Weight Classification
Purpose: Match decision rigor to decision stakes. All 7 gates run at appropriate depth.
The Three Weights
| Weight | Time | When to Use |
|---|---|---|
| Light | <10 min | Reversible decisions, domain expert, lower stakes |
| Medium | 30-40 min | Moderate stakes, some uncertainty, standard rigor |
| Complete | 60-90 min | High stakes, irreversible, significant uncertainty |
Critical: Weight affects depth, not gates. All 7 gates always run.
Weight Selection
When to Suggest Weight
Weight is suggested at the end of Thesis Gate, after the AI understands:
- Financial/strategic impact
- Reversibility
- Time horizon
- User's domain expertise
- Complexity revealed
Suggestion Format
Based on what you've described:
- [Key factor 1]
- [Key factor 2]
- [Key factor 3]
I recommend **[Weight]** depth (~[time] minutes).
[Light: Quick pass through all gates - good for reversible decisions where you have expertise]
[Medium: Standard rigor - balances thoroughness with efficiency]
[Complete: Full analysis - for high-stakes, hard-to-reverse decisions]
Does that feel right, or would you prefer [alternative]?
Weight Selection Criteria
| Factor | Light | Medium | Complete |
|---|---|---|---|
| Financial impact | <$10K | $10K-$100K | >$100K |
| Reversibility | Easy (days) | Moderate (weeks/months) | Hard (years/permanent) |
| Domain expertise | Expert | Familiar | New territory |
| Stakeholders affected | Just me | Team/department | Organization/customers |
| Time to know outcome | Days/weeks | Months | Years |
If any factor suggests Complete, recommend Complete.
Depth by Gate
Gate 1: Thesis
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Questions | 3-4 key questions | 5-6 questions | Full exploration |
| Stakeholders | List only | Identify conflicts | Map relationships |
| Constraints | Non-negotiables only | All constraints | Constraint flexibility analysis |
| Success criteria | 1-2 metrics | 3-4 metrics | Comprehensive metrics + timeline |
Gate 2: Landscape
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Alternatives | 2-3 + do nothing | 4-5 + do nothing | 5-6+ + creative options |
| Decision Points | Surface level (3-5) | Thorough (6-10) | Comprehensive (10+) |
| Intersections | Note obvious ones | Map dependencies | Full intersection analysis |
| Knowns/Unknowns | Quick categorization | Standard matrix | Detailed with confidence |
Gate 3: Research
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Web research | None - user knowledge | 1-2 targeted searches | Thorough multi-source |
| Reference class | Ask user if known | Find basic reference | Deep reference class analysis |
| Document review | Skip unless provided | Review key docs | Comprehensive review |
| Evidence quality | Trust user | Basic assessment | Full quality evaluation |
Gate 4: Calibration
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Fact/Assumption split | Quick pass | Full categorization | Detailed with evidence |
| Confidence levels | High/Low only | High/Medium/Low | Calibrated percentages |
| Assumption inventory | Critical only | Important + critical | Comprehensive inventory |
| Blind spots | Quick check | Identify gaps | Deep blind spot analysis |
Gate 5: Contrarian
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Pre-mortem scenarios | Top 3 risks | 5 scenarios | 5+ with probability/severity |
| Steel-man | Skip | Top rejected alternative | All rejected alternatives |
| Bias audit | Quick self-check | Standard audit | Full bias analysis |
| Second-order effects | Note obvious | One level deep | Two levels deep |
Gate 6: Synthesis
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Must-Be-True | Top 3 conditions | 5-6 conditions | Comprehensive conditions |
| Exit criteria | 1-2 basic tripwires | 3-4 specific tripwires | Detailed thresholds + dates |
| Risk assessment | High-level summary | Consolidated risks | Full risk matrix |
| Recommendation | Brief | Standard with rationale | Comprehensive with confidence |
Gate 7: Decision
| Aspect | Light | Medium | Complete |
|---|---|---|---|
| Summary | One paragraph | Standard summary | Full decision summary |
| Documentation | Minimal artifact | Standard artifact | Complete artifact |
| Rationale capture | Brief | Key factors | Verbatim + deciding factors |
Upgrading Mid-Process
User-Initiated Upgrade
User can say "let's go deeper" at any point. AI responds:
Upgrading to [Weight]. Remaining gates will run at [Weight] depth.
[If past gates were Light, note what was skipped that won't be revisited]
Already-completed gates stay as-is. No going back.
AI-Suggested Upgrade
AI should suggest upgrading when it detects:
| Signal | Example | Suggestion |
|---|---|---|
| Emerging complexity | More alternatives than expected | "This is more complex than it first appeared. Want to upgrade to Medium/Complete?" |
| Low confidence on critical assumptions | Key assumption rated Low | "A critical assumption has low confidence. Recommend upgrading to get more rigor here." |
| Surprising research findings | Reference class shows high failure rate | "The reference class suggests higher risk than expected. Want to go deeper?" |
| Stakes seem higher | Financial impact larger than initially framed | "The stakes seem higher than initially framed. Recommend Complete depth." |
| User uncertainty | User expresses doubt repeatedly | "You seem uncertain about some key points. Want to spend more time here?" |
Upgrade Suggestion Format
⚠️ **Upgrade Suggestion**
I'm noticing [signal]. This suggests [weight] depth might be more appropriate.
Current: [Weight] (~X min remaining)
Suggested: [Weight] (~Y min remaining)
[Brief explanation of what deeper analysis would add]
Continue at current depth, or upgrade?
Weight in Artifact
Document the weight in the decision artifact header:
# Decision: [Statement]
**Weight:** [Light/Medium/Complete]
**Estimated time:** [X minutes]
**Upgraded:** [Yes - from Light at Gate 3 / No]
Commands
Weight can be specified when starting:
/decide → AI suggests weight after Thesis
/decide --light → Start at Light (can upgrade)
/decide --medium → Start at Medium (can upgrade)
/decide --complete → Start at Complete
If weight is pre-specified, skip the weight suggestion step in Thesis but still allow upgrades.