Change Levers¶
Five physical places a change can act. Each lever has a typical effect, a typical failure and a check.
| Lever | What changes | Typical effect | Typical failure | Check |
|---|---|---|---|---|
| Flow | Which gradient is accessed; volume or source | More or different output | Exhausts gradient; overloads downstream constraints | Can downstream constraints carry it? |
| Constraint | Structure channeling the flow: process, tool, role, rule, architecture | Less heat per unit of work; new kinds of work | Rigidity; maintenance cost outgrows benefit | What does the new constraint cost to maintain? |
| Information | What is known, stored, shared; models, docs, skills | Higher leverage; better steering | Information without maintenance decays; overload | Who maintains it, and does the flow it steers fund that? |
| Selector | Who decides continuation; what metric they use; feedback speed | Changes what the system evolves toward | Proxy drift; gaming; capture | Does the new metric track the real flow? |
| Variation | How variants appear; cost of trying; who may try | Adaptability | Novelty without consolidation (if selection is weak) | Is selection strong enough to filter the new variants? |
Rules of thumb¶
- L1. Selector changes have the largest delayed effect. They redirect evolution, not just output. [Hypothesis]
- L2. Constraint changes without information changes revert. People rebuild the old constraint from what they know. [Hypothesis]
- L3. Cutting variation looks like efficiency now and appears as brittleness later. [Framing]
- L4. Adding flow to a constrained system raises waste first. [Hypothesis]
- L5. Pair every lever with its selector check: who will notice whether the change worked, and how fast?