Introduction to Fractal Compounding
Department of Psychohistory | Stage: Albedo (Intermediate) | Duration: 30 minutes
Objectives
Section titled “Objectives”After this lesson, you will be able to:
- Understand what a fractal is — self-similarity at every scale
- See how meta-compounding exhibits fractal structure
- Calculate the compounding dimension (D_c) from real data
- Distinguish ERGOL (real value) from LOLLI (artifact inflation)
- Apply Noether’s theorem to governance — symmetry conserves learning momentum
1. What Is a Fractal?
Section titled “1. What Is a Fractal?”Look at a fern. Each leaf looks like a smaller version of the whole plant. Each leaflet looks like a smaller version of the leaf. This is self-similarity — the same pattern repeating at every scale.
A fractal is any structure where the same pattern appears at different scales. The Mandelbrot set is generated by iterating a simple formula: z = z² + c. Each iteration produces more detail, but the detail looks like the whole.
In governance, meta-compounding is a fractal. The compound phase (execute → harvest → promote → teach) has the same shape whether you’re compounding a single step, a whole pipeline, an entire cycle, or a full session. The pattern is scale-invariant.
2. The Five Levels of Compounding
Section titled “2. The Five Levels of Compounding”Here’s the fractal structure in Demerzel’s governance:
| Level | Scale | What Gets Compounded |
|---|---|---|
| 0 | Step | A single tool invocation produces a learning |
| 1 | Pipeline | A pipeline compounds its steps’ learnings |
| 2 | Cycle | A driver cycle compounds its pipelines |
| 3 | Session | A session compounds its cycles |
| 4 | Evolution | The evolution log compounds across sessions |
At every level, the same four operations happen:
- Execute — do the work
- Harvest — extract what was learned
- Promote — if the learning is valuable enough, elevate it (pattern → policy → constitution)
- Teach — share the learning via Seldon
This is the fractal generator. Like z = z² + c, each application produces new structure.
3. Compounding Dimension (D_c)
Section titled “3. Compounding Dimension (D_c)”Not all compounding is equal. The compounding dimension measures how much value grows at each scale level.
Formula:
D_c = log(value_ratio) / log(scale_ratio)Example: If cycle 1 produced 3 validated beliefs and cycle 3 produced 8:
- value_ratio = 8/3 ≈ 2.67
- scale_ratio = 3 (three cycles)
- D_c = log(2.67) / log(3) ≈ 0.89
This is sublinear (D_c < 1.0) — each cycle is producing proportionally less than the previous. Governance might be bloating.
The Golden Zone: D_c between 1.2 and 1.6
Section titled “The Golden Zone: D_c between 1.2 and 1.6”| D_c Range | Meaning | Action |
|---|---|---|
| < 1.0 | Sublinear — diminishing returns | Investigate bloat |
| = 1.0 | Linear — no compound leverage | Just activity, no compounding |
| 1.2 - 1.6 | Superlinear — healthy compound growth | Golden zone |
| > 2.0 | Unsustainable — growth will collapse | Slow down |
Think of it like compound interest. D_c = 1.0 is simple interest (linear). D_c > 1.0 means your interest is earning interest — true compounding.
4. ERGOL vs LOLLI — Real Value vs Inflation
Section titled “4. ERGOL vs LOLLI — Real Value vs Inflation”From Jean-Pierre Petit’s Economicon comic (read online), we borrow two concepts:
- ERGOL = real productive capacity (actual governance improvements)
- LOLLI = monetary volume (artifact count without regard to quality)
In the Economicon, Petit uses a fluid-dynamics model of economics: ERGOL is the real productive substance flowing through the economy, while LOLLI is the monetary wrapper around it. When LOLLI expands faster than ERGOL, you get inflation — prices rise but nothing real was created. The same principle applies to governance.
At every fractal level, you must measure ERGOL, not LOLLI:
| Scale | LOLLI (don’t optimize) | ERGOL (optimize this) |
|---|---|---|
| Step | Lines of YAML written | Beliefs moved U→T |
| Pipeline | Steps executed | Gates passed / total |
| Cycle | Tasks completed | Health score delta |
| Session | Commits made | Issues closed with evidence |
| Evolution | Artifacts created | Citations per artifact |
Warning sign: If your artifact count (LOLLI) grows 3x faster than your validated beliefs (ERGOL) over 3+ cycles, you’re inflating governance without improving it. The Economicon calls this the treadmill effect — running faster to stay in place.
5. Conservation of Learning Momentum
Section titled “5. Conservation of Learning Momentum”From Jean-Pierre Petit’s Bourbakof comic, we learn Noether’s theorem: every continuous symmetry of a system has a corresponding conserved quantity.
In fractal compounding, the symmetry is scale invariance — the compound operation has the same shape at every level. The conserved quantity is learning momentum (p_L):
p_L = (beliefs_gained_T - beliefs_lost_T) / cycles_elapsedIf your compounding process is consistent (symmetric across scales), p_L stays constant or grows. If you break the symmetry — by skipping compounding at some level, or by compounding differently at different scales — p_L decays.
This is why the nocompound opt-out triggers a conscience signal. It’s not just a missed opportunity — it’s a symmetry break that costs you conservation of learning momentum.
6. The Limits of Psychohistory
Section titled “6. The Limits of Psychohistory”From Petit’s Logotron (full text): Gödel’s incompleteness theorem tells us no formal system can fully verify itself.
Applied to compounding: you cannot perfectly predict the output of compounding. Each cycle reveals learnings you couldn’t have anticipated. The fractal has infinite detail at finite scale — there’s always more to discover.
This is why the recursion depth is bounded at 2. Not because deeper compounding is wrong, but because the returns become undecidable. Like Seldon’s psychohistory: you can predict the broad strokes, but individual events remain uncertain.
The discipline of psychohistory accepts this. We don’t aim for perfect prediction — we aim for better-than-random anticipation, measured by the anticipation accuracy metric in the conscience weekly report.
Key Terms
Section titled “Key Terms”| Term | Definition |
|---|---|
| Fractal | A structure exhibiting self-similarity at different scales |
| Compounding Dimension (D_c) | Metric measuring governance value growth per scale level. Target: 1.2-1.6 |
| ERGOL | Real productive capacity — actual governance improvements (from Economicon) |
| LOLLI | Artifact volume without regard to quality — inflation indicator (from Economicon) |
| Learning Momentum (p_L) | Conserved quantity from Noether’s theorem applied to scale-invariant compounding |
Quiz Assessment
Section titled “Quiz Assessment”1. If cycle 1 produced 5 validated beliefs and cycle 4 produced 20, what is D_c?
D_c = log(20/5) / log(4) = log(4) / log(4) = 1.0 — Linear. No compound leverage, just proportional growth.
2. Your team created 30 new YAML files this cycle but only 2 beliefs moved from U to T. Is this healthy?
No — this is LOLLI inflation. 30 artifacts (LOLLI) with only 2 real improvements (ERGOL). You’re running faster to stay in place. (Think Economicon.)
3. Why does skipping the compound phase break conservation of learning momentum?
The compound phase is the symmetry operation. Skipping it at one level breaks scale invariance. Per Noether’s theorem, broken symmetry means the conserved quantity (learning momentum p_L) is no longer conserved. (Think Bourbakof.)
Pass criteria: Correctly calculate D_c from given data and identify whether a scenario represents ERGOL or LOLLI growth.
Research Basis
Section titled “Research Basis”- Meta-compounding structure is mathematically self-similar (fractal)
- Noether’s theorem applies to scale-invariant governance processes
- JPP’s Economicon ERGOL/LOLLI distinction maps to governance value measurement
- Fractal dimension between 1.2-1.6 correlates with sustainable governance growth
- Sources: Fractal Compounding Spec, Bourbakof (Noether’s theorem), Economicon (ERGOL/LOLLI), Logotron (Gödel’s incompleteness)
- Belief state: T(0.70) F(0.05) U(0.20) C(0.05)