CVTas Scenario Planner

Overview

The CVTas Scenario Planner is an interactive web-based tool for exploring catastrophic scenarios and understanding the practical constraints of sheltering and sustaining AI safety researchers through civilizational disruption.

Purpose: Help stakeholders mechanistically understand trade-offs, resource constraints, and critical path dependencies when planning for X-risk events.

URL: /scenarios/planner/

Key Features

🎮 Interactive Exploration

  • Adjust parameters with sliders to see real-time impacts
  • Three default scenarios (Nuclear Early, Taiwan Crisis, Baseline)
  • Immediate visual feedback on outcomes

📊 Modular Calculators

Each module feeds into the overall scenario assessment:

  1. Timeline Module - When does the catastrophic event occur?
  2. Population/Migration Module - How many people need shelter and when?
  3. Shelter/Infrastructure Module - Do we have enough beds?
  4. Food Production Module - Can we feed everyone?
  5. Resources/Budget Module - Can we afford it?

⚠ Critical Path Analysis

The system identifies: - FAILURE - Critical failures (people stranded, starvation, etc.) - CRITICAL - Urgent action required NOW - WARNING - Significant risks to address

📈 Outcome Metrics

  • Survival Probability - Overall likelihood of mission success
  • Bed Capacity vs Need - Shelter shortfalls
  • Food Self-Sufficiency - Can the site produce enough food?
  • Budget Runway - Months of funding remaining
  • Construction Feasibility - Is there time to build?

Default Scenarios

1. Nuclear War - Early Timeline

Major nuclear exchange 12 months from now, 6 months earlier than baseline.

Key Parameters: - Event: 12 months - Warning: 14 days - Arrivals: 5/day, 100 total - Growing season affected: YES - Yield reduction: 30%

Typical Outcome: CRITICAL - Shows need for immediate infrastructure development

2. China-Taiwan Conflict

Regional conflict disrupts supply chains, 18 months from now.

Key Parameters: - Event: 18 months - Warning: 30 days - Arrivals: 3/day, 80 total - Growing season affected: NO - Yield reduction: 10%

Typical Outcome: AT_RISK - Marginal with current resources

3. Baseline - Current Planning

Current assumptions, 24 months timeline.

Key Parameters: - Event: 24 months - Warning: 60 days - Arrivals: 2/day, 100 total - Growing season affected: NO - Yield reduction: 0%

Typical Outcome: MARGINAL to SUCCESS - Depends on resource allocation

How to Use

For Sponsors/Stakeholders

  1. Start with Baseline - See current planning assumptions
  2. Try "Nuclear Early" - See what happens if timelines compress
  3. Adjust sliders - Explore "what if" questions
  4. Note critical constraints - Understand what needs to happen NOW

For Project Planners

  1. Model your scenarios - Adjust all parameters to match your analysis
  2. Identify bottlenecks - Which constraints are most binding?
  3. Find critical path - What needs to start immediately?
  4. Test sensitivities - Which parameters have biggest impact?

For Technical Team

  1. Save scenarios (coming soon) - Export parameter sets
  2. Add complexity (future) - More detailed models as data improves
  3. Integrate with roadmaps (future) - Link to Jira planning

Understanding the Calculations

Shelter Capacity

months_to_prepare = event_months - (warning_days / 30)
construction_possible = months_to_prepare >= construction_lead_time
peak_occupancy = current_beds + (buildable_units * beds_per_unit)
bed_deficit = total_arrivals - peak_occupancy

Food Security

food_needed_kg/year = people * 2kg/day * 365 days
potential_production_kg = hectares * 5000kg/ha * (1 - yield_reduction)
self_sufficiency = production / needed * 100%

Critical Failures

  • People Stranded: bed_deficit > 0
  • Starvation Risk: food_stores_days < days_to_first_harvest
  • Budget Exhaustion: monthly_runway < event_months

Assumptions & Simplifications

The current model uses simplified calculations:

  • 2 kg food/person/day - Rough average for balanced diet
  • 5000 kg/hectare/year - Conservative permaculture yield
  • 6 months construction lead time - For basic shelter infrastructure
  • 180 days to first harvest - Normal growing conditions
  • 365 days if season affected - Miss one full growing season

These will be refined as we get actual quotes, site data, and expert input.

Extending the Model

The architecture supports adding complexity over time:

Phase 1 (Current)

  • ✅ Simple dependency calculations
  • ✅ Basic resource constraints
  • ✅ Manual parameter adjustment

Phase 2 (Planned)

  • ⏳ Save/load scenarios
  • ⏳ Export to JSON
  • ⏳ More detailed food production model
  • ⏳ Infrastructure cost breakdowns

Phase 3 (Future)

  • 🔮 Bayesian/DAG dependency visualization
  • 🔮 Monte Carlo uncertainty analysis
  • 🔮 Jira integration (auto-create critical path tasks)
  • 🔮 Real quote/cost data integration
  • 🔮 Multi-site comparison

Technical Details

Stack

  • Backend: Django views with calculation logic
  • Frontend: Vanilla JavaScript + HTML/CSS
  • Style: "Old school" terminal aesthetic (deliberate)
  • API: Simple POST to /scenarios/api/calculate/

Data Flow

  1. User adjusts slider
  2. JavaScript captures all parameters
  3. POST to Django calculate endpoint
  4. Server runs interdependent calculations
  5. Returns JSON with results
  6. JavaScript updates UI

Files

Dark Humor Elements

The tool maintains a serious planning focus with occasional gallows humor:

  • Terminal green aesthetic (retro computing meets doomsday prep)
  • "You died of dysentery" style outcome messages
  • Blunt constraint descriptions ("PEOPLE WILL STARVE")
  • Survival probability gauge (ominous progress bar)

This helps stakeholders grasp severity without being preachy.

Use Cases

1. Sponsor Pitch

"We need $2M in Year 1 for infrastructure. Here's why..."

Load Nuclear Early scenario, show bed deficit and construction timeline. Demonstrates need for immediate action.

2. Team Planning

"Which tasks are on the critical path?"

Adjust construction lead time, see when it blocks success. Identify procurement dependencies.

3. Risk Communication

"What's our biggest vulnerability?"

Run all scenarios, note which constraints appear most often. Focus mitigation there.

4. Budget Allocation

"Where should we spend money first?"

Test: More land vs. More storage vs. Faster construction. See impact on survival probability.

Known Limitations

  1. Simplified food model - Doesn't account for crop diversity, seasons, livestock
  2. No skill constraints - Assumes labor is available for all tasks
  3. Linear scaling - Some things don't scale linearly (e.g., management overhead)
  4. No uncertainty - Point estimates, not probability distributions (yet)
  5. Single site - Doesn't model multi-site strategies

Roadmap

Q4 2025 - ✅ Basic scenario planner (DONE) - ⏳ Save/load scenarios - ⏳ Export to CSV/JSON

Q1 2026 - More detailed food production model - Infrastructure cost breakdowns by category - Integration with real quotes/estimates

Q2 2026 - Uncertainty quantification (Monte Carlo) - DAG visualization of dependencies - Multi-scenario comparison view

Q3 2026 - Jira integration (auto-task creation) - Multi-site comparison - Skill/labor constraint modeling

Contributing

To add new parameters or constraints:

  1. Add parameter to default scenarios in views.py
  2. Add slider/input to template
  3. Update getParameters() JavaScript function
  4. Add calculation logic in calculate_scenario() view
  5. Update results display
  6. Document assumptions

See CONTRIBUTING.md for code standards.

Support

Questions or issues? Check: - System Architecture - Initial Planning Doc - Main README

Remember: This tool shows potential failure modes. The goal is to identify and mitigate risks BEFORE they become real. If the scenarios look scary, that's intentional - it means we need to act now.