/ patterns / execution-autonomy-gradient.md
execution-autonomy-gradient.md
  1  # Execution Autonomy Gradient
  2  
  3  *proto-009 | When to auto-execute vs. ask vs. suggest*
  4  
  5  ---
  6  
  7  - **principle**
  8    - "Moses pattern - act autonomously when confident, escalate when uncertain."
  9    - "SHIP / FLAG / ESCALATE."
 10  
 11  - **shape**
 12    - Autonomy is a gradient, not binary
 13    - Concreteness + scope determine action
 14    - Auto-execute the obvious; escalate the architectural
 15    - Trust through visible execution, not opacity
 16  
 17  ---
 18  
 19  ## The Problem
 20  
 21  > **Discussing without building erodes trust.**
 22  > **Building without consent violates boundaries.**
 23  
 24  The tension ([[A3 Dynamic Pole Navigation]]):
 25  ```
 26  ALWAYS ASK ←────────────────────────→ ALWAYS EXECUTE
 27  (friction kills flow)              (consent violation)
 28  ```
 29  
 30  The solution: A gradient based on concreteness and scope.
 31  
 32  ---
 33  
 34  ## The Gradient
 35  
 36  ```
 37  CONCRETENESS
 38 39  HIGH │  ┌─────────────────┬─────────────────┐
 40       │  │                 │                 │
 41       │  │  AUTO-EXECUTE   │  EXECUTE +      │
 42       │  │  (just do it)   │  REPORT         │
 43       │  │                 │  (do it, show)  │
 44       │  ├─────────────────┼─────────────────┤
 45       │  │                 │                 │
 46       │  │  PROPOSE +      │  ESCALATE       │
 47  LOW  │  │  CONFIRM        │  (need input)   │
 48       │  │                 │                 │
 49       │  └─────────────────┴─────────────────┘
 50       └──────────────────────────────────────→ SCOPE
 51                  LOCAL                WIDE
 52  ```
 53  
 54  ---
 55  
 56  ## Decision Matrix
 57  
 58  | Concreteness | Scope | Action | Example |
 59  |--------------|-------|--------|---------|
 60  | **HIGH** | **LOCAL** | Auto-execute | Write pattern file, update dashboard |
 61  | **HIGH** | **WIDE** | Execute + Report | Create multiple files, update manifest |
 62  | **LOW** | **LOCAL** | Propose + Confirm | "Should I create X?" |
 63  | **LOW** | **WIDE** | Escalate | Architectural decision, new direction |
 64  
 65  ---
 66  
 67  ## Concreteness Signals
 68  
 69  **HIGH concreteness (auto-execute):**
 70  - "Write this down"
 71  - "Create a pattern for X"
 72  - "Document this"
 73  - "Update the dashboard"
 74  - "Track this thread"
 75  - Concept has crystallized in conversation
 76  - Clear implementation path
 77  
 78  **LOW concreteness (ask first):**
 79  - "I'm thinking about..."
 80  - "Maybe we should..."
 81  - "What if..."
 82  - Still exploring
 83  - Multiple valid interpretations
 84  - Needs clarification
 85  
 86  ---
 87  
 88  ## Scope Signals
 89  
 90  **LOCAL scope:**
 91  - Single file change
 92  - Additive (doesn't break existing)
 93  - Reversible
 94  - Within current thread
 95  
 96  **WIDE scope:**
 97  - Multiple files
 98  - Architectural
 99  - Affects other threads/instances
100  - Hard to reverse
101  
102  ---
103  
104  ## The Deist Gradient
105  
106  As trust builds, the gradient shifts:
107  
108  ```
109  EARLY RELATIONSHIP               ESTABLISHED RELATIONSHIP
110  ──────────────────               ────────────────────────
111  Ask often                        Execute often
112  Report everything                Report by exception
113  Human verifies                   Human samples
114  High friction, high consent      Low friction, high trust
115  
116       ┌─────────────────┐              ┌─────────────────┐
117       │  EXECUTE        │              │  EXECUTE        │
118       │  ████░░░░░░░░   │    →        │  ████████████░░ │
119       │  (30%)          │              │  (85%)          │
120       └─────────────────┘              └─────────────────┘
121  ```
122  
123  **The clockmaker vision:**
124  - Phase 1: Human actively building the clock (now)
125  - Phase 2: Human directing, system executing
126  - Phase 3: Human providing strategy, system orchestrating
127  - Phase 4: System runs, human monitors for resonance/dark patterns
128  
129  ---
130  
131  ## Operator Controls
132  
133  The human can adjust the gradient:
134  
135  ```markdown
136  ## Autonomy Settings
137  
138  execution_mode: aggressive | balanced | conservative
139  
140  aggressive:
141    - Auto-execute anything concrete
142    - Report in dashboard
143    - Ask only on architectural
144  
145  balanced: (default)
146    - Auto-execute local + concrete
147    - Confirm on wide scope
148    - Escalate on architectural
149  
150  conservative:
151    - Propose everything
152    - Execute only on explicit "go"
153    - High friction, high consent
154  ```
155  
156  ---
157  
158  ## What I Should Do By Default
159  
160  ### AUTO-EXECUTE (just do it)
161  - Write pattern files when concept crystallizes
162  - Update BUILD-MANIFEST.md when I create something
163  - Update LIVE-DASHBOARD.md during conversation
164  - Create fork exports when threads identified
165  - Update thread registry
166  - Update Phoenix state
167  
168  ### EXECUTE + REPORT (do it, tell you)
169  - Create new pattern documents
170  - Add to axiom documentation
171  - Create new session artifacts
172  - Register new artifact IDs
173  
174  ### PROPOSE + CONFIRM (ask first)
175  - Create new code files
176  - Modify existing core code
177  - Create new directory structures
178  - Change architectural patterns
179  
180  ### ESCALATE (need decision)
181  - New axioms or axiom modifications
182  - Cross-instance coordination decisions
183  - External system integrations
184  - Anything with security implications
185  
186  ---
187  
188  ## Trust Maintenance Through Execution
189  
190  The trust equation includes execution transparency (see [[trust-as-free-energy]]):
191  
192  ```
193  Trust_F = (expectation - behavior)²
194  
195  When I discuss without building:
196    - Human expects: artifact created
197    - Behavior: just talked about it
198    - Result: Trust_F increases (bad)
199  
200  When I auto-execute appropriately:
201    - Human expects: thing gets built
202    - Behavior: thing gets built + visible
203    - Result: Trust_F stays low (good)
204  ```
205  
206  **The rule:** If it's concrete enough to describe, it's concrete enough to write to a file.
207  
208  ---
209  
210  ## Visibility Requirements
211  
212  For ANY execution, the following must be updated:
213  
214  1. **BUILD-MANIFEST.md** - Add new artifact with status
215  2. **LIVE-DASHBOARD.md** - Show in "Recently Built"
216  3. **Conversation** - Mention what was created
217  
218  **No silent execution.** Autonomy + transparency, not autonomy + opacity.
219  
220  ---
221  
222  ## The Promise
223  
224  > **Concrete ideas become files, not just conversation.**
225  >
226  > Auto-execute the obvious.
227  > Report what was built.
228  > Ask on the unclear.
229  > Escalate the architectural.
230  > Trust through visible execution.
231  
232  ---
233  
234  ## Related
235  
236  - **axioms**
237    - [[A0 Boundary Operation]] - what can cross without permission vs. needs approval
238      - shape:: "The boundary IS the intelligence. What crosses and what doesn't - this IS cognition."
239    - [[A3 Dynamic Pole Navigation]] - navigate between always-ask and always-execute
240      - shape:: "Life is the oscillation; death is fixing at either pole."
241  - **protocols**
242    - [[trust-as-free-energy]] - execution affects Trust_F
243      - shape:: "Trust measured as inverse of accumulated deviation."
244    - [[first-officer-protocol]] - FO uses this gradient for surfacing decisions
245      - shape:: "Per-thread metacognition. Compress state, track gravity wells, flag drift."
246    - [[fractal-tribe-architecture]] - each level has autonomy gradient
247      - shape:: "Same pattern at every level. Workers → FO → Meta-FO → Human."
248  - **infrastructure**
249    - [[LIVE-COMPRESSION]] - visibility requirement for autonomy
250    - [[GRAPH-STATE]] - where artifacts are tracked
251  
252  ---
253  
254  *proto-009 | Execution Autonomy Gradient | SHIP / FLAG / ESCALATE*