HOW INTELLIGENCE WORKS · SPIRAL PITCH · LEARNING RATE · CAPABILITY · TRANSFER · eduKateSG

The Pitch of the Spiral

Why going around more often does not necessarily mean becoming more intelligent.

An intelligence can complete thousands of feedback loops and improve almost nothing. Another can complete one carefully designed loop and permanently change what it is capable of doing. Loop count is therefore not the same as learning.

The Spiral Closed Loop gives us a useful geometric language for this difference. A spiral has not only a direction and a speed. It has a pitch: how much durable displacement occurs per revolution.

Loop speed tells us how quickly intelligence returns. Spiral pitch tells us how much useful capability survives the return.

1. Two factories can run the same number of cycles

Imagine two car manufacturers. Both collect warranty data every week. Both hold engineering reviews. Both issue corrective actions. Both can truthfully say they run closed-loop quality systems.

Factory A repeatedly fixes symptoms. A seal leaks, so technicians replace it. A software fault appears, so support resets the system. A vibration occurs, so a service bulletin tells dealers how to reduce it.

Factory B traces recurring failures to deeper mechanisms. It changes the interface that creates the leak, redesigns the software state transition, alters the tolerance stack that creates the vibration, verifies the changes and transfers the lessons into later platforms.

Both factories loop. Factory B gains more height from each meaningful revolution.

2. Correction depth changes spiral pitch

A correction can occur at several depths. The shallowest response repairs the visible outcome. A deeper response changes the process that produced it. Deeper still, the system may revise the model, assumption or architecture that made the process seem sensible.

Symptom repair → process repair → model repair → architecture repair

Deeper is not automatically better. A loose fastener does not require a philosophical reconstruction of the vehicle. Intelligence includes matching correction depth to causal depth. But repeated shallow repairs to the same mechanism are evidence that the loop is closing without climbing.

3. Practice can have low pitch

Education makes low-pitch looping easy to see. A student completes fifty nearly identical questions. Accuracy rises because the surface becomes familiar. Then the examination changes the wording and performance collapses.

The learner completed many loops: attempt, mark, correction, repeat. Yet little transferable structure survived.

Now compare a smaller set of problems in which the learner must identify the underlying relationship, explain why a method applies, encounter counterexamples, retrieve the method after delay and use it under a changed surface. Fewer repetitions can produce greater durable displacement.

Repetition counts revolutions. Learning asks what changed because of them.

4. Retained learning creates vertical displacement

The previous article introduced Retained Learning. That mechanism gives spiral pitch its memory component.

If an insight disappears after the episode, the system may have moved temporarily but not durably. If a repair becomes a reusable schema, procedure, design principle or transferable capability, the next loop begins from a different state.

We can write the idea simply:

Spiral displacement ≈ useful change × retention × transfer

This is not a literal universal equation. It is an architectural reminder. A large immediate improvement multiplied by near-zero retention produces little future advantage. A well-retained lesson with zero transfer may remain trapped in one case.

5. Learning rate and spiral pitch are different

Learning Rate Control asks how much one piece of evidence should change a model. Spiral pitch asks how much useful capability the whole completed loop leaves behind.

A large model update can produce negative pitch if the evidence was misleading. A small update can eventually produce large cumulative gains if it is reliable and compounds across many loops.

This distinction protects us from equating dramatic change with intelligent change.

6. Creativity can increase pitch—or destroy it

Creativity can move the dots outward and discover a route that escapes years of incremental optimisation. A new architecture can produce a large jump in capability.

But creative change can also reset learning before evidence accumulates. If a company redesigns constantly, a school replaces programmes annually or a learner changes methods every week, the system may generate novelty faster than it can close its loops.

High creative output is therefore not high spiral pitch by itself. The candidate still needs World Return, evaluation, retention and transfer.

7. Measurement can mistake motion for height

Many systems measure activity because activity is visible. Number of experiments, lessons, meetings, iterations, releases, practice questions, reports and improvement projects are easy to count.

But these are revolutions, not necessarily displacement.

A better measurement question is: what can the system now do reliably that it could not do before? What error no longer recurs? What class of problem became easier? What knowledge became transferable? What uncertainty was genuinely reduced?

8. Negative pitch

The spiral can descend. A loop can preserve the wrong lesson, optimise a proxy, remove useful redundancy or consolidate a misconception.

Negative pitch is therefore retained deterioration. The system does not merely fail once; it changes itself in a way that makes later performance worse.

This is more dangerous than an isolated error because versioning now propagates the mistake forward.

9. Nested loops have different pitches

A fast inner loop may have tiny pitch and still be essential. A vehicle controller makes countless small corrections whose purpose is stability, not redesign. A slower engineering loop may produce larger structural changes. A product-strategy loop may occasionally redefine the whole architecture.

So pitch should be judged relative to the job of the loop. Stability loops often seek low displacement around a desired state. Learning loops seek durable improvement. Creative strategic loops sometimes seek discontinuous jumps.

Intelligence coordinates these different geometries rather than forcing every loop to behave the same way.

10. Civilisation can have enormous pitch

At civilisation scale, one discovery can permanently change the starting point for billions of later loops. Writing, positional number systems, germ theory, electricity, computing and other durable knowledge structures alter what later generations can do without rediscovering the entire path.

The civilisational ratchet amplifies spiral pitch because useful learning can be inherited by people who did not participate in the original experiment.

Education, libraries, standards, institutions and digital systems become mechanisms for transmitting vertical displacement across generations.

11. Intelligence should optimise learning yield, not loop count

This gives us a more useful design question for intelligent systems:

How much reliable, transferable, retained capability does this loop produce for the cost of running it?

That question brings together Value of Information, Cognitive Effort Allocation, Exploration–Exploitation Arbitration, Outcome Evaluation and Retained Learning. Some loops are not worth closing. Some need more evidence. Some need deeper diagnosis. Some need to stop repeating and change the representation itself.

12. The geometry is becoming complete

Our Intelligence architecture now has several dimensions. Dots describe possibilities. Connections describe structure. Creativity expands the field. World Return closes the loop. Retention allows versioning. The spiral describes cumulative change. Nested loops describe scale. Loop speed describes time. Spiral pitch describes how much durable capability is gained per revolution.

That gives us a stronger way to recognise intelligence developing through time. Do not merely ask whether the system is busy, adaptive or repeatedly updating.

Ask whether each return from reality leaves the next version meaningfully more capable.