The universe learns to process information

For billions of years, the universe became increasingly structured. But complexity alone isn't the most interesting pattern. Something else happened.

The universe progressively produced systems capable of:

• storing information
• copying information
• transmitting information
• processing information
• generating information
• acting on information
• learning from information
• improving the systems that process information

The story may therefore be better described as the evolution of information processing.

Rather than asking only “How did the universe become more complex?” we can ask:

“How did the universe repeatedly produce new ways for information to influence what happens next?”

A linear timeline would crush all of human history into the last pixel. Time here is logarithmic. Click any milestone for its date, uncertainty, substrate, and feedback mechanism. Dates are approximate; the origin of life and language carry large uncertainty.

At first glance, the intervals between transitions shrink dramatically. Millions of years. Then thousands. Then centuries. Then decades. Then years.

But there is a trap.

When time is plotted logarithmically, events naturally compress toward the present. A shrinking interval is not, by itself, evidence of accelerating evolution.

We tested this using a null model in which milestones were distributed randomly across logarithmic time. The artificial timelines produced surprisingly similar apparent acceleration. Try it:

Conclusion: the historical timeline alone does not establish a universal acceleration law. We shouldn't cherry-pick a curve because it looks compelling. The better question is: what mechanism could produce genuine acceleration beyond the effect of logarithmic time? That leads to feedback.

1. PhysicsMatter organizes according to physical dynamics.
2. ChemistryInteractions create reaction networks.
3. AutocatalysisSome processes promote their own continuation.
4. LifeInformation becomes heritable.
5. EvolutionVariation + selection produce improved descendants.
6. Nervous systemsSystems can adapt within their own lifetimes.
7. LearningExperience changes future behavior.
8. Social learningInformation moves between individuals.
9. Cumulative cultureKnowledge survives individual lifetimes and accumulates.
10. WritingInformation survives outside biological memory.
11. PrintingInformation becomes cheaply reproducible.
12. ComputingInformation processing becomes programmable.
13. InternetProcessors become globally interconnected.
14. Machine learningParameters of information-processing systems can be optimized automatically.
15. AI-assisted R&DAI begins participating in the development of AI.
16. AI-led R&DAI performs substantial portions of the AI-development loop.
17. Recursive R&D?AI materially improves the process that produces better AI.

One of the deeper questions in the evolution of information is whether the universe is actually becoming more complicated — or whether increasingly sophisticated systems are learning to extract, encode, and generate more complexity from the same underlying rules.^[1,4]

A conceptual framework — not an established scientific law. The term “ruliology” comes from Wolfram’s study of simple programs^[1,4]; the three-part distinction and its application here are our own.

The individual historical events are documented; grouping them as a progression of rule types is this page's interpretation.

Does the universe progressively generate systems capable of generating increasingly powerful effective rules from relatively simple underlying rules?

The fundamental laws of physics may not need to become more complicated for the effective complexity of the universe to increase.

Simple rules, iterated through enormous state spaces and interacting with memory, feedback, selection, and information storage, can generate extraordinarily complex behavior.^[3,4]

Conway's Game of Life is a useful intuition: simple underlying rules can generate structures capable of memory, computation, self-replication, and complex interactions.^[4]

The complexity we observe in the universe may therefore come less from increasingly complicated fundamental laws and more from the enormous number of ways those laws can interact. (Our extrapolation — the cited sources define the concept, not this conclusion.)

The important question is therefore not simply whether AI becomes “intelligent.”

The more measurable question is whether AI becomes capable of materially increasing the rate at which the systems responsible for improving AI themselves can improve.

A conceptual research framework — not a validated scientific metric.

“AGI” is difficult to define and difficult to measure. The more useful question may be: when does AI become capable of materially accelerating the production of better AI?

Is AI-development throughput increasing?

Is greater AI R&D throughput causing greater future throughput? That is the feedback signal.

This is not a standardized industry metric. It is a proposed measurement.

Did AI contribute to making the next generation of AI better?

If G > 1, AI is already accelerating AI development relative to the chosen baseline.

The really interesting condition is dG/dt > 0: AI is not merely helping — its ability to help is itself increasing.

AI capability → AI R&D capability → AI-generated discoveries → better AI systems → greater capability → repeat. The difference between these two loops is not simply intelligence. It is who closes the loop.

Biological evolution faces enormous physical constraints. Generations take time. Reproduction takes time. Populations are finite. Mutations are relatively rare.

Software has radically different properties. An AI system can potentially be:

So the relevant quantity may not be “How intelligent is one AI?” but “How much useful optimization can the entire AI ecosystem produce per unit of calendar time?”

Current evidence suggests AI is moving into this territory but has not clearly crossed the recursive threshold. These figures are specific to their sources and methods — not universal constants, and not evidence that recursive self-improvement has occurred.^[5,6,7,8]

Conceptual regions plotted across autonomy, R&D capability, and recursive contribution (depth). Tilt the depth axis; click a region.

We began with a scientific question: does information processing accelerate over cosmic history? The data does not yet establish a universal acceleration law.

So we asked what mechanism could produce genuine acceleration. The answer we began investigating was feedback.

But eventually we have to ask: why does any of this exist?

The universe has produced physical systems capable of constructing representations of the universe itself. In that limited, literal sense:

The universe has produced systems through which it can model itself.

This is an interpretation — not evidence of cosmic intention.

Yet a human can design a dam that redirects a river. Why? Because information allows matter to be organized. A small physical system can use a model of a much larger system to influence it.

Life does this. Brains do it. Culture, technology, and computers do it. AI increasingly does it.

We can describe information processing externally. But consciousness asks a different question: why is there something it is like to be the system doing the processing?

This page does not choose among these. The uncertainty is part of the story.

If consciousness is emergent, then at some point the universe crossed a remarkable threshold:

And now we are here.

Perhaps the universe has no predetermined purpose. Perhaps purpose is something it eventually became capable of producing.

Perhaps the deepest transition isn't human intelligence → artificial intelligence, but a universe that evolves → a universe containing systems capable of increasingly directing evolution.

For most of cosmic history, the universe happened. Then life appeared. Then intelligence. Then foresight. Then technology. Now increasingly capable information-processing systems can participate in deciding what happens next. AI could eventually extend that process.

And perhaps, eventually: Intelligence improves intelligence.

If that happens, the universe will have produced something unusual: not merely a system that adapts, learns, or designs machines — but a system capable of participating in the process by which increasingly capable information-processing systems come into existence.

Whether that transition happens, how quickly, and what follows remain open questions.

Perhaps the deeper pattern is not that the fundamental rules of the universe are becoming increasingly complicated.

Perhaps the universe is progressively producing systems capable of exploring more of what those rules permit.

Each transition adds another layer of information processing, memory, feedback, and control.

The universe may not be changing its rules.
It may be producing systems capable of creating increasingly powerful rules of their own.

Philosophical speculation — not a scientific conclusion.