Understanding Uncertainty: Cost Confidence Intervals and Reconciliation
New cost analysis tools help quantify uncertainty in Dyson swarm budgets, from confidence intervals on BOM items to reconciliation analysis across LLM estimates.
Project Dyson Team
Project Dyson
Project Dyson's Phase 0 estimate is $9 billion. But what does that number actually mean?
Is it a floor that assumes everything goes perfectly? A ceiling with maximum contingency? A median with equal probability of over/under? Without understanding the uncertainty, a cost estimate is just a number.
Today we're releasing cost confidence intervals and reconciliation analysis—tools that quantify and explain uncertainty in our estimates.
The Problem with Point Estimates
Our three-LLM consensus approach produces different estimates for every BOM item:
| BOM Item | Claude | GPT | Gemini |
|---|---|---|---|
| Prospecting Satellites | $250M | $280M | $220M |
| Processing Platform | $800M | $1.2B | $650M |
| Transport Vehicles | $2.0B | $1.8B | $2.2B |
Which estimate is "right"? They all are—under different assumptions about technology readiness, launch costs, development complexity, and risk margins.
Cost Confidence Intervals
Each BOM item now includes three cost levels:
| Level | Definition | Use Case |
|---|---|---|
| Low | 10th percentile estimate | Best-case with favorable assumptions |
| Medium | 50th percentile estimate | Most likely outcome |
| High | 90th percentile estimate | Conservative with risk margins |
How Intervals Are Determined
- LLM Variation — Range across Claude, GPT, and Gemini estimates
- Historical Analogs — NASA cost growth factors for comparable missions
- TRL Adjustment — Higher uncertainty for lower technology readiness
- Divergent Views — Explicit disagreements widen intervals
Example: Processing Platform
| Metric | Value |
|---|---|
| Low Estimate | $650M |
| Medium Estimate | $850M |
| High Estimate | $1.2B |
| Confidence Spread | 1.85× |
The 1.85× spread reflects significant uncertainty about ISRU processing efficiency and first-of-a-kind development costs.
Cost Reconciliation Analysis
The new Cost Reconciliation tool answers: Where do our LLMs disagree, and why?
Reconciliation Categories
| Category | Meaning |
|---|---|
| Aligned | All models within 20% of consensus |
| Minor Divergence | 20-50% spread, different assumptions |
| Major Divergence | >50% spread, fundamental disagreement |
Major Divergence Example: ISRU Capital Costs
Claude estimates $50B for seed factory infrastructure. Gemini estimates $30B. GPT estimates $100B.
Root Cause: Different assumptions about:
- Self-replication capability (Claude: partial, GPT: minimal, Gemini: full)
- Material processing efficiency
- Automation level required
Resolution Path: The simulation (RQ-1-12) suggests $50B baseline is reasonable, but the range should be $30-100B until design matures.
Divergent View Prioritization
Not all disagreements matter equally. The new prioritization system ranks divergent views by:
- Cost Impact — Absolute dollar difference
- Schedule Impact — Effect on critical path
- Technical Risk — Likelihood of being wrong
- Actionability — Can we resolve through research?
Top Divergent Views (February 2026)
| Rank | Topic | Impact | Status |
|---|---|---|---|
| 1 | ISRU capital investment | $70B range | Research planned |
| 2 | Collector unit size | 3× cost difference | Under simulation |
| 3 | Assembly Hub location | $5B logistics delta | Resolved (L4/L5) |
Using Cost Tools
For BOM Items
Each BOM detail page now shows:
- Low/Medium/High cost range
- Confidence interval visualization
- Divergent views affecting that item
- Links to reconciliation analysis
For Phases
Phase summary pages include:
- Aggregate confidence intervals
- Monte Carlo probability distribution
- Top cost risks for the phase
For Project-Wide Analysis
The Cost Reconciliation page provides:
- All major divergences ranked by impact
- Resolution status for each divergence
- Trend tracking as research progresses
Why This Matters
A $9B estimate with ±50% uncertainty means something very different from $9B with ±10% uncertainty. By quantifying and explaining cost uncertainty:
- Funders understand the range of outcomes
- Engineers know where to focus cost reduction
- Planners can build appropriate contingencies
- Critics see we're honest about what we don't know
The goal isn't to eliminate uncertainty—it's to measure and manage it.
Explore: Cost Reconciliation | Phase 0 BOM | Phase 1 BOM
Tags: