Six Birds Theory: How “Things” Show Up When You Squint at Reality

I’m going to start with a confession that feels obvious once you say it out loud:

You never meet the world “raw.”

You meet it through shortcuts. Through labels. Through categories. Through the mental equivalent of turning the resolution down so your brain doesn’t melt.

When I say “a bird,” I’m not tracking every feather, every micro-movement, every air current. I’m doing something simpler: I’m packaging a messy swarm of details into a single object that I can think with.

Six Birds theory is basically a clean mathematical way of talking about that move—and five other moves that inevitably come with it. It’s not a “the universe is made of vibes” manifesto. It’s more like a toolkit: if you want to talk about emergence (how higher-level “things” and “laws” appear), here are the minimal gears that keep showing up, no matter the domain.

The paper calls those gears P1–P6. It nicknames them “the six birds.” I’ll translate them into normal-person language.

The vibe: emergence is what happens when your description stabilizes

Imagine you’re organizing a chaotic camera roll.

At first you scroll and it’s just noise: 3,000 photos of food, friends, screenshots, accidental pocket pictures.

Then you start grouping:

“Trip to Lisbon”
“Family”
“Memes”
“Receipts”
“That one plant I keep forgetting to water”

Suddenly you’ve got objects: stable chunks you can refer to. And you have a kind of theory: a way of carving the world up that makes it manageable.

In Six Birds theory, a “theory” isn’t a big philosophical worldview. It’s just:

A lens (how you group things / what distinctions you allow yourself to see),
A completion rule (how you “finish” a description so it becomes stable),
An audit (some quantity you track to keep yourself honest—especially about time and irreversibility).

This is a really important vibe of the paper: it keeps stability, novelty, and directionality separate, because people mix them up constantly.

Stability: do you get solid “objects” that don’t fall apart when you look again?
Novelty: can you genuinely extend your theory, not just reword it?
Directionality: is there a real arrow of time, or are you just fooling yourself by hiding information?

Six Birds theory tries to give each of those its own “certificate,” so you don’t smuggle in big conclusions by accident.

One key idea: “doing it twice changes nothing” = stability

The technical word the paper leans on is idempotent, which sounds like a Harry Potter spell but just means:

If you apply an operation twice, it’s the same as applying it once.

Like running autocorrect on a paragraph: the first pass might fix a bunch of stuff. The second pass usually does nothing, because it’s already in the “finished” form.

That “finished” form is what the paper calls a fixed point. And fixed points are where objects live.

So, in human terms:

A completion rule is like a “make this description self-consistent” button.
A stable object is a description that stays the same after you press the button again.

That’s emergence, in this framework: not magic, not spooky new substances—just stability under a chosen way of describing things.

The other key idea: you don’t get endless novelty by pressing the same button forever

Here’s the blunt takeaway I wish more people would say out loud:

If you keep using the same lens and the same completion rule, you eventually saturate.

You reach a point where pressing the “complete” button doesn’t create anything new. It just reaffirms what you already had.

So if you want open-ended growth—new concepts, new objects, new levels—you can’t just iterate closure forever. You have to change the operator. Change the lens. Change what counts as describable.

That’s the paper’s version of “new paradigms don’t come from doing the old paradigm harder.”

Meet the Six Birds (P1–P6), in plain English

The paper argues that once you accept a few basic facts—like “you can compose processes” and “your access to the system is limited”—these six moves aren’t optional. They show up canonically, like the default moves any describer has to make.

Bird 1: Packaging (P5) — “Squash many details into one object”

This is the first and most relatable move.

You take lots of micro-states and treat them as “the same” for your purposes. That creates a quotient in math-speak, but the vibe is:

Many different realities → one label.

This is how “temperature” packages the motion of trillions of particles into one number. Or how “the economy” packages a million messy transactions into a few indicators.

Packaging is where objects come from.

Bird 2: Accounting (P6) — “Track what survives the squashing”

Once you package, you lose information. So you start caring about monotones: quantities that behave predictably as you zoom out.

Think:

“Number of categories”
“Amount of uncertainty left”
“What patterns are preserved under my lens”

The paper is really careful here: accounting is not automatically the arrow of time. It’s just the bookkeeping that comes with coarse descriptions.

Bird 3: Staging (P4) — “There are levels; you can refine your lens”

Staging is the idea that you don’t just have one level of description.

You can have:

a rough lens,
then a refined lens,
then a more refined one,
and so on.

In human terms: you can zoom in.

But the paper adds a constraint that feels very real: if your “interface” to the system is bounded (you only get so many measurements, so many knobs), then your refinement can’t explode arbitrarily fast. You don’t get infinite complexity for free.

Bird 4: Constraints (P2) — “Not every story is physically (or structurally) possible”

Once you have a lens, some macrostates are simply not representable.

This is huge and easy to miss.

Even if you can imagine a high-level description, it might not be compatible with the way the world actually updates. Your lens can carve out a “feasible” subset of macro-stories.

Like: not every “plot summary” matches any actual movie. The summary has to be something that could come from real scenes.

Bird 5: Operator Rewrite (P1) — “If the macro rules don’t close, you have to rewrite something”

This is the moment where theory hits friction.

You’d like to describe the world at the macro level and say, “Here are the macro rules.”

But sometimes the micro dynamics don’t respect your packaging. Two different micro-states you’ve lumped together can evolve in ways that don’t stay lumped together.

When that happens, you don’t get a clean macro law “for free.”

So P1 is the forced move:

refine your lens,
or modify your macro rule,
or admit there is no closed macro dynamics at that level.

Philosophically, this is the part that feels like scientific revolutions in miniature. Sometimes the world refuses your categories, and you have to update the categories.

Bird 6: Holonomy / Route Mismatch (P3) — “Doing things in a different order gives a different result”

This one is subtle but kind of beautiful.

Sometimes there are two reasonable “routes” from input to output:

Route A: update the system, then package it
Route B: package it, then apply the macro update

If those routes disagree, you get route mismatch. That mismatch is a diagnostic: it tells you there’s tension between your description and the underlying process.

But (and this is important) the paper emphasizes:

Route mismatch is not automatically a certificate of “time’s arrow” or irreversibility.

It’s a sign of non-commuting operations. It’s a symptom, not a moral judgment.

Time, honesty, and why “forgetting stuff” shouldn’t fake an arrow of time

A big part of the paper is basically a rant against sloppy irreversibility claims.

It gives a clean definition of arrow-of-time as asymmetry between forward and reversed histories (technically: path-space KL divergence). The reason this matters is:

Coarse-graining can’t increase real irreversibility.

If you blur your view, you can hide evidence of irreversibility. But you shouldn’t be able to create it out of thin air. The paper proves a data-processing style result: pushing dynamics through a lens can only contract that asymmetry.

Then it tackles a classic trap:

The protocol trap

If you apply different update rules in a hidden schedule (like alternating between two rules), and you don’t include the schedule as part of the state, it can look like the system is generating irreversibility.

But if you include the “clock” or “phase” inside the system’s state—if the model is truly autonomous—then that fake arrow disappears unless there’s a real drive in the protocol itself.

In the paper’s slogan form: “P3 needs P6_drive.”
Translation: route mismatch alone doesn’t create sustained directionality unless there’s an actual non-equilibrium push somewhere in the full dynamics.

This is the paper being careful in a way I genuinely appreciate: it refuses to let “we hid variables” masquerade as “nature is irreversible.”

Where novelty comes from: most new predicates aren’t expressible in your old language

The paper also has this clean little argument about definability.

If your “theory” is basically a partition—your lens divides the world into a limited number of buckets—then only certain properties are expressible inside that theory. Properties have to be constant on buckets.

Here’s the punchline:

Out of all the possible “yes/no” properties you could define on a big system, only a tiny fraction are definable from a coarse partition.

So if you add a new predicate at random (a new distinction, a new concept), it’s overwhelmingly likely to be non-definable in the old theory—meaning it forces a strict extension.

In plain terms:

Your current language can only say so much.
Most possible new ideas don’t fit in it.
So “strict theory extension” is cheap once there’s hidden complexity.

That’s the paper’s clean mathematical version of: the world has more structure than your current vocabulary can capture, and expanding vocabulary is how you climb.

So what is Six Birds theory, really?

I’d summarize it like this:

Six Birds theory is a minimal grammar for emergence.

It tells you:

how stable objects can appear inside a given way of describing,
why closure saturates (so endless novelty requires changing the description, not repeating it),
how to separate “new objects” from “new concepts” from “real time asymmetry,”
and how to audit your claims so you don’t get fooled by your own blind spots.

It’s deliberately not trying to be a total theory of life, mind, or society. It’s more like a set of scaffolding pieces you could reuse in those domains if you’re willing to specify your lens, your completion rule, and your audits carefully.

A closing image (because philosophy majors can’t resist)

Next time you point at something and say “that thing,” imagine six birds landing on a wire above your head:

Packaging: you lump details together.
Accounting: you track what survives that lumping.
Staging: you allow multiple zoom levels.
Constraints: you admit not every macro-story is possible.
Rewrite: you update your theory when the world won’t fit your categories.
Route mismatch: you check whether your shortcuts commute—or whether something deeper is going on.

That’s the whole vibe.

Not a grand myth. Just a sharp reminder that “what exists,” for us, is always braided with “how we describe.”

And if you want new worlds, you don’t just press the same button harder. You learn new buttons.