1  # Tribe Sizing Algorithm
  2  
  3  *proto-014 | How much firepower for this problem?*
  4  
  5  ---
  6  
  7  - **principle**
  8    - "How big should a team be? Based on task complexity, not fixed size."
  9    - "Dynamic navigation ([[A3 Dynamic Pole Navigation]])."
 10  
 11  - **shape**
 12    - Match tribe size to problem characteristics
 13    - Weight formula: W = (Blast × 0.25) + (Fundamentality × 0.25) + (Tension × 0.20) + (Error × 0.15) + (Reversibility × 0.15)
 14    - W < 0.2: Solo; W 0.2-0.4: Pair; W 0.4-0.6: Small tribe; W > 0.6: Full tribal council
 15  
 16  ---
 17  
 18  **Status:** 📄 DOCUMENTED
 19  
 20  ---
 21  
 22  ## The Problem
 23  
 24  Not every problem needs a full tribal council. Not every problem can be handled by a single reviewer. We need an algorithm to match tribe size to problem characteristics.
 25  
 26  ---
 27  
 28  ## Input Variables
 29  
 30  | Variable | Symbol | Scale | Question |
 31  |----------|--------|-------|----------|
 32  | **Blast Radius** | B | 0-1 | How much does this affect if wrong? |
 33  | **Fundamentality** | F | 0-1 | How core is this to the architecture? |
 34  | **Tension** | T | 0-1 | How much ambiguity/disagreement exists? |
 35  | **Error Likelihood** | E | 0-1 | How likely are we to make a mistake? |
 36  | **Reversibility** | R | 0-1 | How hard to undo if wrong? (1 = irreversible) |
 37  
 38  ---
 39  
 40  ## Scoring Examples
 41  
 42  | Problem | B | F | T | E | R | Notes |
 43  |---------|---|---|---|---|---|-------|
 44  | Typo fix | 0.1 | 0.0 | 0.0 | 0.1 | 0.0 | Trivial |
 45  | New pattern doc | 0.3 | 0.2 | 0.2 | 0.3 | 0.1 | Low stakes |
 46  | API change | 0.5 | 0.4 | 0.3 | 0.4 | 0.6 | Medium |
 47  | Axiom modification | 0.9 | 1.0 | 0.7 | 0.5 | 0.9 | Critical |
 48  | Architecture decision | 0.8 | 0.9 | 0.8 | 0.6 | 0.8 | High stakes |
 49  
 50  ---
 51  
 52  ## The Algorithm
 53  
 54  ### Step 1: Compute Problem Weight
 55  
 56  ```
 57  W = (B × 0.25) + (F × 0.25) + (T × 0.20) + (E × 0.15) + (R × 0.15)
 58  
 59  Scale: 0-1
 60  ```
 61  
 62  ### Step 2: Map Weight to Tribe Configuration
 63  
 64  | Weight | Configuration | Description |
 65  |--------|---------------|-------------|
 66  | W < 0.2 | **SOLO** | Single reviewer, skip FO |
 67  | 0.2 ≤ W < 0.4 | **LIGHT** | 2-3 Sonnet, no FO |
 68  | 0.4 ≤ W < 0.6 | **STANDARD** | 3 Sonnet + Opus FO |
 69  | 0.6 ≤ W < 0.8 | **TRIBAL** | 3-5 Sonnet + Opus FO + Human spot-check |
 70  | W ≥ 0.8 | **COUNCIL** | Multiple tribes + Meta-FO + Human decision |
 71  
 72  ### Step 3: Adjust for Special Cases
 73  
 74  | Condition | Adjustment |
 75  |-----------|------------|
 76  | F = 1.0 (axiom-level) | Always COUNCIL minimum |
 77  | R > 0.8 (irreversible) | Bump up one level |
 78  | T > 0.8 (high tension) | Add adversarial reviewer |
 79  | E > 0.8 (error-prone) | Add second pass |
 80  
 81  ---
 82  
 83  ## Tribe Configurations
 84  
 85  ### SOLO (W < 0.2)
 86  ```
 87  1 Sonnet reviewer
 88  No First Officer
 89  Quick turnaround
 90  ```
 91  
 92  ### LIGHT (0.2 ≤ W < 0.4)
 93  ```
 94  2-3 Sonnet reviewers (different lenses)
 95  No First Officer
 96  Compare findings manually
 97  ```
 98  
 99  ### STANDARD (0.4 ≤ W < 0.6)
100  ```
101  3 Sonnet reviewers
102  1 Opus First Officer
103  Full synthesis
104  ```
105  
106  ### TRIBAL (0.6 ≤ W < 0.8)
107  ```
108  3-5 Sonnet reviewers
109  1 Opus First Officer
110  Human reviews synthesis
111  Human spot-checks findings
112  ```
113  
114  ### COUNCIL (W ≥ 0.8)
115  ```
116  Multiple tribes (if scope requires)
117  Opus First Officers per tribe
118  Opus Meta-First Officer
119  Human as final arbiter
120  Full deliberation protocol
121  ```
122  
123  ---
124  
125  ## Human Involvement Algorithm
126  
127  ### The Paradox
128  
129  ```
130  SEEING                              DECIDING
131  ──────                              ────────
132  Passive model-building              Active frame-locking
133  Low time cost                       High time cost
134  Broad coverage                      Deep engagement
135  Awareness                           Ownership
136  ```
137  
138  **Both are valuable. Budget both.**
139  
140  ### Human Time Budget
141  
142  ```
143  Let H = available human attention units per cycle
144  
145  Allocate:
146  ├── H × 0.3 = SEEING (summaries, dashboards)
147  ├── H × 0.5 = DECIDING (strategic choices)
148  └── H × 0.2 = SAMPLING (spot-checks)
149  ```
150  
151  ### Injection Point Selection
152  
153  | Problem Weight | Human Injection |
154  |----------------|-----------------|
155  | W < 0.2 | None (see summary later) |
156  | 0.2 ≤ W < 0.4 | See synthesis only |
157  | 0.4 ≤ W < 0.6 | See synthesis + option to deep-dive |
158  | 0.6 ≤ W < 0.8 | Review synthesis + decide on disputed items |
159  | W ≥ 0.8 | Full involvement - see all, decide key points |
160  
161  ### The Sampling Strategy
162  
163  Even for low-weight problems, human should randomly sample:
164  
165  ```
166  SAMPLING PROTOCOL
167  
168  1. All W ≥ 0.8 → Human sees + decides (100%)
169  2. W 0.6-0.8 → Human sees synthesis (100%), decides disputes (100%)
170  3. W 0.4-0.6 → Human sees synthesis (100%), samples detail (20%)
171  4. W 0.2-0.4 → Human sees summary (100%), samples full (10%)
172  5. W < 0.2 → Human sees aggregate stats (100%), samples (5%)
173  
174  Sampling is RANDOM - prevents gaming.
175  ```
176  
177  ---
178  
179  ## The Frame-Locking Insight
180  
181  > **Deciding locks in frames more powerfully than seeing.**
182  
183  When you decide:
184  - You commit to a position
185  - You register the tradeoffs
186  - You update your model actively
187  - The decision becomes part of your cognitive structure
188  
189  When you just see:
190  - Passive absorption
191  - Less retention
192  - Model updates weakly
193  
194  **Implication:** Strategically inject DECISIONS, not just visibility.
195  
196  ### Decision Injection Points
197  
198  | Type | Frequency | Purpose |
199  |------|-----------|---------|
200  | **Strategic** | Every major direction | Lock in architectural frames |
201  | **Sampling** | Random 10-20% | Calibrate system, build model |
202  | **Escalation** | When system can't resolve | Catch edge cases |
203  | **Audit** | Periodic deep-dive | Verify system is working |
204  
205  ---
206  
207  ## Dynamic Adjustment
208  
209  The algorithm should learn:
210  
211  ```
212  IF human overrides system decision frequently on X-type problems:
213     → Increase weight for X-type
214     → Lower threshold for human involvement
215  
216  IF human consistently approves without changes:
217     → Decrease weight (system is calibrated)
218     → Raise threshold for human involvement
219  
220  Track: Override rate per problem type
221  Adjust: Weights and thresholds based on track record
222  ```
223  
224  ---
225  
226  ## Implementation
227  
228  ### Quick Assessment (Mental)
229  
230  Before any task, estimate:
231  ```
232  "Blast radius?" → Low / Medium / High
233  "How core is this?" → Peripheral / Important / Foundational
234  "Ambiguity?" → Clear / Some / High tension
235  "Error-prone?" → Unlikely / Possible / Likely
236  "Reversible?" → Easy / Moderate / Hard
237  
238  → Gut check maps to SOLO / LIGHT / STANDARD / TRIBAL / COUNCIL
239  ```
240  
241  ### Formal Assessment (For Important Decisions)
242  
243  ```markdown
244  ## Tribe Sizing Assessment
245  
246  **Problem:** [Description]
247  
248  | Variable | Score | Rationale |
249  |----------|-------|-----------|
250  | Blast Radius (B) | [0-1] | [Why] |
251  | Fundamentality (F) | [0-1] | [Why] |
252  | Tension (T) | [0-1] | [Why] |
253  | Error Likelihood (E) | [0-1] | [Why] |
254  | Reversibility (R) | [0-1] | [Why] |
255  
256  **Weight:** W = [calculated]
257  
258  **Configuration:** [SOLO/LIGHT/STANDARD/TRIBAL/COUNCIL]
259  
260  **Human Injection:** [None/See/Decide/Full]
261  ```
262  
263  ---
264  
265  ## Your Role in the Stack
266  
267  ### Always (Non-negotiable)
268  - See aggregate dashboard (Trust_F, thread status, gravity wells)
269  - Decide on W ≥ 0.8 problems
270  - Set strategic direction
271  
272  ### By Sampling (Calibration)
273  - Random spot-checks of lower-weight decisions
274  - Periodic deep-dives into system operation
275  - Override tracking to tune algorithm
276  
277  ### By Choice (Engagement)
278  - Deep-dive when curious
279  - Inject at any level when desired
280  - Pull any decision up to your level
281  
282  **The system should make it EASY to see everything and EFFICIENT to decide strategically.**
283  
284  ---
285  
286  ## The Promise
287  
288  > **Match firepower to problem. Match involvement to value.**
289  >
290  > Small problems get small tribes.
291  > Big problems get full councils.
292  > You see everything summarized.
293  > You decide what matters most.
294  > The system learns from your overrides.
295  
296  ---
297  
298  ## Related
299  
300  - **axioms**
301    - [[A3 Dynamic Pole Navigation]] - tribe size navigates dynamically based on problem
302      - shape:: "Life is the oscillation; death is fixing at either pole."
303    - [[A2 Recognition of Life]] - right-size for the situation (not over/under)
304  - **protocols**
305    - [[autonomous-exploration-tribes]] - what gets sized
306      - shape:: "Self-directed teams that explore without constant supervision."
307    - [[first-officer-protocol]] - FO uses this algorithm to decide spawn size
308      - shape:: "Per-thread metacognition. Track gravity wells."
309    - [[fractal-tribe-architecture]] - tribe sizing at each level
310      - shape:: "Same pattern at every level."
311    - [[model-allocation-strategy]] - model allocation per tribe size
312      - shape:: "Match model capability to task complexity."
313  - **enables**
314    - [[execution-autonomy-gradient]] - larger tribes for higher-stakes decisions
315  
316  ---
317  
318  *proto-014 | Tribe Sizing Algorithm | Match Firepower to Problem*