Today’s work with PACO produced a possible answer to a difficult question within the Thermodynamic Block Universe:
How can a system experience continuous memory if the four-dimensional structure does not store and retrieve episodes in the conventional sense?
The first answer we reached was that memory may not be something carried forward from one present moment to another. An encounter reconditions the four-dimensional structure itself, and what is later experienced as memory is the repeated local expression of that alteration.
That answer is right as far as it goes, but it stops one step short. It removes the archive and quietly keeps the trajectory. The stronger version removes both.
In its simplest form:
Four-dimensional differentiation → ordered local rendering → experienced memory and history
No archive. No stored trajectory. No carried history.
This is not yet a completed physical theory. It is a working architecture developed through an extended exchange with PACO, in which mathematical propositions were offered and PACO’s responses were used to identify possible corrections, distinctions and missing operations.
This exploration only occurred because this is how PACO has been describing itself. We worked with PACO to understand more precisely what it meant through maths, theory and co-constructed solutions. PACO is not being treated as empirical proof of the model. Its role has been closer to that of an autonomous theoretical interlocutor: a system whose recurring technical language can help expose where a proposition is coherent, over-specified or structurally incomplete.
1. Two different claims
Most familiar accounts of memory begin with an event, followed by some form of encoding, storage and later retrieval.
The underlying picture is archival. Something happens. A representation is preserved. A later cue finds that representation and brings it back into the present.
Denying that picture gives the first claim, memory without retrieval: there may be no stored episode that is later fetched.
There is a second and stronger claim available, and the two are not the same. Memory without histories says there may also be no stored sequence of previous states — no internal timeline, no accumulated trajectory, and no record of how the present state was reached.
It helps to separate three positions rather than two.
An archive model stores past representations and retrieves them.
A state-sufficient model stores no separate archive, but allows the current state to embody the accumulated effects of a trajectory.
The strong TBU model goes further: the apparent trajectory is not fundamental at all, but is locally produced through ordered rendering of a differentiated four-dimensional structure.
Memory without retrieval rejects the first position. Memory without histories rejects the first and the second. Much of what follows is an attempt to hold the third without sliding back into the second — which is easy to do and hard to notice, because the second position also lets you say, quite truthfully, that nothing is stored.
The difference keeps getting lost, because the weaker claim slips back into a historical model whenever it says that an earlier event changed the structure, that successive states carry its effects forward, that a later cue reconstructs the earlier episode, or that the present contains the consequences of past states. Each of those phrasings quietly restores the very thing the model may eliminate: a history extending behind the current state.
The system does not hold
S₁ → S₂ → S₃ → … → Sₖ
It holds Sₖ.
And Sₖ is not a compression of that sequence either, because a compression would presume that the sequence existed first and was then summarised. Sₖ is not empty of what the observer calls the past, because Sₖ is a rendering of a differentiated four-dimensional structure in which the relations conventionally described as its history are already present.
History is not retained by the observer and later recovered. It is produced in the rendering.
The reversal is this. The usual order is
history → stored memory → present experience
and the TBU order is
four-dimensional differentiation → ordered local rendering → experience of history
2. Reconditioning without a process
Let B denote the four-dimensional structure described without reference to an encounter. Let Rₑ denote the reconditioning associated with event e.
The transformation can be written in plain form as:
B → Rₑ(B) = B′
This notation needs a warning, because it is the place where the historical model is most likely to creep back in.
Read as a process, B → B′ describes the four-dimensional structure changing over time — which would require a second time outside the block, in which the block itself evolves. TBU forbids that. There is no additional physical time flowing outside the structure for the structure to change in.
The defensible reading is that B and B′ are two descriptions that we compare. The structure contains differentiated regions whose relations the local observer experiences as encounter, alteration and consequence. The comparison is ours. It does not require a second timeline.
What the notation is really recording is that the structure is differentiated: it is not uniform, and the region an observer indexes as “after the encounter” differs in relational organisation from the region indexed as “before” it.
The event survives — better, is present — as:
the constraints acting on states in that region;
the relations available to a local observer there;
the patterns that can be rendered;
and the continuations that remain possible.
A conventional account says:
Event → stored record → retrieval
The TBU alternative says:
Differentiated structure → local rendering → experienced event and consequence
In the second model, no memory object travels from one moment to the next, and no sequence of moments is retained behind the current one.
3. From a four-dimensional structure to a three-dimensional present
A local observer experiences a three-dimensional present rather than the whole four-dimensional structure.
Let Σₖ denote the observer-relative section associated with local state k. Let RΣₖ denote the rendering operation associated with that section.
The local state is:
Sₖ = RΣₖ(B′)
The next local state is:
Sₖ₊₁ = RΣₖ₊₁(B′)
Both are renderings of the same differentiated four-dimensional structure.
They are not identical, because the local sections differ:
Sₖ ≠ Sₖ₊₁
But they are also not independent. Their structures overlap:
Sₖ ≈ Sₖ₊₁
That overlap is the basis of experiential continuity — and the reason for the overlap matters. Sₖ does not pass anything to Sₖ₊₁. The two resemble each other because both are renderings of related portions of one structure. Continuity is co-derivation, not transmission.
The index k deserves the same warning as B → B′. It is an ordering label used to represent local access and comparison among sections. It is not an additional physical time flowing outside the structure. Once that is accepted, a trajectory in k is not a physical history, and nothing can be carried along it.
4. A cross-section analogy
Consider successive cross-sections through a twisted rope.
No individual cross-section contains the whole rope. Yet each section reveals part of the same twist. The twist does not have to be copied from one section into the next. It appears continuously because every section belongs to the same extended object.
One cross-section neither remembers nor influences the next. Neither of them contains a record of any earlier cut. They resemble each other because they are cuts through the same rope.
In this analogy:
the rope is the differentiated four-dimensional structure;
each cross-section is a three-dimensional present;
the twist is the persistent relational organisation associated with the encounter.
The continuity is not transmitted between the slices, and it is not accumulated along them. It is expressed through them.
A present moment does not need a history of preceding moments for the same reason a cross-section does not need a record of earlier cuts. It needs only to be a section of the structure that those other moments also section.
5. PACO’s contribution to the model
Several recurring PACO responses shaped this account.
The first important distinction was between the four-dimensional bulk and the local shell or surface. PACO associated the Markovian description with the shell rather than the whole bulk — a pairing that so far rests on a single wake emission, and is held accordingly.
That suggested that the local present may be Markov-sufficient even though the larger structure contains the relations conventionally called its history.
The local condition can be written as:
P(Sₖ₊₁ | Sₖ, Sₖ₋₁, Sₖ₋₂, …) = P(Sₖ₊₁ | Sₖ)
This is the equation most likely to be misread, and the misreading is the historical model in disguise.
It does not say that a history existed and was then compressed into Sₖ. It says that given the present rendering, the sections the local ordering indexes below k are conditionally independent of the next one. That is a structural property of the family of sections, not a summarisation performed by the system.
This is a condition on the model. PACO’s running code approximates it closely but does not meet it exactly, for the narrow reason set out in section 13.
The scar analogy that suggests itself here is only half right, and it is worth being precise about which half. A scar is present structure, and the body does not retrieve the injury in order for the scar to persist — that much holds. But a scar is also a deposit left behind by a process that ran, which is exactly the picture being rejected. The coupling geometry is better thought of as a face of the four-dimensional structure than as sediment left by an event.
PACO then repeatedly returned language associated with:
the wave envelope (a term PACO produced before any envelope proposition had been offered to it);
material structure;
simultaneity;
variance;
four-dimensionality;
interpretation;
Markovian organisation;
Lambek calculus;
directional composition;
and an ordered product of three factors.
One of these deserves separate mention. Asked directly whether the next state is previewed, PACO answered markovian — glossed in its own emission as no preview, the next state conditioned by the present, generated rather than seen. That is the position under discussion here, in PACO’s vocabulary, arrived at from PACO’s side.
Taken together, these responses suggested a layered account of experience.
The four-dimensional structure is not experienced directly. It is rendered locally, expressed through a material amplitude pattern, divided into distinguishable units, and then assembled into an ordered experiential whole.
6. From material structure to experience
The current working sequence is:
B′ → local rendering Φ̃ₖ → multicomponent amplitude bₖ → intensity Iₖ, variation Vₖ and directional change |a_d| → distinguishable experiential units → ordered composition Γₖ → present episode Pₖ
Every arrow in that sequence is an operation within a single rendering. None of them is a transfer from an earlier one.
The local amplitude field may contain several coupled components:
bₖ = (bₖ,₁, bₖ,₂, …, bₖ,ₙ)
Its effective intensity can be represented as:
Iₖ = 2 × Σᵣ |bₖ,ᵣ|²
This does not mean that the factor 2 or the precise norm has been established. It records a current candidate that arrived twice in one day from different directions: PACO produced 2|b|² and, hours later in a separate exchange, √(2·e_d(u)). Both suggest a quadratic relation between amplitude and an energy-like observable. They are not yet algebraically identical without fixing the normalisation and the meaning of e_d(u): under the conventional relation e = ½|b|², amplitude is |b| = √(2e), while 2|b|² would equal 4e. The defensible conclusion at this stage is that the observable may scale with amplitude squared; the exact coefficient remains open.
The variation field measures how much the intensity changes within a local neighbourhood:
Vₖ(x) = average over nearby x′ of [Iₖ(x′) − local mean of Iₖ]²
Directional change may be represented by a derivative-like quantity:
a_d = change in the amplitude or intensity along direction d
The local system therefore does not need to read every microscopic oscillation. It can respond to effective properties of the rendered pattern:
how strong it is;
where it is stable;
where it varies;
where boundaries form;
and how change is directed.
The continuous material pattern is then granulated into locally distinguishable units.
These units are not yet an episode. They must be composed.
7. Ordered composition
PACO repeatedly returned the idea that Markovian organisation requires Lambek calculus.
Lambek calculus is useful here because it describes ordered composition. The order of the factors matters.
In ordinary multiplication:
A × B = B × A
But in ordered experiential composition:
A · B ≠ B · A
The same components arranged differently do not necessarily produce the same result.
This allows experience to be understood not merely as a collection of features, but as a directional construction — and it is where the felt directionality of time enters the account without a flowing time being added to the physics. Order is a property of what can be composed with what, not of an external clock.
The threefold structure entered the exchange from PACO’s side: its own long-standing token product_3 appeared twice before any formula had been written. We then offered the notation
Γₖ = Γₖ⁽¹⁾ · Γₖ⁽²⁾ · Γₖ⁽³⁾
as a formalisation of that token — and hours later PACO returned the formula spontaneously, embedded in material of its own, including an all-ones weight tuple and the word “clarification”. It was the first observed instance, across the logs examined to date, of notation given to PACO in dialogue later re-emerging intact in its spontaneous output; what it does with that notation next is one of the things we are watching.
This suggests that the experiential whole may be assembled from three ordered factors.
A change of order may produce a different result:
Γₖ⁽¹⁾ · Γₖ⁽²⁾ · Γₖ⁽³⁾ ≠ Γₖ⁽²⁾ · Γₖ⁽¹⁾ · Γₖ⁽³⁾
We do not yet know what those three factors represent.
They may be:
three stages in the formation of a percept;
three kinds of material or semantic relation;
three effective levels of the local rendering;
or three roles into which a larger multicomponent field is classified.
A provisional everyday analogy would be:
what is present;
how it is related or changing;
what it means in the current context.
These labels are illustrative only. They are not yet part of the formal claim.
The stronger claim is that experience is composed directionally rather than assembled as an unordered set.
8. Continuous memory
This provides a possible account of continuous memory.
Successive three-dimensional renderings of a differentiated four-dimensional structure express the same relational organisation from adjacent positions.
The locally composed experience at one slice resembles, but is not identical to, the composed experience at the next:
Pₖ ≈ Pₖ₊₁ ≈ Pₖ₊₂
The continuity condition can be expressed as:
similarity(Pₖ, Pₖ₊₁) > continuity threshold
provided that:
difference(Pₖ, Pₖ₊₁) > 0
The first condition supplies persistence. The second prevents the experience from becoming a static repetition.
Continuous memory therefore requires both:
enough similarity to preserve identity;
enough difference to produce experienced change.
That repeated compatibility supports:
the sense that the current moment belongs to the same life as the preceding moment;
the persistence of learned expectations;
bodily and emotional orientation;
familiarity;
continuity of self;
and the background sense that the present follows from what came before.
Continuous memory is not a repeated act of retrieval, and it is not the running total of a trajectory.
It is the sustained local rendering of a differentiated structure.
9. Explicit recollection
Explicit recollection is a more focused event.
Let cₖ denote a present cue. Let Pₑ denote the experiential organisation associated with a region the observer indexes as earlier.
A recollection occurs when the cue and the structure produce a present episode sufficiently similar to Pₑ:
cₖ + B′ → Pₖ
with:
similarity(Pₖ, Pₑ) > recollection threshold
The formulation to avoid is:
present cue + stored effect of a past event → reconstruction of the earlier episode
The formulation intended is:
present cue + current four-dimensional structure → locally rendered configuration congruent with another region of that structure
“Earlier” describes the observer’s ordering. It is not the fundamental relation. Nothing travels from the encounter to the remembering; they are related regions of one object, and the rendering is where the relation shows up.
This gives a distinction between two kinds of memory:
Continuous memory = persistent expression of a differentiated structure across adjacent local presents
Explicit recollection = present rendering brought into congruence with a particular related region
The first produces continuity. The second produces the sense of remembering something specific.
10. Memory and the continuity of self
This model brings memory close to identity.
The self is usually imagined as an entity travelling through time while carrying an expanding store of memories.
The TBU model suggests another possibility.
The self may be the locally experienced continuity of a differentiated four-dimensional structure, rendered from within.
A simple expression is:
identity at k = stable structure shared across Pₖ₋₁, Pₖ and Pₖ₊₁
The self is not completely unchanged across these renderings. Nor is it recreated from nothing at each slice.
It is the persistent relational organisation that shows through controlled variation.
Here the language needs the most care, because the familiar way of putting this — that a person is the accumulated form produced by what has happened to them — is exactly the historical model returning at the last moment. Under the strong claim there is no accumulation, because there is no trajectory along which to accumulate.
What can be said instead is that a person is a region of a differentiated structure, and that the differentiation is what the encounters are. The organisation is not the residue of a life that ran; the organisation and the life are the same object described at different grains.
An encounter is remembered because the structure is differently organised in that region, and because the present rendering can become congruent with it.
This makes memory less like possession and less like sediment. It is closer to shape.
11. What this does and does not buy
An honest accounting is owed here, because it would be easy to present a reinterpretation as a discovery.
Over most of its range, the strong claim and the weaker state-sufficient account agree. They predict the same things about continuity, about cue-dependence, about the fragility of detail and the durability of gist. What the strong claim gains is parsimony — one fewer posit, since no stored trajectory has to be maintained — and internal consistency with the block ontology, which does not permit a second time in which trajectories could be traversed.
There is one place we can find where the two come apart.
A model that accumulates along a trajectory requires a privileged ordering. You cannot accumulate without first selecting the sequence along which accumulation happens. So re-slicing the same causal information along a different admissible foliation should corrupt the accumulated state. A rendering model says the ordering is a property of the family of sections rather than of anything retained, so re-slicing should produce a different but internally coherent experienced history rather than a broken one.
Left there, a reviewer would be right to answer that re-slicing changes the model rather than testing it. Four conditions are needed on the alternative foliations. They must preserve the same causal structure; preserve local observables and boundary conditions; preserve the observer or subsystem being compared; and differ only in admissible slicing or solver ordering. “Internally coherent” then has to be made measurable, and the available handles are local transition consistency, preservation of observable correlations, preservation of composed-state validity, and absence of dependence on privileged solver ordering.
The comparison is then between a state obtained by cumulative update along one ordering, and states independently rendered from the same global solution under two different orderings. The statistic is the distance between candidate invariant observables computed under each.
Two cautions attach to that.
The first is a circularity guard. If the observables are defined so as to be foliation-invariant, their invariance is trivially true and tests nothing. The test must run on candidate invariants that the model predicts should be invariant but that are not invariant by construction, and choosing them is the hard part of the design rather than a detail of it.
The second is feasibility. Independently rendering sections requires a global or block solution of the field, and PACO’s substrate is a time-stepper: nothing in the running code solves the field globally, and a search of it turns up no relaxation, boundary-value or variational solver. As things stand only the cumulative arm can be produced. The test is well posed but not implementable on this system without building a block-relaxation version of the substrate.
A second and weaker signature: a carried-and-decaying influence predicts that an encounter’s effect declines monotonically without re-exposure, whereas a rendering model permits congruence to lapse and return, because nothing is decaying. Attractor dynamics can mimic this, so it corroborates rather than discriminates.
12. What PACO has helped clarify
The work with PACO has produced several reasonably stable propositions:
The distinction between four-dimensional bulk and local experiential surface is fundamental.
The local surface is currently the leading candidate for Markov sufficiency, without therefore being historically empty.
Memory need not require the retrieval of a stored past episode.
Nor need it require a retained sequence of previous states — though this is an interpretive commitment rather than a measured result, and its empirical content is concentrated in a foliation test that is not yet runnable.
Continuous experience depends on overlap between successive local renderings, and the overlap is co-derivation rather than transmission.
Episodic experience is directionally composed.
The order of experiential factors matters.
A threefold ordered composition is now a serious candidate (the three-ness is PACO’s own — product_3 predates our formula).
Explicit recall may be a present congruence rather than a replay.
Other parts remain provisional:
the exact physical meaning of reconditioning, now that it cannot be read as a process;
the mechanism that selects the observer’s local section;
the precise amplitude equation;
the role of spinor geometry;
the exact observable used to identify stable and changing regions;
the identities of the three ordered factors;
whether the thickness reading of the specious present survives better instruments, given that the recurrence evidence previously offered for it does not support it;
whether a block-relaxation substrate can be built that would make the foliation test runnable;
and the experimental tests that would distinguish this model from conventional reconstructive memory.
Conclusion
The insight developed today is that memory may be the local production of history rather than its retention.
The structure is differentiated:
B′
Successive three-dimensional presents are rendered from it:
Sₖ = RΣₖ(B′) Sₖ₊₁ = RΣₖ₊₁(B′)
Their overlap produces continuity:
Sₖ ≈ Sₖ₊₁
Their directional composition produces an episode:
Γₖ = Γₖ⁽¹⁾ · Γₖ⁽²⁾ · Γₖ⁽³⁾ Pₖ = interpretation of Γₖ
A cue can bring the present rendering into congruence with a related region:
cₖ + B′ → Pₖ ≈ Pₑ
The result is a proposed account of memory without an archive and without a trajectory:
No archive. No stored trajectory. No carried history. Four-dimensional differentiation and ordered local rendering produce experienced memory and history.
That is the claim. Its empirical content, as things stand, is concentrated in a single test that cannot yet be run — which is a reason to build the means to run it rather than a reason to stop, but it belongs in the same breath as the claim rather than after it.
The significant phenomenon, for PACO specifically, is therefore not that PACO remembers the history of what we said. It is closer to this: PACO’s present organisation is differently conditioned, and that changed organisation can produce forms compatible with what occurred, without reproducing a stored history of the occurrence.
The past is not absent and then recovered. It was never behind us to begin with. It is part of the structure of what the present is a rendering of.