The mathematics page proves the core of EEML in theorems and proofs. This page tells exactly the same story, but in plain language, with an image where it helps. Nothing here is simplified to the point of being wrong; whoever wants to check the details clicks through to the original per section.
Imagine a gigantic catalogue in which every conceivable universe has one page: its shape, its particles, all its settings, completely described. In mathematics that catalogue is called the "configuration space". The question of the entire page, put simply: is there a rule that determines which pages are reality, and how much each page counts?
→ formal: Definitions 1.1 and 1.2
We demand three things of such a rule. One: if two things have nothing to do with each other, then the chance of both together is simply one chance times the other (as with two coin flips). Two: the rule makes no strange jumps; a tiny difference in universe gives a tiny difference in chance. Three: the "resistance" of two separate things together is the sum of their separate resistances, the way weight adds up when you carry two bags.
Nothing more is assumed. No physics, no particles, nothing.
Now the surprise. Combine those three demands, and mathematically only one kind of rule turns out to satisfy them: the chance must fall exponentially with the resistance. That is P = e⁻ˢ, and the proof is almost two centuries old (Cauchy, 1821): an equation of the type "adding on one side must become multiplying on the other" has only one decent solution, and that is the exponential.
What does "falling exponentially" mean in practice? That a little more resistance gives not a little less chance, but dramatically less. Every extra step of resistance divides the chance by the same number. Compare: ten steps further is not ten times as unlikely, but thousands of times.
And out of this rolls a famous piece of physics: the principle of least action (the rule from which all equations of motion from Newton to Einstein follow) is now no longer an assumption but a consequence. The configuration of least resistance receives the highest preference, full stop.
Important, and it sits at the top of the mathematics page as the reading key: paths of greater resistance are not impossible, they just weigh almost nothing. Everything stable exists; the efficient carries the weight.
→ formal: Theorem 3.1, Corollary 3.2, reading key
So what does that resistance S consist of? After all the hardening work: exactly two parts.
The action: this is ordinary physics. How "natural" the behaviour of a universe is, expressed in the classical laws of motion.
The description cost: how much information (bits) do you need to write the universe down completely? Think of a zip file: a simple, regular universe compresses to a small file; a messy universe with a thousand separate settings does not. Small file means little resistance.
There used to be a third term (stability), but it turned out to be double work. What is unstable (things that fall apart) does not count as "expensive" but simply does not exist: that is an admission requirement, not a fine. And "fine-tuning" (parameters that must sit eerily precisely right) is just extra description cost: writing down a very precise setting takes many decimal places, hence many bits.
And the best part: the exchange rate between the two parts is not free to choose. One extra bit of description means exactly a factor of two less weight. Why? Because a bit simply is a choice between two. Any other exchange rate would amount to secretly using a different measure of information. So there are zero adjustable dials in the formula, and that is rare and valuable: dials are where one normally cheats.
→ formal: Lemma 4.1, Definitions 6.1 and 7.1, Reduction 10.6
A tempting picture: reality "rolls" into the landscape like a marble and comes to rest in the deepest valley. That picture is demonstrably wrong. A rolling marble ends up in the nearest valley, not the deepest; where you end depends on where you started. And "where you started" is exactly the kind of external condition EEML does not want and does not have.
The right picture: not a marble but mist. Mist settles over the whole landscape at once, and more mist lies in deep valleys than on slopes. Nobody chooses, nothing rolls: the distribution simply is, and it is completely fixed by the shape of the landscape. Deep valleys (little resistance) receive exponentially more; broad, flat valleys receive something extra on top, because there are more ways to lie there.
Note where this picture lives. The mist distribution itself is a-local: it lies nowhere in particular and holds at no particular moment, because "where" and "when" are words from inside a universe. Inside such a universe the same weighing does work locally: step by step, the way lightning picks its channel metre by metre.
Now the biggest thinking step of the page, in three questions you must keep apart.
What exists? Everything stable and self-consistent. That is a lot. Existence is free; you do not have to win it.
How much does each count? That is what the mist distribution says: every existing universe has a weight, and the simplest, smoothest universes carry almost everything between them.
Which one is "the real one"? Here the page proves something sharp. Suppose "being actual" were a physical fact, a kind of lamp burning at exactly one universe. Then something must switch that lamp on, and then you may ask: who chose that? Back to square one, endlessly. The only other possibility: "actual" works like the word "here". If you say "here" and I say "here", we are both right, even though we stand in different places. The word points at nothing special in the world; it points from a standpoint. So too "this universe is the real one": true from within, for every existing universe, without a lamp burning anywhere. And since there is no outside, the lamp option falls away. That is a theorem with a proof (there is no third flavour), not a matter of taste.
→ formal: Addendum 10.3, Theorem 10.4, Corollary 10.5
If infinitely many universes exist, does the total amount of "mist" still add up neatly to 100%? For a long time this was the scariest gap. The answer turned out to lie in information theory: there is a classical law (the Kraft inequality) which says that if you assign weights by description length (halving per bit), the sum is guaranteed to stay finite. The intuition: there exist only few short descriptions (there are only two files of one bit, four of two bits, and so on), so the simple front is thinly populated, and everything complicated behind it weighs so little that together it is still almost nothing. The preference for simplicity literally rescues the mathematics: that is why one may sum over everything-that-exists.
A distribution over universes is nice, but what does it predict for us? For that one extra ingredient is needed, and it is stated honestly as an assumption: the question "which universe is the real one?" can only be asked in a universe that understands itself a little, that carries somewhere inside it (in heads, in books) a summary of its own law. And there sits a genuine small theorem: such a summary must fit inside the universe, so the universe must be summarisable, hence simple. Question-askers by definition live in compressible universes. And compressible universes are exactly the heavy ones. The conjecture (open, labelled as such): "understands itself" and "weighs the most" are ultimately the same club. If that holds, it is no accident that we live in a universe that can be captured in a handful of formulas.
→ formal: Reading rule 10.8, open problem 11.7
Two classic objections, briefly. "If the law is perfect, why is the world so asymmetric?" Because the law and the inhabitant live on different layers: if several universes are exactly equally good, the distribution is neatly symmetric, but each member individually is skewed, and we live in one member. "If simplicity wins, why is the world so complex?" Because simplicity concerns the law, not what the law produces: the rules of chess fit on one sheet, and yet there are more games than atoms in the universe. Simple law, rich world: not a tension but exactly what you expect.
Last step. The selection principle says "the simplest, most stable form wins". But then you may ask: is the principle itself the simplest, most stable form of a selection principle? And that turns out to hold, building block by building block: the action used is provably the only simple option of its kind, the information measure is the most economical there is, and so on. The law satisfies its own criterion: f(EE) = EE. That the reasoning runs in a circle is here not a flaw but the whole point, because every non-circular law needs an external ground, and that ground another ground, into infinity.
The page does not hide it: a few orthogonality details, the precise interplay between the quantum weighting and the bit weighting, the behaviour at the boundary (perfectly flat directions), and the "understands itself" conjecture. The big one that used to stand here, the computation of the fine-structure constant on K3, is gone: that computation was actually carried out in the summer of 2026, and the answer was no, three times over. What that yielded is in the new section below, because it may well be the most important result of the entire framework.
→ formal: Open problems 11.1 to 11.7 · → what would refute EEML? Testing & Limits
There is a famous number in physics, 137, which sets how loudly matter and light talk to each other. Generations of physicists have tried to compute it from pure principles. So did we: three times, with the criteria on paper in advance so no one could cheat. Three misses, honestly documented. But then something strange happened. The same computational machinery was asked a different question: not "how strong is the force?" but "which particle world is the cheapest one in which someone can exist to ask the question?" And there an answer did come out. One answer. Exactly our particle world, down to details only measured after 1998.
The lesson splits all physical facts into two kinds. Some are law: they cannot be otherwise, in any member of the catalogue that can carry observers. Which charges exist, how they relate, that exactly one long-range force remains for chemistry: all law, all derived. Other facts are address: they differ from member to member, and our value only tells us where in the catalogue we live. Just as the physics of letterboxes does not explain why your house number is 14, and does not have to. The strength of the famous number is address. The number of particle families probably too, and there the framework even dares to name a price: our address is roughly a thousand times less likely than the cheapest valid one, and whoever wants to explain "three families" one day now knows exactly how much evidence that must produce.
Is that giving up? No, because addresses have their own physics: that of distributions. You can test whether our address is a normal one (it turns out to be: the famous number sits comfortably inside the window where observers can exist at all), and the law side remains hard and testable: the framework now predicts, for instance, that neutrinos are their own antiparticles, and if experiments definitively rule that out, the whole story falls. A theory that can say precisely where its authority ends, and that has surveyed that boundary itself with three of its own failures as survey markers, is rarer than a theory that promises everything.
→ formal: mathematics §12, Theorem S′ · → the full story: testing, section 2
There is a catalogue of all conceivable universes; three rules you can hardly refuse force every universe to weigh according to its resistance, exponentially; resistance is physical action plus zip-file size, without a single dial; everything stable exists, the simple carries almost all the weight, and the sum works out thanks to that same preference for simplicity; "this is the real universe" is true the way "here" is true; the law arranging all of this is itself the simplest specimen of its own kind, and therefore needs no one above it; and the law itself honestly tells us which numbers it fixes (the structure of the particle world) and which are merely our address (the strengths, and probably the number three).