHUB_TOKEN=github_pat_your_token_here
+
+# OPTIONAL, defaults to false.
+# Where used: app/convex/lib/logger.ts
+# true = enable debug/info logging in Convex backend logger
+CONVEX_DEBUG=false
diff --git a/app/CLAUDE.md b/app/CLAUDE.md
index 7130041..e0d6297 100644
--- a/app/CLAUDE.md
+++ b/app/CLAUDE.md
@@ -4,7 +4,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project Overview
 
-OpenPasture - an open-source AI-powered decision support system that translates satellite-derived vegetation data into daily grazing recommendations. The core experience is a "Morning Farm Brief" that recommends which pasture to graze today.
+OpenPasture - an open-source adaptive grazing intelligence platform. See [docs/vision.md](../docs/vision.md) for the product thesis.
 
 ## Commands
 
diff --git a/app/README.md b/app/README.md
index a22c47a..f25bc54 100644
--- a/app/README.md
+++ b/app/README.md
@@ -13,13 +13,20 @@ npm run dev
 
 ## Environment
 
-Required env vars (set in `app/.env.local`):
+Copy and fill the app env template:
 
-- `VITE_CONVEX_URL` - Convex deployment URL
-- `VITE_DEV_AUTH=true` - bypass Clerk sign-in locally
-- `VITE_CLERK_PUBLISHABLE_KEY` - required when `VITE_DEV_AUTH` is not set
-- `CLERK_JWT_ISSUER_DOMAIN` (or `CLERK_FRONTEND_API_URL`) - Clerk issuer domain used by `convex/auth.config.ts`
-- Clerk JWT template named `convex` with audience `convex` - required for `ConvexProviderWithClerk`
+```bash
+cp .env.example .env.local
+```
+
+Primary app variables:
+
+- `VITE_CONVEX_URL` - required
+- `VITE_DEV_AUTH=true` - optional local auth bypass
+- `VITE_CLERK_PUBLISHABLE_KEY` - required when `VITE_DEV_AUTH=false`
+- `VITE_PAYWALL_DISABLED=true` - optional local paywall bypass
+
+Server-side Convex variables (`ANTHROPIC_API_KEY`, `BRAINTRUST_API_KEY`, etc.) are documented in [docs/environment.md](../docs/environment.md) and should be set in Convex Dashboard env vars.
 
 ## Scripts
 
diff --git a/app/index.html b/app/index.html
index e4aa675..7bb1d72 100644
--- a/app/index.html
+++ b/app/index.html
@@ -4,39 +4,35 @@
     <meta charset="UTF-8" />
     <link rel="icon" type="image/svg+xml" href="/vite.svg" />
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-    <meta name="description" content="Get daily land status reports and grazing recommendations powered by Sentinel-2 satellite imagery. Decision support for adaptive grazing." />
+    <meta name="description" content="AI grazing recommendations that predict what a trained farmer would do. Daily movement plans from satellite imagery and grazing science. Open source." />
     <meta name="keywords" content="openpasture, adaptive grazing, satellite imagery, NDVI, farm management, grazing recommendations, Morning Farm Brief" />
 
     <!-- Open Graph / Facebook -->
     <meta property="og:type" content="website" />
-    <meta property="og:title" content="OpenPasture | AI-Powered Adaptive Grazing" />
-    <meta property="og:description" content="Get daily land status reports and grazing recommendations powered by Sentinel-2 satellite imagery. Decision support for adaptive grazing." />
+    <meta property="og:title" content="OpenPasture | Predict What a Trained Farmer Would Do" />
+    <meta property="og:description" content="AI grazing recommendations that predict what a trained farmer would do. Daily movement plans from satellite imagery and grazing science. Open source." />
 
     <!-- Twitter -->
     <meta name="twitter:card" content="summary_large_image" />
-    <meta name="twitter:title" content="OpenPasture | AI-Powered Adaptive Grazing" />
-    <meta name="twitter:description" content="Daily grazing decisions powered by satellite intelligence. Get your Morning Farm Brief." />
+    <meta name="twitter:title" content="OpenPasture | Predict What a Trained Farmer Would Do" />
+    <meta name="twitter:description" content="Daily movement recommendations from satellite imagery and grazing science. Predict what a trained farmer would do -- every morning, every paddock." />
+
+    <title>OpenPasture | Predict What a Trained Farmer Would Do</title>
 
-    <title>OpenPasture | AI-Powered Adaptive Grazing</title>
-    
     <!-- Structured Data -->
     <script type="application/ld+json">
     {
       "@context": "https://schema.org",
       "@type": "SoftwareApplication",
-      "name": "Morning Farm Brief",
+      "name": "OpenPasture",
       "applicationCategory": "FarmManagementApplication",
-      "description": "Daily grazing decisions powered by satellite intelligence. Get land status reports and grazing recommendations for adaptive grazing.",
-      "offers": {
-        "@type": "Offer",
-        "price": "0",
-        "priceCurrency": "USD"
-      },
+      "description": "AI grazing recommendations that predict what a trained farmer would do. Daily movement plans from satellite imagery and grazing science.",
       "featureList": [
-        "Daily land status reports",
-        "AI-powered grazing recommendations",
-        "Satellite-derived vegetation health metrics",
-        "Digital farm mapping",
+        "Daily movement recommendations with full reasoning",
+        "NDVI satellite vegetation health by paddock",
+        "Targeted field observation requests",
+        "Farmer feedback learning loop",
+        "Pasture recovery and rest period tracking",
         "Exportable grazing plans"
       ]
     }
diff --git a/app/src/components/marketing/Footer.tsx b/app/src/components/marketing/Footer.tsx
index 5dd3b70..774c52b 100644
--- a/app/src/components/marketing/Footer.tsx
+++ b/app/src/components/marketing/Footer.tsx
@@ -14,7 +14,7 @@ export function Footer() {
                   to="/technology"
                   className="hover:text-cream transition-colors"
                 >
-                  Technology
+                  How It Works
                 </Link>
               </li>
               <li>
diff --git a/app/src/components/marketing/MarketingHeader.tsx b/app/src/components/marketing/MarketingHeader.tsx
index ed85359..6876193 100644
--- a/app/src/components/marketing/MarketingHeader.tsx
+++ b/app/src/components/marketing/MarketingHeader.tsx
@@ -17,7 +17,7 @@ export function MarketingHeader() {
             to="/technology"
             className="hover:opacity-70 transition-opacity text-muted-foreground"
           >
-            technology
+            how-it-works
           </Link>
           <Link
             to="/investors"
diff --git a/app/src/components/marketing/diagrams/SystemArchitectureDiagram.tsx b/app/src/components/marketing/diagrams/SystemArchitectureDiagram.tsx
index 5ca95fc..f5a6173 100644
--- a/app/src/components/marketing/diagrams/SystemArchitectureDiagram.tsx
+++ b/app/src/components/marketing/diagrams/SystemArchitectureDiagram.tsx
@@ -43,7 +43,7 @@ export function SystemArchitectureDiagram() {
           </div>
           <div className="flex items-center gap-2 text-xs text-muted-foreground">
             <Brain className="h-3 w-3 text-terracotta" />
-            <span>Scores paddocks, generates daily grazing plans with confidence levels</span>
+            <span>applies grazing science to your conditions. draws a polygon. explains its reasoning.</span>
           </div>
         </div>
 
diff --git a/app/src/components/marketing/diagrams/TechnologyConvergence.tsx b/app/src/components/marketing/diagrams/TechnologyConvergence.tsx
index 3b50c5b..afb2b0a 100644
--- a/app/src/components/marketing/diagrams/TechnologyConvergence.tsx
+++ b/app/src/components/marketing/diagrams/TechnologyConvergence.tsx
@@ -1,30 +1,24 @@
-import { Satellite, MapPin, Cpu, Brain, Check, Sparkles } from 'lucide-react'
+import { Satellite, Brain, Plane, Check, Sparkles } from 'lucide-react'
 
 const technologies = [
   {
     icon: Satellite,
-    label: 'Daily satellite observation',
+    label: 'Satellite resolution -- SpaceX/Blue Origin era',
     year: '2015+',
     highlight: false,
   },
-  {
-    icon: MapPin,
-    label: 'GPS-enabled animal tracking',
-    year: '2018+',
-    highlight: false,
-  },
-  {
-    icon: Cpu,
-    label: 'Cheap sensors & cloud',
-    year: '2020+',
-    highlight: false,
-  },
   {
     icon: Brain,
     label: 'LLMs that can reason & draw',
     year: '2022+',
     highlight: true,
   },
+  {
+    icon: Plane,
+    label: 'Drone accessibility -- falling costs, loosening regs',
+    year: '2023+',
+    highlight: false,
+  },
 ]
 
 export function TechnologyConvergence() {
@@ -90,14 +84,14 @@ export function TechnologyConvergence() {
         {/* Expandable detail */}
         <details className="mt-3">
           <summary className="text-[10px] text-muted-foreground cursor-pointer hover:text-foreground transition-colors text-center">
-            Why 2022 changed everything
+            Why now changed everything
           </summary>
           <div className="mt-2 pt-2 border-t border-border text-[10px] text-muted-foreground space-y-1.5">
             <p>
               Before LLMs, you'd need a custom model for each farm's unique conditions -- prohibitively expensive.
             </p>
             <p>
-              <span className="text-foreground">Modern LLMs already know how to reason.</span>{' '}
+              <span className="text-foreground">Modern language models already know how to reason.</span>{' '}
               They weigh tradeoffs, understand cause and effect, and can follow complex instructions.
               We just tell them about your land.
             </p>
diff --git a/app/src/components/ui/page-section.tsx b/app/src/components/ui/page-section.tsx
index 3784928..bfb049c 100644
--- a/app/src/components/ui/page-section.tsx
+++ b/app/src/components/ui/page-section.tsx
@@ -28,7 +28,7 @@ export function PageSection({
   children,
 }: PageSectionProps) {
   return (
-    <section className={cn('px-6 py-24', bgMap[bg], className)}>
+    <section className={cn('px-6 py-12', bgMap[bg], className)}>
       <div className={cn('mx-auto', maxWidthMap[maxWidth])}>
         {children}
       </div>
diff --git a/app/src/index.css b/app/src/index.css
index 3124d68..12c0619 100644
--- a/app/src/index.css
+++ b/app/src/index.css
@@ -116,21 +116,21 @@
     @apply border-border outline-ring/50;
   }
   html {
-    font-size: 14px;
+    font-size: 13px;
   }
   @media (min-width: 1024px) {
     html {
-      font-size: 18px;
+      font-size: 14px;
     }
   }
   @media (min-width: 1280px) {
     html {
-      font-size: 20px;
+      font-size: 15px;
     }
   }
   @media (min-width: 1536px) {
     html {
-      font-size: 21px;
+      font-size: 15px;
     }
   }
   body {
diff --git a/app/src/lib/docs/content/getting-started/installation.ts b/app/src/lib/docs/content/getting-started/installation.ts
index 6d9f153..d480eee 100644
--- a/app/src/lib/docs/content/getting-started/installation.ts
+++ b/app/src/lib/docs/content/getting-started/installation.ts
@@ -29,7 +29,10 @@ npm install`,
     },
     {
       heading: 'Environment Variables',
-      content: `Create a \`.env.local\` file with the following variables:
+      content: `Copy \`app/.env.example\` to \`app/.env.local\` and fill in values.
+The complete variable matrix (including ingestion and Convex dashboard-only keys) is in \`docs/environment.md\`.
+
+App-local keys:
 
 **Required:**
 - \`VITE_CONVEX_URL\` - Your Convex deployment URL (e.g., https://your-deployment.convex.cloud)
@@ -38,14 +41,19 @@ npm install`,
 - \`VITE_CLERK_PUBLISHABLE_KEY\` - For Clerk authentication in production
 - \`VITE_DEV_AUTH=true\` - For development mode without authentication
 
-**Optional:**
-- \`BRAINTRUST_API_KEY\` - For agent observability (set in Convex dashboard)
-- \`BRAINTRUST_PROJECT_NAME\` - Defaults to 'grazing-agent'`,
+**Optional app key:**
+- \`VITE_PAYWALL_DISABLED=true\` - Disable paywall checks locally
+
+**Set in Convex dashboard (server-side):**
+- \`ANTHROPIC_API_KEY\` - Required for AI recommendations
+- \`BRAINTRUST_API_KEY\` - Optional agent observability
+- \`BRAINTRUST_PROJECT_NAME\` - Optional, defaults to 'grazing-agent'`,
       codeExample: {
         language: 'bash',
         filename: '.env.local',
         code: `VITE_CONVEX_URL=https://your-deployment.convex.cloud
-VITE_DEV_AUTH=true`,
+VITE_DEV_AUTH=true
+VITE_PAYWALL_DISABLED=false`,
       },
     },
     {
diff --git a/app/src/routes/_public/investors.tsx b/app/src/routes/_public/investors.tsx
index db5c1b2..4098ebd 100644
--- a/app/src/routes/_public/investors.tsx
+++ b/app/src/routes/_public/investors.tsx
@@ -1,5 +1,5 @@
 import { createFileRoute, Link } from '@tanstack/react-router'
-import { Factory, Leaf, TrendingUp, Users, Globe, RefreshCw, Lock } from 'lucide-react'
+import { Factory, Leaf, TrendingUp, Users, Globe, RefreshCw, Lock, Target, Zap, Cpu } from 'lucide-react'
 import { MarketingHeader, Footer } from '@/components/marketing'
 import {
   DecisionScaleDiagram,
@@ -30,16 +30,15 @@ function InvestorsPage() {
           </h1>
 
           <p className="text-xl md:text-2xl font-light text-muted-foreground mb-4">
-            Pasture farming works. It just couldn't scale.
+            regenerative grazing increases stocking density. this system increases operational scale.
           </p>
 
           <p className="text-xl md:text-2xl font-light text-terracotta mb-6">
-            Until now.
+            multiplicative, not additive.
           </p>
 
           <p className="text-sm text-muted-foreground max-w-xl mx-auto">
-            How AI-automated decision-making unlocks industrial-scale output from regenerative methods --
-            and why this is the inflection point for sustainable livestock.
+            how small and mid-size farmers begin to compete with operations far larger than their own. and why the timing is now.
           </p>
         </div>
       </section>
@@ -63,7 +62,7 @@ function InvestorsPage() {
               The limiting factor isn't land, livestock, or biology.
             </p>
             <p className="text-base text-terracotta font-semibold mt-1">
-              It's human bandwidth for decisions.
+              it's time on pasture. the attention required to observe conditions and make daily movement decisions across a whole farm.
             </p>
           </div>
         </div>
@@ -71,63 +70,168 @@ function InvestorsPage() {
 
       <SectionDivider />
 
-      {/* AI Breakthrough */}
+      {/* The Value Proposition */}
       <PageSection className="py-10">
+        <div className="max-w-3xl mx-auto">
+          <div className="flex items-center gap-3 mb-4 justify-center">
+            <TrendingUp className="h-5 w-5 text-terracotta" />
+            <h2 className="text-2xl md:text-3xl font-bold">
+              The Value Proposition
+            </h2>
+          </div>
+          <p className="text-sm text-muted-foreground text-center mb-6">
+            two forces. one unlock.
+          </p>
+
+          <div className="grid md:grid-cols-2 gap-4 mb-6">
+            <div className="border-2 border-border p-5 shadow-hard-sm">
+              <div className="flex items-center gap-2 mb-3">
+                <Leaf className="h-4 w-4 text-olive" />
+                <h3 className="text-sm font-semibold">Regenerative Grazing</h3>
+              </div>
+              <p className="text-xs text-muted-foreground mb-3">increases stocking density</p>
+              <p className="text-sm">
+                more animals on the same acreage. the land is healthier and more productive.
+              </p>
+            </div>
+
+            <div className="border-2 border-olive/50 p-5 bg-olive-light shadow-hard-sm">
+              <div className="flex items-center gap-2 mb-3">
+                <TrendingUp className="h-4 w-4 text-terracotta" />
+                <h3 className="text-sm font-semibold">This System</h3>
+              </div>
+              <p className="text-xs text-terracotta mb-3">increases operational scale</p>
+              <p className="text-sm">
+                more acres managed by the same farmer. daily observation and decision-making handled autonomously.
+              </p>
+            </div>
+          </div>
+
+          <div className="border-2 border-border p-5 shadow-hard-sm text-center">
+            <p className="text-sm font-medium mb-1">
+              these are multiplicative, not additive.
+            </p>
+            <p className="text-xs text-muted-foreground">
+              the unlock for small and mid-size farmers competing with operations far larger than their own.
+            </p>
+          </div>
+        </div>
+      </PageSection>
+
+      <SectionDivider />
+
+      {/* The Product */}
+      <PageSection bg="white" className="py-10">
         <div className="max-w-3xl mx-auto">
           <h2 className="text-2xl md:text-3xl font-bold text-center mb-2">
-            The AI Breakthrough (2022-2024)
+            The Product
           </h2>
           <p className="text-sm text-muted-foreground text-center mb-6">
-            Why this wasn't possible before -- and why it is now.
+            a complete stack from raw data to actionable recommendations.
           </p>
 
           <div className="border-2 border-border p-5 mb-6 shadow-hard-sm">
             <p className="text-sm mb-4">
-              Before large language models, automating grazing decisions required building
-              a custom mathematical model for every farm -- capturing every variable, every
-              interaction, every edge case. That's prohibitively expensive for each
-              operation's unique conditions.
+              imagine <a href="https://youtu.be/wZagcP4U8-0?si=-oa0dDgLkYShvuAg" target="_blank" rel="noopener noreferrer" className="text-olive underline hover:text-terracotta transition-colors">greg judy</a>, joel salatin, or richard perkins staring at images of your farm all day. every morning they send you a text and a map: "move the animals here today." the suggestion is adjacent to where the animals currently are, timed and measured to benefit both the paddock being vacated and the broader farm system -- accounting for forward projections, weather, rest periods, and herd pressure.
             </p>
 
             <p className="text-sm text-terracotta font-semibold mb-4">
-              The insight: Modern LLMs already know how to reason.
+              that is what this system builds.
             </p>
 
+            <p className="text-sm text-muted-foreground mb-4">
+              the recommendation is rendered as a polygon on an orthomosaic representation of the farm. the farmer approves, adjusts, or rejects. that feedback improves future suggestions.
+            </p>
+
+            <p className="text-sm text-muted-foreground">
+              critically, the system does not just receive farmer input -- it directs the farmer's observational attention. each morning report includes a request for the specific field observation that would most reduce uncertainty in that day's recommendation: a grass height sample, a gut read on time left in the current paddock. the farmer spends less time walking the farm randomly and more time collecting targeted signal. this is how the system scales a farmer's attention rather than replacing it.
+            </p>
+          </div>
+
+          <SystemArchitectureDiagram />
+        </div>
+      </PageSection>
+
+      <SectionDivider />
+
+      {/* The Farmer as Benchmark */}
+      <PageSection className="py-10">
+        <div className="max-w-3xl mx-auto">
+          <div className="flex items-center gap-3 mb-4 justify-center">
+            <Target className="h-5 w-5 text-olive" />
+            <h2 className="text-2xl md:text-3xl font-bold">
+              The Farmer as Benchmark
+            </h2>
+          </div>
+          <p className="text-sm text-muted-foreground text-center mb-6">
+            not the ceiling.
+          </p>
+
+          <div className="border-2 border-border p-5 mb-4 shadow-hard-sm">
             <p className="text-sm mb-4">
-              They understand cause and effect, weigh tradeoffs, and explain their thinking.
-              They just need context about <em>your</em> land.
+              the system earns trust by matching farmer judgment. it earns the right to exceed farmer judgment by showing better outcomes -- tracked over time through pasture health trends, stocking performance, and land recovery rates. the land itself is the feedback mechanism.
             </p>
+            <p className="text-sm text-muted-foreground">
+              this design prevents the atrophy risk: a farmer who always defers to the system does not lose their intuition as long as the outcomes confirm the decisions were right. outcome tracking is how the system proves it is working -- and how it surfaces when it isn't.
+            </p>
+          </div>
+
+          <div className="grid md:grid-cols-2 gap-4">
+            <div className="border-2 border-border p-4 shadow-hard-sm bg-olive-light">
+              <div className="text-xs font-semibold text-olive uppercase tracking-wider mb-2">Success</div>
+              <p className="text-sm">
+                "this matches how i think about my land" -- and they say it without having walked the pasture that morning.
+              </p>
+            </div>
+            <div className="border-2 border-border p-4 shadow-hard-sm bg-olive-light">
+              <div className="text-xs font-semibold text-terracotta uppercase tracking-wider mb-2">Wins</div>
+              <p className="text-sm">
+                "i couldn't run this many acres without it."
+              </p>
+            </div>
+          </div>
+        </div>
+      </PageSection>
+
+      <SectionDivider />
+
+      {/* Why Now */}
+      <PageSection bg="white" className="py-10">
+        <div className="max-w-3xl mx-auto">
+          <div className="flex items-center gap-3 mb-4 justify-center">
+            <Zap className="h-5 w-5 text-terracotta" />
+            <h2 className="text-2xl md:text-3xl font-bold">
+              Why Now
+            </h2>
+          </div>
+          <p className="text-sm text-muted-foreground text-center mb-6">
+            three vectors converging simultaneously.
+          </p>
 
-            <div className="space-y-2 mb-4">
-              <h4 className="text-xs font-semibold text-muted-foreground uppercase tracking-wider">Our Architecture</h4>
-              <div className="grid gap-2">
-                <div className="flex items-start gap-2">
-                  <span className="text-olive font-bold text-xs">1.</span>
-                  <div>
-                    <span className="text-sm font-medium">Data Pipeline</span>
-                    <span className="text-sm text-muted-foreground"> &rarr; Gather the facts (satellite, weather, history)</span>
-                  </div>
+          <div className="border-2 border-border p-5 mb-6 shadow-hard-sm">
+            <div className="space-y-4">
+              <div className="flex items-start gap-3">
+                <span className="text-olive font-bold text-xs mt-1">01</span>
+                <div>
+                  <span className="text-sm font-medium">Satellite Resolution</span>
+                  <p className="text-xs text-muted-foreground mt-1">SpaceX and Blue Origin dramatically reduced the cost of reaching orbit. cheaper launch economics enabled commercial operators like Planet Labs to deploy large constellations, increasing revisit frequency and resolution while driving down cost. daily 3m imagery at commercial prices is a direct downstream effect of the reusable rocket era.</p>
                 </div>
-                <div className="flex items-start gap-2">
-                  <span className="text-olive font-bold text-xs">2.</span>
-                  <div>
-                    <span className="text-sm font-medium">Encoded Wisdom</span>
-                    <span className="text-sm text-muted-foreground"> &rarr; Explain grazing principles in the prompt</span>
-                  </div>
+              </div>
+              <div className="flex items-start gap-3">
+                <span className="text-terracotta font-bold text-xs mt-1">02</span>
+                <div>
+                  <span className="text-sm font-medium">LLM and Agent Maturity</span>
+                  <p className="text-xs text-muted-foreground mt-1">the ability to reason over complex, multi-variable decisions using codified domain knowledge -- and to do so as an autonomous agent operating on a daily loop -- has only recently become viable and affordable.</p>
                 </div>
-                <div className="flex items-start gap-2">
-                  <span className="text-olive font-bold text-xs">3.</span>
-                  <div>
-                    <span className="text-sm font-medium">Structured Output</span>
-                    <span className="text-sm text-muted-foreground"> &rarr; Let the model draw a polygon and justify its choice</span>
-                  </div>
+              </div>
+              <div className="flex items-start gap-3">
+                <span className="text-cobalt font-bold text-xs mt-1">03</span>
+                <div>
+                  <span className="text-sm font-medium">Drone Accessibility</span>
+                  <p className="text-xs text-muted-foreground mt-1">falling hardware costs and loosening regulations make drone imagery a credible near-term input for ground-truth validation of satellite estimates.</p>
                 </div>
               </div>
             </div>
-
-            <p className="text-sm text-muted-foreground">
-              This became possible in 2022. It became <em>practical</em> with 2024's frontier models.
-            </p>
           </div>
 
           <div className="mt-6">
@@ -139,7 +243,7 @@ function InvestorsPage() {
               to="/technology"
               className="text-sm text-olive hover:text-terracotta transition-colors"
             >
-              Deep dive: How LLMs became agricultural reasoning engines &rarr;
+              Deep dive: how we predict what a trained farmer would do &rarr;
             </Link>
           </div>
         </div>
@@ -147,24 +251,46 @@ function InvestorsPage() {
 
       <SectionDivider />
 
-      {/* The System */}
-      <PageSection bg="white" className="py-10">
+      {/* The Collar Dependency */}
+      <PageSection className="py-10">
         <div className="max-w-3xl mx-auto">
-          <h2 className="text-2xl md:text-3xl font-bold text-center mb-2">
-            The System: Three Layers
-          </h2>
+          <div className="flex items-center gap-3 mb-4 justify-center">
+            <Cpu className="h-5 w-5 text-cobalt" />
+            <h2 className="text-2xl md:text-3xl font-bold">
+              The Collar Dependency
+            </h2>
+          </div>
           <p className="text-sm text-muted-foreground text-center mb-6">
-            A complete stack from raw data to actionable recommendations.
+            decision support today. autonomous operation tomorrow.
           </p>
 
-          <SystemArchitectureDiagram />
+          <div className="border-2 border-border p-5 shadow-hard-sm">
+            <p className="text-sm mb-4">
+              collar integration is the unlock that converts this system from decision support into autonomous operation. without it, the system reduces cognitive burden but not labor burden. with it, the farmer's role shifts from daily physical execution to daily review and approval.
+            </p>
+
+            <p className="text-sm text-muted-foreground mb-4">
+              the collar market is currently locked behind proprietary software with no open apis.
+            </p>
+
+            <div className="bg-olive-light p-4 border-2 border-border mb-4">
+              <div className="text-xs font-semibold text-olive uppercase tracking-wider mb-2">Strategy</div>
+              <p className="text-sm text-muted-foreground">
+                build the decision-making layer aggressively and let its value create pressure on collar companies to integrate. if they don't open their apis, compete at the hardware layer with a cheap, cots-based, open-source collar design that commoditizes the market and breaks the lock.
+              </p>
+            </div>
+
+            <p className="text-sm font-semibold text-terracotta">
+              the brain is the defensible asset. the collar is not.
+            </p>
+          </div>
         </div>
       </PageSection>
 
       <SectionDivider />
 
       {/* Industrial Scale */}
-      <PageSection className="py-10">
+      <PageSection bg="white" className="py-10">
         <div className="max-w-3xl mx-auto text-center">
           <h2 className="text-2xl md:text-3xl font-bold mb-2">
             Industrial Scale, Biological Methods
@@ -210,7 +336,7 @@ function InvestorsPage() {
       <SectionDivider />
 
       {/* Market Opportunity */}
-      <PageSection bg="white" className="py-10">
+      <PageSection className="py-10">
         <div className="max-w-3xl mx-auto">
           <div className="flex items-center gap-3 mb-4 justify-center">
             <TrendingUp className="h-5 w-5 text-terracotta" />
@@ -244,6 +370,9 @@ function InvestorsPage() {
             <p className="text-sm text-terracotta font-medium mb-3">
               We remove that bottleneck.
             </p>
+            <p className="text-sm text-muted-foreground mb-3">
+              higher stocking density combined with greater operational scale means the same acreage supports more animals, managed by fewer people. the addressable market expands as adoption compounds.
+            </p>
             <div className="pt-3 border-t-2 border-border">
               <p className="text-xs text-muted-foreground">
                 <span className="text-olive font-semibold">Currently in beta</span> with early adopter farmers -- building the training data flywheel now.
@@ -256,7 +385,7 @@ function InvestorsPage() {
       <SectionDivider />
 
       {/* Platform Vision */}
-      <PageSection className="py-10">
+      <PageSection bg="white" className="py-10">
         <div className="max-w-3xl mx-auto">
           <div className="flex items-center gap-3 mb-4 justify-center">
             <Globe className="h-5 w-5 text-olive" />
@@ -265,12 +394,12 @@ function InvestorsPage() {
             </h2>
           </div>
           <p className="text-sm text-muted-foreground text-center mb-6">
-            Not just software. An integrated stack for regenerative livestock at scale.
+            not just software. a recommendation engine that becomes an operating system.
           </p>
 
           <div className="border-2 border-border p-5 mb-4 shadow-hard-sm">
             <p className="text-sm font-medium mb-4">
-              We're building what John Deere built for industrial farming -- but for rotational grazing.
+              not just software. a recommendation engine that becomes an operating system for regenerative grazing at scale.
             </p>
             <div className="grid md:grid-cols-3 gap-4">
               <div>
@@ -282,7 +411,7 @@ function InvestorsPage() {
               <div>
                 <div className="text-xs font-semibold text-olive uppercase tracking-wider mb-2">Sensors</div>
                 <p className="text-xs text-muted-foreground">
-                  Integration with satellite, collar, soil, and forage measurement systems
+                  satellite imagery today. collar integration as the unlock. open-source hardware if proprietary apis remain closed.
                 </p>
               </div>
               <div>
@@ -308,7 +437,7 @@ function InvestorsPage() {
       <SectionDivider />
 
       {/* The Moat */}
-      <PageSection bg="white" className="py-10">
+      <PageSection className="py-10">
         <div className="max-w-3xl mx-auto">
           <div className="flex items-center gap-3 mb-4 justify-center">
             <Lock className="h-5 w-5 text-cobalt" />
@@ -332,7 +461,7 @@ function InvestorsPage() {
                     has thousands of farmer corrections paired with satellite observations.
                   </p>
                   <p className="text-xs text-terracotta font-medium">
-                    We're building it. One decision at a time. Across thousands of farms.
+                    one approval, one correction, one field note at a time.
                   </p>
                 </div>
               </div>
@@ -361,7 +490,7 @@ function InvestorsPage() {
       <SectionDivider />
 
       {/* Get Involved */}
-      <PageSection className="py-10">
+      <PageSection bg="white" className="py-10">
         <div className="max-w-3xl mx-auto text-center">
           <h2 className="text-2xl md:text-3xl font-bold mb-2">
             Learn More
@@ -395,7 +524,7 @@ function InvestorsPage() {
               to="/sign-in"
               className="border-2 border-border p-4 shadow-hard-sm hover:border-olive hover:shadow-hard hover:-translate-x-0.5 hover:-translate-y-0.5 transition-all block"
             >
-              <h3 className="text-sm font-semibold mb-1">Try It Free</h3>
+              <h3 className="text-sm font-semibold mb-1">Try the Beta</h3>
               <p className="text-xs text-muted-foreground">
                 See it working today
               </p>
diff --git a/app/src/routes/_public/research.tsx b/app/src/routes/_public/research.tsx
index 1e53081..d59f437 100644
--- a/app/src/routes/_public/research.tsx
+++ b/app/src/routes/_public/research.tsx
@@ -58,6 +58,7 @@ function ResearchPage() {
                 <li>-- Forage utilization efficiency</li>
                 <li>-- Grazing season length extension</li>
                 <li>-- Input cost reduction</li>
+                <li>-- recommendation accuracy vs. experienced farmer judgment</li>
               </ul>
             </div>
 
@@ -81,7 +82,7 @@ function ResearchPage() {
                 <span className="rsc-badge-solid ml-3" style={{ verticalAlign: 'middle' }}>Roadmap</span>
               </h3>
               <p className="text-muted-foreground mb-4">
-                GPS collar integration is on our roadmap and will open new research possibilities in livestock behavior and welfare:
+                GPS collar integration is on our roadmap and will open new research possibilities in livestock behavior and welfare. collar data also enables the shift from decision support to autonomous operation -- the unlock that removes labor burden, not just cognitive burden.
               </p>
               <ul className="text-muted-foreground space-y-2 text-sm">
                 <li>-- Grazing distribution and preference patterns</li>
diff --git a/app/src/routes/_public/technology.tsx b/app/src/routes/_public/technology.tsx
index 185e9e4..e055bd0 100644
--- a/app/src/routes/_public/technology.tsx
+++ b/app/src/routes/_public/technology.tsx
@@ -27,7 +27,7 @@ function TechnologyPage() {
           </h1>
 
           <p className="text-lg md:text-xl text-muted-foreground max-w-2xl">
-            How we turn satellite data, weather patterns, and farmer knowledge into daily grazing decisions.
+            how we predict what a trained farmer would do, using satellite imagery, grazing science, and farmer feedback.
           </p>
         </div>
       </PageSection>
@@ -117,7 +117,7 @@ function TechnologyPage() {
             </p>
 
             <p className="text-terracotta font-semibold text-lg mb-5">
-              Modern AI models like Claude already know how to reason. They understand cause
+              modern language models already know how to reason. They understand cause
               and effect. They can weigh tradeoffs. They can explain their thinking. They
               just need context about YOUR land.
             </p>
@@ -194,14 +194,14 @@ function TechnologyPage() {
               <Cpu className="h-7 w-7 text-terracotta mx-auto mb-3" />
               <h3 className="font-semibold mb-2">Scoring Engine</h3>
               <p className="text-sm text-muted-foreground">
-                Multi-factor analysis considering recovery, weather, and grazing history
+                the system applies accumulated grazing science to your specific conditions
               </p>
             </div>
             <div className="text-center">
               <Code className="h-7 w-7 text-cobalt mx-auto mb-3" />
               <h3 className="font-semibold mb-2">Recommendation AI</h3>
               <p className="text-sm text-muted-foreground">
-                Natural language explanations with confidence scores and assumptions
+                a movement recommendation with transparent reasoning. adjust or approve.
               </p>
             </div>
           </div>
@@ -334,8 +334,8 @@ function TechnologyPage() {
                 <div className="flex items-start gap-3">
                   <Plus className="h-4 w-4 text-terracotta mt-0.5 flex-shrink-0" />
                   <div>
-                    <div className="font-medium text-sm">GPS Collars</div>
-                    <div className="text-muted-foreground text-sm">Animal behavior insights</div>
+                    <div className="font-medium text-sm">GPS Collars (the unlock)</div>
+                    <div className="text-muted-foreground text-sm">decision support becomes autonomous operation</div>
                   </div>
                 </div>
               </div>
diff --git a/app/src/routes/index.tsx b/app/src/routes/index.tsx
index 8e43bf8..3fcfefa 100644
--- a/app/src/routes/index.tsx
+++ b/app/src/routes/index.tsx
@@ -11,23 +11,22 @@ function LandingPage() {
       <MarketingHeader />
 
       {/* Hero */}
-      <section className="px-6 py-28 relative">
+      <section className="px-6 py-16 relative">
         <div className="rsc-grid-bg" />
         <div className="max-w-5xl mx-auto relative z-10">
           <div className="flex items-center gap-3 mb-8 rsc-fade rsc-d1">
-            <span className="rsc-badge">satellite-input</span>
+            <span className="rsc-badge">satellite-imagery</span>
             <span className="font-bold text-olive">&rarr;</span>
-            <span className="rsc-badge">probability-output</span>
+            <span className="rsc-badge">farmer-judgment</span>
           </div>
 
           <h1 className="font-serif text-5xl md:text-6xl lg:text-8xl font-bold leading-[0.92] mb-8 rsc-fade rsc-d2" style={{ textWrap: 'balance' }}>
-            Two Layers.<br />
-            <span className="text-olive italic">One Intelligence.</span>
+            what a trained farmer would do.<br />
+            <span className="text-olive italic">every morning. every paddock.</span>
           </h1>
 
           <p className="text-lg md:text-xl max-w-3xl leading-relaxed mb-10 text-muted-foreground rsc-fade rsc-d3">
-            Satellite imagery meets probability theory. Your morning brief delivers
-            gradient confidence, not binary answers. All data, all reasoning, all open source.
+            satellite imagery and grazing science, combined into a daily movement recommendation. the system predicts what an experienced farmer would decide -- then asks you to confirm, adjust, or override.
           </p>
 
           <div className="flex flex-col sm:flex-row gap-4 rsc-fade rsc-d3">
@@ -58,10 +57,10 @@ function LandingPage() {
         <div className="max-w-4xl mx-auto text-center">
           <span className="rsc-badge-solid mb-4 inline-block">early-access-beta</span>
           <h2 className="text-2xl md:text-3xl font-bold mb-3" style={{ textWrap: 'balance' }}>
-            join ranchers testing probability-powered grazing
+            join farmers testing ai-augmented grazing
           </h2>
           <p className="text-sm text-muted-foreground">
-            be among the first to experience gradient intelligence. free during beta.
+            the system learns how you think about your land. reduced pricing during beta.
           </p>
         </div>
       </section>
@@ -72,18 +71,17 @@ function LandingPage() {
           <div>
             <span className="rsc-badge mb-6 inline-block">available-now</span>
             <h2 className="text-3xl md:text-4xl font-bold mb-6 leading-tight">
-              <span className="rsc-prompt">morning_brief</span><br />
-              <span className="text-olive">--probability-scored</span>
+              <span className="rsc-prompt">morning_report</span><br />
+              <span className="text-olive">--daily</span>
             </h2>
             <p className="text-sm mb-8 leading-relaxed text-muted-foreground">
-              AI analyzes satellite data overnight and delivers a recommended grazing plan
-              with probability distributions and reasoning you can interrogate.
+              imagine <a href="https://youtu.be/wZagcP4U8-0?si=-oa0dDgLkYShvuAg" target="_blank" rel="noopener noreferrer" className="text-olive underline hover:text-terracotta transition-colors">greg judy</a> staring at satellite images of your farm all day. every morning, a text and a map: &apos;move the animals here today.&apos; the reasoning is transparent. adjust or approve with one tap.
             </p>
             <ul className="space-y-3">
               {[
-                'confidence scores replace yes/no',
-                'probability reasoning for each suggestion',
-                'override --force when conditions change',
+                'daily movement recommendation with full reasoning',
+                'targeted observation request to close data gaps',
+                'approve, adjust, or override. your call.',
               ].map((item, idx) => (
                 <li key={idx} className="flex items-start gap-3 text-xs">
                   <span className="font-bold mt-0.5 text-olive">[x]</span>
@@ -94,7 +92,7 @@ function LandingPage() {
           </div>
 
           <WindowChrome title="brief --daily">
-            <img src="/marketing/morning-brief.png" alt="AI-powered grazing recommendation with probability reasoning" className="w-full" />
+            <img src="/marketing/morning-brief.png" alt="AI-powered grazing recommendation with movement plan and reasoning" className="w-full" />
           </WindowChrome>
         </div>
       </PageSection>
@@ -114,14 +112,13 @@ function LandingPage() {
               see what&apos;s invisible<br /><span className="text-terracotta">from the ground</span>
             </h2>
             <p className="text-sm mb-8 leading-relaxed text-muted-foreground">
-              NDVI heatmaps reveal vegetation health as probability distributions.
-              Historical patterns show recovery trends as continuous fields.
+              daily satellite imagery shows vegetation health across your whole farm. track recovery paddock by paddock. spot what you&apos;d miss on foot.
             </p>
             <ul className="space-y-3">
               {[
-                'probability-colored vegetation health',
-                'recovery distributions over time',
-                'uncertainty zones flagged early',
+                'ndvi vegetation health by paddock',
+                'rest period and recovery tracking',
+                'known gap: grass height needs field notes (for now)',
               ].map((item, idx) => (
                 <li key={idx} className="flex items-start gap-3 text-xs">
                   <span className="font-bold mt-0.5 text-terracotta">[x]</span>
@@ -141,17 +138,16 @@ function LandingPage() {
           <div>
             <span className="rsc-badge mb-6 inline-block">available-now</span>
             <h2 className="text-3xl md:text-4xl font-bold mb-6 leading-tight">
-              track recovery<br /><span className="text-cobalt">quantify certainty</span>
+              track recovery<br /><span className="text-cobalt">earn trust.</span>
             </h2>
             <p className="text-sm mb-8 leading-relaxed text-muted-foreground">
-              Monitor rest periods as probability distributions. Recovery rates as data fields.
-              Know not just if ready, but how confident.
+              the system earns trust by matching your judgment. it earns the right to exceed it by showing better outcomes. land health is the feedback mechanism.
             </p>
             <ul className="space-y-3">
               {[
-                'real-time pasture probability tracking',
-                'historical confidence trend analysis',
-                'uncertainty-aware rotation planning',
+                'pasture health trends over time',
+                'recommendation accuracy vs. your decisions',
+                'the land tells you if the system is working',
               ].map((item, idx) => (
                 <li key={idx} className="flex items-start gap-3 text-xs">
                   <span className="font-bold mt-0.5 text-cobalt">[x]</span>
@@ -162,7 +158,7 @@ function LandingPage() {
           </div>
 
           <WindowChrome title="analytics --recovery --trends">
-            <img src="/marketing/analytics-dashboard.png" alt="Analytics dashboard with pasture recovery, rest periods, NDVI trends, and AI partnership stats" className="w-full" />
+            <img src="/marketing/analytics-dashboard.png" alt="Analytics dashboard with pasture recovery, rest periods, NDVI trends, and recommendation accuracy" className="w-full" />
           </WindowChrome>
         </div>
       </PageSection>
@@ -173,13 +169,13 @@ function LandingPage() {
       <PageSection bg="white" maxWidth="6xl">
         <div className="text-center mb-16">
           <h2 className="text-3xl md:text-4xl font-bold mb-3">how_it_works()</h2>
-          <p className="text-sm text-muted-foreground">three steps to probability-powered grazing</p>
+          <p className="text-sm text-muted-foreground">three steps. satellite to suggestion.</p>
         </div>
         <div className="grid md:grid-cols-3 gap-6">
           {[
-            { num: '01', title: 'capture', desc: 'sentinel-2 delivers multispectral imagery every 5 days. discrete pixels become raw material.', color: 'text-olive' },
-            { num: '02', title: 'process', desc: 'models transform point measurements into continuous probability fields. uncertainty is a feature.', color: 'text-terracotta' },
-            { num: '03', title: 'harvest', desc: 'wake up to probability-scored guidance. not just where to graze but how certain we are.', color: 'text-cobalt' },
+            { num: '01', title: 'observe', desc: 'satellite imagery captures vegetation health across your farm daily. resolution: 3m. coverage: every paddock.', color: 'text-olive' },
+            { num: '02', title: 'reason', desc: 'the system applies grazing science to your specific conditions. weather, rest periods, recovery rates, herd pressure. then it draws a plan.', color: 'text-terracotta' },
+            { num: '03', title: 'decide', desc: 'you wake up to a recommendation and a map. approve, adjust, or override. your correction makes the next suggestion better.', color: 'text-cobalt' },
           ].map((step, idx) => (
             <div key={idx} className="rsc-card">
               <div className={`text-4xl font-bold mb-4 ${step.color}`}>{step.num}</div>
@@ -216,21 +212,21 @@ function LandingPage() {
       </PageSection>
 
       {/* CTA */}
-      <section className="px-6 py-28 bg-dark">
+      <section className="px-6 py-16 bg-dark">
         <div className="max-w-4xl mx-auto text-center">
           <h2 className="text-3xl md:text-4xl lg:text-5xl font-bold mb-8 text-white" style={{ textWrap: 'balance' }}>
-            ready to harvest<br />
-            <span className="text-terracotta">probability?</span>
+            ready to scale<br />
+            <span className="text-terracotta">your judgment?</span>
           </h2>
           <p className="text-sm mb-12 max-w-2xl mx-auto text-terracotta-muted">
-            join the beta. start using data intelligence for adaptive grazing.
+            join the beta. the system learns how you think about your land.
           </p>
           <div className="flex flex-col sm:flex-row gap-4 justify-center">
             <Link
               to="/sign-in"
               className="px-10 py-4 bg-white font-bold border-2 border-white transition-all shadow-[4px_4px_0_rgba(168,58,50,0.5)] hover:translate-x-0.5 hover:translate-y-0.5 hover:shadow-[2px_2px_0_rgba(168,58,50,0.5)] flex items-center justify-center gap-2 text-xs uppercase tracking-wider group text-dark"
             >
-              ./signup --free
+              ./signup --beta
               <ArrowRight className="w-4 h-4 group-hover:translate-x-1 transition-transform" strokeWidth={2.5} />
             </Link>
             <Link
diff --git a/app/src/vite-env.d.ts b/app/src/vite-env.d.ts
index aad6263..1972185 100644
--- a/app/src/vite-env.d.ts
+++ b/app/src/vite-env.d.ts
@@ -2,6 +2,9 @@
 
 interface ImportMetaEnv {
   readonly VITE_CONVEX_URL?: string
+  readonly VITE_DEV_AUTH?: string
+  readonly VITE_CLERK_PUBLISHABLE_KEY?: string
+  readonly VITE_PAYWALL_DISABLED?: string
 }
 
 interface ImportMeta {
diff --git a/docs/environment.md b/docs/environment.md
new file mode 100644
index 0000000..683787a
--- /dev/null
+++ b/docs/environment.md
@@ -0,0 +1,100 @@
+# Environment Variables
+
+This is the single source of truth for environment configuration across the repo.
+
+## App Local Env (`app/.env.local`)
+
+From repo root:
+
+```bash
+cp app/.env.example app/.env.local
+```
+
+Frontend (Vite) variables:
+
+- `VITE_CONVEX_URL` (required)
+- `VITE_DEV_AUTH` (optional, defaults to `false`)
+- `VITE_CLERK_PUBLISHABLE_KEY` (required when `VITE_DEV_AUTH=false`)
+- `VITE_PAYWALL_DISABLED` (optional, defaults to `false`)
+
+## Ingestion Local Env (`src/ingestion/.env.local`)
+
+From repo root:
+
+```bash
+cp src/ingestion/.env.example src/ingestion/.env.local
+```
+
+Minimum required for Convex writeback:
+
+- `CONVEX_DEPLOYMENT_URL`
+- `CONVEX_API_KEY`
+
+Provider/storage variables are optional and depend on enabled integrations.
+
+## Convex Dashboard Env Vars (Server-side only)
+
+Set these in Convex Dashboard for backend actions:
+
+1. Open [Convex Dashboard](https://dashboard.convex.dev/)
+2. Select your deployment
+3. Go to `Deployment Settings` -> `Environment Variables`
+4. Add/update keys below
+
+Required:
+
+- `ANTHROPIC_API_KEY`
+
+Optional:
+
+- `BRAINTRUST_API_KEY`
+- `BRAINTRUST_PROJECT_NAME`
+- `GITHUB_TOKEN`
+- `CONVEX_DEBUG`
+
+These are server-side only. Do not prefix them with `VITE_`.
+
+## Variable Matrix
+
+| Variable | Scope | Required | Default | Where Used | Acquisition |
+|---|---|---|---|---|---|
+| `VITE_CONVEX_URL` | Vite | Yes | none | `app/src/main.tsx` | Convex deployment URL from [Deployment Settings](https://docs.convex.dev/dashboard/deployments/deployment-settings) |
+| `VITE_DEV_AUTH` | Vite | No | `false` | `app/src/lib/auth/index.tsx` | Local toggle, no external provider needed |
+| `VITE_CLERK_PUBLISHABLE_KEY` | Vite | Conditional (`VITE_DEV_AUTH=false`) | none | `app/src/lib/auth/index.tsx` | Clerk dashboard publishable key per [Clerk env docs](https://clerk.com/docs/deployments/clerk-environment-variables) |
+| `VITE_PAYWALL_DISABLED` | Vite | No | `false` | `app/src/routes/app.tsx` | Local toggle, no external provider needed |
+| `ANTHROPIC_API_KEY` | Convex | Yes (AI path) | none | `app/convex/grazingAgentDirect.ts` | Anthropic console key per [Anthropic getting started](https://docs.anthropic.com/en/api/getting-started) |
+| `BRAINTRUST_API_KEY` | Convex | No | none | `app/lib/braintrust.ts` | Braintrust org API key per [Braintrust API intro](https://www.braintrust.dev/docs/api-reference/introduction) |
+| `BRAINTRUST_PROJECT_NAME` | Convex | No | `grazing-agent` | `app/lib/braintrust.ts` | Optional name from [Braintrust projects](https://www.braintrust.dev/docs/core/projects) |
+| `GITHUB_TOKEN` | Convex | No | none | `app/convex/bugReportsAction.ts` and `app/convex/featureRequestsAction.ts` | GitHub PAT per [token docs](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/creating-a-personal-access-token) |
+| `CONVEX_DEBUG` | Convex | No | `false` | `app/convex/lib/logger.ts` | Local/Convex toggle, no provider key |
+| `CONVEX_DEPLOYMENT_URL` | Ingestion | Yes (writeback path) | none | `src/ingestion/writer.py` | Convex deployment URL from [Deployment Settings](https://docs.convex.dev/dashboard/deployments/deployment-settings) |
+| `CONVEX_API_KEY` | Ingestion | Yes (writeback path) | none | `src/ingestion/writer.py` and `src/ingestion/scheduler.py` | Convex deploy key per [deploy key docs](https://docs.convex.dev/cli/deploy-key-types) |
+| `COPERNICUS_CLIENT_ID` | Ingestion | No | none | `src/ingestion/providers/copernicus.py` | CDSE OAuth client per [CDSE auth docs](https://documentation.dataspace.copernicus.eu/APIs/SentinelHub/Overview/Authentication.html) |
+| `COPERNICUS_CLIENT_SECRET` | Ingestion | No | none | `src/ingestion/providers/copernicus.py` | CDSE OAuth secret from same client setup |
+| `PL_API_KEY` | Ingestion | No | none | `src/ingestion/providers/planet_scope.py` | Planet API key per [Planet auth docs](https://docs.planet.com/develop/authentication/) |
+| `PL_CLIENT_ID` | Ingestion | No | none | `src/ingestion/providers/planet_scope.py` | Planet OAuth client credentials (optional) per [Planet auth docs](https://docs.planet.com/develop/authentication/) |
+| `PL_CLIENT_SECRET` | Ingestion | No | none | `src/ingestion/providers/planet_scope.py` | Planet OAuth client credentials (optional) per [Planet auth docs](https://docs.planet.com/develop/authentication/) |
+| `R2_ACCOUNT_ID` | Ingestion | No | none | `src/ingestion/storage/r2.py` | Cloudflare account ID per [R2 S3 token docs](https://developers.cloudflare.com/r2/api/s3/tokens/) |
+| `R2_ACCESS_KEY_ID` | Ingestion | No | none | `src/ingestion/storage/r2.py` | Cloudflare R2 S3 access key per [R2 get started](https://developers.cloudflare.com/r2/get-started/cli/) |
+| `R2_SECRET_ACCESS_KEY` | Ingestion | No | none | `src/ingestion/storage/r2.py` | Cloudflare R2 S3 secret key per [R2 get started](https://developers.cloudflare.com/r2/get-started/cli/) |
+| `R2_BUCKET_NAME` | Ingestion | No | `grazing-satellite-tiles` | `src/ingestion/storage/r2.py` | Existing/new R2 bucket name |
+| `R2_PUBLIC_URL_BASE` | Ingestion | No | none | `src/ingestion/storage/r2.py` | Optional public/custom domain URL |
+| `COMPOSITE_WINDOW_DAYS` | Ingestion | No | `21` | `src/ingestion/config.py` | Local tuning value |
+| `MAX_CLOUD_COVER` | Ingestion | No | `50` | `src/ingestion/config.py` | Local tuning value |
+| `MIN_CLOUD_FREE_PCT` | Ingestion | No | `0.3` | `src/ingestion/config.py` | Local tuning value |
+| `DEFAULT_PROVIDER` | Ingestion | No | `sentinel2` | `src/ingestion/config.py` | Local tuning value (`copernicus` or `sentinel2`) |
+| `ENABLE_PLANET_SCOPE` | Ingestion | No | `false` | `src/ingestion/config.py` | Local toggle for PlanetScope integration |
+| `WRITE_TO_CONVEX` | Ingestion | No | `true` | `src/ingestion/config.py` | Local toggle for writeback |
+| `OUTPUT_DIR` | Ingestion | No | `output` | `src/ingestion/config.py` | Local filesystem path |
+| `LOG_LEVEL` | Ingestion | No | `INFO` | `src/ingestion/config.py` | Local logging config |
+
+## Convex CLI Parity
+
+Reference: [Convex env docs](https://docs.convex.dev/production/environment-variables).
+
+```bash
+npx convex env list
+npx convex env get ANTHROPIC_API_KEY
+npx convex env set ANTHROPIC_API_KEY "your_value"
+npx convex env remove CONVEX_DEBUG
+```
diff --git a/docs/phasing.md b/docs/phasing.md
index 437b163..4c78991 100644
--- a/docs/phasing.md
+++ b/docs/phasing.md
@@ -119,7 +119,7 @@ python src/ingestion/pipeline.py --dev --write-convex --output /tmp/pan_output
 
 ### Why This Phase Matters
 
-The product IS the Morning Farm Brief. If this experience isn't crisp, nothing else matters. Skipping this results in UI rework loops later.
+The recommendation UI is the primary farmer touchpoint. If it isn't crisp, nothing else matters. Skipping this results in UI rework loops later.
 
 ---
 
diff --git a/docs/vision.md b/docs/vision.md
new file mode 100644
index 0000000..3c749ee
--- /dev/null
+++ b/docs/vision.md
@@ -0,0 +1,104 @@
+# Vision
+
+## Founding Principle
+
+Complex systems — like nature itself — are best utilized when you work within them, not when you abstract away from them. Regenerative agriculture was built on this insight. This product is built on the same one.
+
+The decision of when to move animals across pasture is a multi-variable analysis: labor availability, animal stress and condition, the health of the currently grazed paddock, the next paddock, the pasture as a whole, the farm as a whole, projected weather in the near and intermediate term, and the historic performance of the land over time. A farmer with the right training has proven capable of weighing all of this — some of it empirically, much of it intuitively. We seek to augment that decision-making layer by predicting what a trained farmer would do.
+
+## The Real Bottleneck
+
+The limiting factor in regenerative grazing is not data. Farmers don't say they're blocked by lack of soil sensors or nutrient analysis. The real bottleneck is time on pasture — the attention required to physically observe land conditions and make daily movement decisions across a whole farm.
+
+Soil analysis and dense sensor networks feel like solutions, but they address a problem farmers don't actually have. Threading the needle means avoiding two failure modes:
+
+- **Over-indexing on trackable metrics** reduces a complex living system to numbers, repeating the reductionist mistakes regenerative agriculture was built to correct.
+- **Over-indexing on physical observation** doesn't scale. A farmer cannot be everywhere every day.
+
+The optimal solution scales the farmer's visual monitoring capability without replacing their judgment.
+
+## The Value Proposition
+
+Regenerative grazing increases stocking density — more animals on the same acreage because the land is healthier and more productive. Our system increases operational scale — more acres managed by the same farmer because daily observation and decision-making is handled autonomously.
+
+These are multiplicative, not additive. A farmer who adopts regenerative practices and uses this system doesn't just farm more profitably — they begin to compete with operations far larger than their own. This is the unlock for the small and mid-size farmer.
+
+## Thesis
+
+A farmer needs two things to make a movement decision:
+
+1. **Accumulated industry knowledge** — the principles developed by practitioners like Greg Judy, Joel Salatin, and Richard Perkins: move at least weekly, ideally daily; don't overgraze; leave residual; rest periods matter; context of the broader ecosystem.
+2. **Visual confirmation of pasture health** plus a mental model of prior conditions.
+
+The first can be captured in text and reasoned over by an LLM. Combined with pasture imagery, this creates a feedback loop capable of generating movement suggestions comparable to farmer intuition.
+
+## The Product
+
+Imagine Greg Judy, Joel Salatin, or Richard Perkins staring at images of your farm all day. Every morning they send you a text and a map: *"Move the animals here today."* The suggestion is adjacent to where the animals currently are, timed and measured to benefit both the paddock being vacated and the broader farm system — accounting for forward projections, weather, rest periods, and herd pressure.
+
+That is what this system builds.
+
+The recommendation is rendered as a polygon on an orthomosaic representation of the farm. The farmer approves, adjusts, or rejects. That feedback improves future suggestions.
+
+Critically, the system does not just receive farmer input — it directs the farmer's observational attention. Each morning report includes a request for the specific field observation that would most reduce uncertainty in that day's recommendation: a grass height sample, a gut read on time left in the current paddock. The farmer spends less time walking the farm randomly and more time collecting targeted signal. This is how the system scales a farmer's attention rather than replacing it.
+
+## The Farmer as Benchmark
+
+The farmer is the benchmark of efficiency, not the ceiling.
+
+The system earns trust by matching farmer judgment. It earns the right to exceed farmer judgment by showing better outcomes — tracked over time through pasture health trends, stocking performance, and land recovery rates. The land itself is the feedback mechanism.
+
+This design prevents the atrophy risk: a farmer who always defers to the system does not lose their intuition as long as the outcomes confirm the decisions were right. Outcome tracking is how the system proves it is working — and how it surfaces when it isn't.
+
+## What We Need to Build It
+
+**Starting inputs:**
+- Satellite imagery — daily, global, up to 3m resolution on paid plans, providing NDVI and biomass proxies sufficient for farm and pasture-level health assessment
+- Farmer field notes, optionally with photos
+
+These two inputs are sufficient to architect the core decision engine. The farm's digital representation lives in imagery — not soil probes or collar telemetry.
+
+**Known gap:** Satellite cannot reliably assess precise grass height or growth stage at the paddock level. The decision to graze just before seed, taking only the top third of the plant, requires this information. Field notes close this gap initially. Fixed pasture cameras are the likely near-term bridge. Drone imagery or improved satellite analysis resolves it long-term.
+
+**Optimization path (once the engine exists):**
+- Farmer approval rate as a training signal (RLHF)
+- Fixed pasture cameras for growth stage visibility
+- Drone imagery for higher resolution
+- Direct push to virtual fencing collars
+
+## The Collar Dependency
+
+Collar integration is the unlock that converts this system from decision support into autonomous operation. Without it, the system reduces cognitive burden but not labor burden — a harder value proposition to sustain at scale. With it, the farmer's role shifts from daily physical execution to daily review and approval. That is the equivalent shift that coding agents brought to software development.
+
+The collar market is currently locked behind proprietary software with no open APIs.
+
+Our strategy: build the decision-making layer aggressively and let its value create pressure on collar companies to integrate. If they don't open their APIs, we will compete at the hardware layer with a cheap, COTS-based, open-source collar design that commoditizes the market and breaks the lock. The brain is the defensible asset. The collar is not.
+
+## What We Are Not Building
+
+This is a recommendation engine, not a control system. We are not building:
+
+- Automated fence actuation
+- Livestock tracking or collar firmware
+- Soil monitoring infrastructure
+- Real-time herd movement systems
+
+Those are downstream of a working recommendation engine. We build the brain first.
+
+## Why Now
+
+Three vectors are converging:
+
+**Satellite resolution** — SpaceX and Blue Origin dramatically reduced the cost of reaching orbit. Cheaper launch economics enabled commercial operators like Planet Labs to deploy large constellations, increasing revisit frequency and resolution while driving down cost. Daily 3m imagery at commercial prices is a direct downstream effect of the reusable rocket era.
+
+**LLM and agent maturity** — The ability to reason over complex, multi-variable decisions using codified domain knowledge — and to do so as an autonomous agent operating on a daily loop — has only recently become viable and affordable.
+
+**Drone accessibility** — Falling hardware costs and loosening regulations make drone imagery a credible near-term input, though it is not required to begin.
+
+These three vectors arriving simultaneously is what makes the core architecture buildable today.
+
+## Success
+
+The product succeeds when a farmer says: "This matches how I think about my land" — and they say it without having walked the pasture that morning.
+
+The product wins when a farmer says: "I couldn't run this many acres without it."
diff --git a/src/ingestion/.env.example b/src/ingestion/.env.example
index 9faad74..dffaa87 100644
--- a/src/ingestion/.env.example
+++ b/src/ingestion/.env.example
@@ -3,46 +3,53 @@
 # =============================================================================
 
 # =============================================================================
-# Convex Connection
+# 1) REQUIRED: Convex Writeback
 # =============================================================================
-# Your Convex deployment URL
+# Where used: writer.py, scheduler.py
+# Acquire: Convex Dashboard -> Deployments -> Settings -> URL and Deploy Key
 CONVEX_DEPLOYMENT_URL=https://your-project.convex.cloud
 
-# Convex deploy key for API mutations
-# Generate from Convex dashboard → Settings → Deploy Keys
+# Deploy key used for Convex HTTP mutations
 CONVEX_API_KEY=prod:your_deploy_key_here
 
 # =============================================================================
-# Copernicus Data Space (Sentinel-2 Imagery)
+# 2) Sentinel-2 Provider Auth (Copernicus Data Space)
 # =============================================================================
-# Register at: https://dataspace.copernicus.eu/
-# Create OAuth client at: Dashboard → OAuth Clients
-# Use "client_credentials" grant type
+# Where used: imagery_checker.py, providers/copernicus.py
+# Acquire:
+# 1) Register account at https://dataspace.copernicus.eu/
+# 2) Create OAuth client (client_credentials) in CDSE dashboard
 COPERNICUS_CLIENT_ID=your_client_id_here
 COPERNICUS_CLIENT_SECRET=your_client_secret_here
 
 # =============================================================================
-# Cloudflare R2 Storage (For Paid Tiers)
+# 3) OPTIONAL: Planet Provider Auth (premium/professional tiers)
 # =============================================================================
-# Get from Cloudflare dashboard → R2 → Manage R2 API Tokens
+# Where used: providers/planet_scope.py, providers/__init__.py
+# Preferred for Planet Data API access:
+PL_API_KEY=your_planet_api_key
+
+# Optional OAuth credentials (Sentinel Hub / future provider flows)
+PL_CLIENT_ID=your_planet_client_id
+PL_CLIENT_SECRET=your_planet_client_secret
+
+# =============================================================================
+# 4) OPTIONAL: Cloudflare R2 Storage
+# =============================================================================
+# Where used: storage/r2.py
+# Acquire: Cloudflare Dashboard -> R2 -> API tokens (S3 compatible keys)
 R2_ACCOUNT_ID=your_cloudflare_account_id
 R2_ACCESS_KEY_ID=your_r2_access_key_id
 R2_SECRET_ACCESS_KEY=your_r2_secret_access_key
 R2_BUCKET_NAME=grazing-satellite-tiles
 
-# Optional: Public URL base for R2 bucket (if using custom domain)
-# R2_PUBLIC_URL_BASE=https://tiles.yourdomain.com
+# Optional if you serve tiles via public/custom domain
+R2_PUBLIC_URL_BASE=https://tiles.yourdomain.com
 
 # =============================================================================
-# Planet Labs (Optional - For Premium Tiers)
-# =============================================================================
-# Get from Planet dashboard → Account Settings → API Key
-# Only needed for professional/enterprise tiers
-PL_API_KEY=your_planet_api_key
-
-# =============================================================================
-# Pipeline Settings (Optional - Defaults shown)
+# 5) OPTIONAL: Pipeline Tuning (defaults shown)
 # =============================================================================
+# Where used: config.py
 # Number of days to look back for composite imagery
 COMPOSITE_WINDOW_DAYS=21
 
@@ -55,8 +62,14 @@ MIN_CLOUD_FREE_PCT=0.3
 # Default satellite provider ("copernicus" or "sentinel2" for Planetary Computer)
 DEFAULT_PROVIDER=copernicus
 
+# Enable PlanetScope integration in pipeline configuration
+ENABLE_PLANET_SCOPE=false
+
 # Whether to write results to Convex
 WRITE_TO_CONVEX=true
 
 # Output directory for local files
 OUTPUT_DIR=output
+
+# Logging level: DEBUG, INFO, WARNING, ERROR
+LOG_LEVEL=INFO
diff --git a/start.sh b/start.sh
new file mode 100755
index 0000000..d27bbb9
--- /dev/null
+++ b/start.sh
@@ -0,0 +1,23 @@
+#!/bin/bash
+
+# Start OpenPasture dev environment
+# Runs Convex backend + Vite dev server concurrently
+
+cd "$(dirname "$0")/app"
+
+cleanup() {
+  echo "Shutting down..."
+  kill "$CONVEX_PID" "$VITE_PID" 2>/dev/null
+  wait "$CONVEX_PID" "$VITE_PID" 2>/dev/null
+  exit 0
+}
+
+trap cleanup INT TERM
+
+npx convex dev &
+CONVEX_PID=$!
+
+pnpm dev &
+VITE_PID=$!
+
+wait "$CONVEX_PID" "$VITE_PID"