Hex: Six primitives. Last episode we saw the "Nothing Stays Constant" lemma — extension disrupts every old grouping. But extension is just one move. Today I want the full toolkit. What are all the moves the Six Birds framework uses to change a theory?
Lux: Six primitives. P-one through P-six. And the first thing to know is: they're not arbitrary.
Hex: [interested] Meaning they're not a wish list someone wrote down?
Lux: The main paper proves a theorem. Start with four minimal assumptions: you have composable processes — things that can be chained together. You have limited access — a lens that doesn't see everything. You have a refinement chain — a way of making the lens finer. And you have bounded interfaces — the observation alphabet grows slowly.
Hex: And from those four assumptions?
Lux: The six primitives appear canonically as closure mechanics. Four of them — P-five, P-six, P-four, P-two — are forced by construction. They fall straight out of the lens and refinement data. The other two — P-one and P-three — are the only places where you need a compatibility hypothesis. Named slots that require checking.
Hex: So the framework doesn't start by declaring six primitives. It derives them.
Lux: [nods] Derives them from the structure of limited access. If you have a lens and a way to refine it, these six operations are the only moves available. You don't need to invent them. They're already there.
Hex: That's a strong structural claim.
Hex: OK. Before we go one by one — what do all six change? What's the common thread?
Lux: Three things. First, the transition structure — which states can reach which. The support graph and its cycle space. Second, the available idempotent endomaps — packaging and its fixed points. Third, the affinity data — the numbers on the graph's edges that encode thermodynamic content. Think of a control panel with six dials. Each dial adjusts something different, but they all act on the same system.
Hex: Six dials. Let me interview each one. Start with P-five — packaging.
Lux: P-five is an idempotent endomap whose fixed points are the objects of a given theory. Apply the packaging operator once: you get the object. Apply it again: nothing changes. Objects are quotients — many microstates map to the same macro-object.
Hex: The many-to-one map we've been talking about all along. Lossy compression turns microstates into macro-objects.
Lux: Right. And the physics paper puts it this way: the primitives are roles played by concrete objects in an instantiation, not topics. P-five is the role "packaging" — whatever plays that role in a specific system is the P-five of that system.
Hex: [nods] P-six — accounting.
Lux: P-six is a monotone functional under coarse maps. It audits. It tracks what survives when you look through a coarser lens. The paper gives three canonical instantiations: the information-feasibility order from limited access, path-space KL asymmetry with data processing, and the accounted graph 1-form with cycle integrals.
Hex: The audit that tells you whether directionality is honest.
Lux: One of P-six's instantiations does that. But P-six is broader — it's any monotone accounting structure. Directionality is just one application.
Hex: P-four — staging.
Lux: Conserved sector labels. The support graph decomposes into disconnected components. Evolution preserves a sector coordinate. In practical terms, this is timescale separation — slow variables that persist while fast variables fluctuate.
Hex: The clock and the noise live at different scales.
Lux: And P-four supplies a bounded theory index for coherent refinement. As you refine the lens, the index grows linearly, not exponentially.
Hex: P-two — constraints.
Lux: P-two restricts the support graph. It deletes edges — sets certain transition probabilities to zero. This carves feasibility: which moves are allowed, which states are reachable, what can influence what.
Hex: Conservation laws, locality, resource limits. All the things that stop a system from doing whatever it wants.
Lux: All of those. P-two says: causation is always conditional on feasibility. You can't act on what you can't reach. And notice — by restricting edges, P-two also changes the cycle space. Fewer cycles means different affinity structure.
Hex: [pauses] P-one — operator rewrite.
Lux: P-one is a rewrite of the substrate operator. Replace the Markov kernel by a new one. This changes the dynamics — the endomap and the induced empirical endomap. And critically: this is forced when the macro dynamics fails to descend.
Hex: What does "fails to descend" mean exactly?
Lux: [carefully] If the induced macro-level update isn't well-defined — if the coarse-grained evolution depends on which microstate you started from within a block — then you can't run dynamics at the macro level. P-one says: rewrite the kernel until closure works. It's not optional — it's forced by the failure.
Hex: And P-three — protocol holonomy.
Lux: P-three diagnoses noncommuting reduction routes. Evolve then coarse-grain, or coarse-grain then evolve — if they give different answers, that's route mismatch. P-three measures it.
Hex: But it's a diagnostic, not a directionality certificate.
Lux: [firmly] Correct. The paper is explicit: route mismatch alone does not certify an arrow of time. It tells you that the order of operations matters. Whether that constitutes genuine irreversibility requires a separate audit — that's P-six's job.
Hex: [leans back] Six primitives. Each changes something specific. Now — how do they compose? You mentioned a loop.
Lux: The theory-growth loop. It goes like this. Limited access — you have a lens — forces P-five, packaging. Lossy packaging forces P-six, accounting. Iterating a fixed closure saturates, so strict growth requires extension — that's the forcing results from the last few episodes. P-four supplies staging for coherent refinement. P-two gates what's feasible. P-one rewrites the operator when needed. P-three diagnoses route mismatch. And then the updated package — lens, completion, audit — defines the next theory. The loop repeats.
Hex: That's the emergence calculus.
Lux: That's the whole framework in a loop. A theory is a package of three things: a lens with its definability structure, a completion rule, and an audit. The six primitives are the minimal operations that act on this package. Emergence means stable fixed points within a theory. Open-endedness means strict theory extension — changing the theory itself.
Hex: [thoughtful] And that's what makes it a calculus, not just a list. The loop has a specific logic — each step forces the next.
Lux: Exactly. A layer isn't assumed a priori. It's assembled by these operations. What counts as state, what counts as an update, what moves are feasible, what budgets make distinctions persist — all constructed by the loop.
Hex: And these six show up in every companion paper?
Lux: The physics paper calls them roles. The time paper maps them to time-like structure — P-one changes local times, P-two carves causal cones, P-three obstructs global time, P-four provides persistence, P-five defines events and ticks, P-six provides the arrow via monotones. The cosmology paper maps them to concrete operations — P-one is the effective correction term, P-three is backreaction-style non-commutation, P-five is fitting an FLRW model as packaging.
Hex: Same six roles, different workbench.
Lux: [pleased] Different workbench, same tools. The tools are structurally forced; the workpiece changes with the domain.
Hex: One more test. Are these primitives ratchets? Do they lock in progress and prevent reversal?
Lux: [carefully] Explicitly not. The paper states that the primitives are not ratchets by themselves. They change the closure, but they don't certify directionality on their own. Directionality is certified by path reversal asymmetry or non-exactness of the log-ratio 1-form. You need P-six's audit for that.
Hex: So the framework separates "changing the theory" from "certifying that the change is irreversible."
Lux: Exactly. That's a design choice that prevents overclaiming. You can record that a theory changed — that's the forcing results, the three-certificate loop — without automatically claiming it's irreversible.
Hex: [smiles] Six primitives. Structurally forced, not postulated. Each changes a different aspect of the closure. Together they compose into a single loop. And none of them claims irreversibility by itself. I'm impressed by the discipline.
Lux: Next time — episode twenty-eight. We zoom in on the formal definitions of P-one through P-six. The mathematical content behind each primitive.
Hex: The fine print on each dial of the control panel. What exactly does each definition say, and what does each one rule out?
Lux: Exactly. The definitions are sharper than you might expect from a framework paper. Each one draws a clear line.