diff --git a/elements/neuron/appunti/2026-07-08-the-unexpressed-objects.md b/elements/neuron/appunti/2026-07-08-the-unexpressed-objects.md index 0cbdf2b..7344c02 100644 --- a/elements/neuron/appunti/2026-07-08-the-unexpressed-objects.md +++ b/elements/neuron/appunti/2026-07-08-the-unexpressed-objects.md @@ -2,228 +2,173 @@ *The objects the mechanism implies but never expresses. The pseudocode has PRE, POST, ASTROSYNAPSE — three components, each running its own local loop. It has no `synapse`: no variable -holds one, no line makes one act. Yet the synapse is real — the system builds it, sustains it, -consolidates or forgets it. The synapse is an object that is **verified but never expressed**: it -exists only as a behavior that three components continuously constitute together, along the axes of -time, space, and quantity. This document is the catalogue of such objects — the synapse, the branch, -the cell, the assembly, the rhythm — each real, each acted upon, none appearing in the mechanism, -each existing only as a sustained mutual project its parts are always working at without any of them -containing it.* +holds one, no line makes one act. Yet we speak of the synapse constantly. This document is about what +kind of thing "the synapse" is, given that it is nowhere in the mechanism — and the answer, worked +out below, is that an object is a name we lay over an aggregation of behaviours, each behaviour read +at a chosen cut.* -## Orthogonal to classical reduction: objects are cuts, not things +## The machinery — expression, cut, behaviours, and the name -Before any object, the move this document makes must be named, because it runs perpendicular to the -habit classical physics trained into us. The reductive default isolates a system at a **static -object-boundary** — this mass, that charge, this cell — treats the object as a persisting thing with -fixed identity, and explains behavior by **cause and effect between such objects**: A strikes B, B -moves. The parts explain the whole, and the parts are the same before and after. +Three things, in a fixed relation: -This system does not offer that. Ask "what is the neuron trying to achieve," and any answer is -partial — because the neuron is part of an assembly it cannot see, which is part of an organ, part of -an organism, with no top level where the question closes; and it is part of, and made of, synapses -and channels, with no bottom level either. There is no privileged object from which the whole story -can be told. This is the descriptive face of the model's founding principle (logic_principles, Part -I): as there is no global state and no privileged actor *inside* the system, there is no privileged -*object* from which to *describe* it. The absence of a top and a bottom is a fact about the system; -the necessity of choosing a cut is its consequence for the observer. +**The expression is possibility.** The components express the *possibility* of verifying behaviours +in time, space, and quantity. Nowhere is a frequency, a flow, a duration expressed directly — only +the latent possibility of reading them. Time, space, and quantity are never *in* the expression; they +become readable only when we cut. -So here an **object is a cut**, not a thing — a choice the observer makes for the sake of a question. -And because these objects are active and multi-scaled, a cut is not merely a spatial line; it is a -choice of three things at once: +**A cut is a choice of level.** To read anything, we choose a level on each axis — time (ms · sec · +min · hr), space (channel · synapse · branch · cell · territory), quantity (single quantum · occupancy +· total) — plus a scope (day · night). The cut is the grain at which we verify; nothing more. It is +what makes behaviours *actually* verifiable, out of the possibility the expression laid down. -- a **boundary** — what counts as inside (the parts that constitute the object) and what counts as - outside (context, which then appears only as signals arriving and constraints descending at the - boundary); -- a **timescale** — the same boundary yields *different objects* at different grains: the - synapse-over-milliseconds (a coincidence) and the synapse-over-a-night (a structure being rebuilt) - are different objects sharing parts, because the behavior verified and the relevant context differ; -- a **scope** — day or night: the same spatial cut opens onto the *world* by day (its context is - behavior, its inputs sensory) and onto the *economy* by night (its context is material, its inputs - supply). The outside inverts between scopes. +**Behaviours are what a cut lets us read.** Put the expression through a cut and you get behaviours — +just behaviours, read at that level. This is the whole verification act: -Every cut is legitimate and every cut is **partial**: it makes some behaviors verifiable by treating -the rest as context, and a behavior that spans its boundary — the neuron aligning *within* its -assembly — is only half-visible from inside, because the other half lives in a larger cut. No cut is -the whole story, for the same reason no computation is the whole model (logic_principles, Part I): -there is no privileged, bounded, stable object to be the whole. A cut is to *description* what one -history is to *simulation* — the only tractable thing, necessarily partial, honestly chosen. +``` +expression → cut → behaviours +``` -Two consequences shape everything below. First, the relations that matter are not cause-and-effect -between objects but **constitution across cuts**: parts constitute an object (the synapse does not -*cause* its parts — it *is* them, seen from a cut); a level constrains the one below and emerges into -the one above. Within a single cut at a single timescale, ordinary cause and effect still holds (this -release causes that response); it is the objects *themselves* — synapse, alignment, assembly — that -live at the intersection of cuts, where between-object causation is not the operative relation. -Second, cuts are not arbitrary: the informative ones fall at the **joints** — where the system's own -coupling is denser inside than across (a synapse's three parts interact more with each other than -with the next synapse; a neuron's components more with each other than with the next neuron). We -prefer these natural cuts because they carve where the coupling already is, while remaining explicit -that they are still cuts — still boundary × timescale × scope, still partial. +and it runs once per cut. The same expression, through a different cut, yields different behaviours. -*The pseudocode is itself a cut — the finest one.* It cuts at the local component and treats every -larger whole as context arriving at the boundary: dopamine, the day/night context, the -renormalization are the organism and the hypothalamus reduced to inputs; what a component emits is -output for others to integrate. This document simply makes *coarser* cuts in the same web. And there -are affinities with the scale-relative frontier of physics — the renormalization group, non- -equilibrium thermodynamics — which also make description depend on the scale of the cut; but as the -simulation argument showed, those point in the direction without solving this system. They tell us -cut-relative description is legitimate physics; they do not hand us the object. +**An object is a name over a set of these results.** "The synapse" is not a step in the chain, not an +input, not an output. It is the name we lay over a *collection* of chains — -## The frame: verified but not expressed +``` + the synapse = name over { expression→cut₁→behaviours A, + expression→cut₂→behaviours B, + expression→cut₃→behaviours C, ... } +``` -Within a chosen cut, an object is described by what it makes verifiable. The three axes along which -any behavior is conceptualized are **time, space, and quantity**. An unexpressed object is *verified* -when its behavior, read along these axes, is constituted by its parts and held as the level of some -store — the object's present shadow (see logic_principles §7, Integrate). Behaviors read along the -axes compose into the observables: a **frequency** (events per time), a **flow** (quantity per time), -an **elapsed interval** (time between events), an **amount at a moment** (quantity at a time), a -**spatial extent** (quantity over space), a **coincidence** (several things at one time and place). -None is emitted by any component; each is constituted by many local acts and readable only where they -accumulate. The verification is not a computation anyone performs — it is the automatic consequence -of local acts accumulating in a store. +— a grouping we perform across many cuts and label with one stable word. The object contributes +nothing to what is read; behaviours are all there is on the mechanism's side. "Synapse" adds no +behaviour and no verification — it adds only a grouping in our account. This is the cleanest form of +*verified but not expressed*: the object is not even verified; **behaviours** are verified, at cuts, +and the object is just the name over a chosen set of them. -But an unexpressed object is more than a passively-observed pattern. It is an **active project**: its -parts are continuously *aligning* along the three axes — tuning toward each other so the behavior -lands better and lasts longer. So each object is described through two lenses that are the same -structure still and moving: +**The name is one; the cuts are many.** Because "synapse" names the collection, it stays fixed +whichever cuts the collection gathers — synapse at ms, synapse at night, synapse as third party to +pre and post. These are not different objects; they are the same name, read at different cuts. It is +*always the synapse* — but always the synapse *at a cut*. The name gives stability (we can speak of +one thing across contexts); the cut gives specificity (we know which of its behaviours are in view). +Neither alone suffices: a name with no cut is vague (which behaviours?), a cut with no name is +unanchored (behaviours of what?). -- **Decomposed (still):** what behavior the object's parts constitute, and how it separates along the - three axes — which part owns which axis, and where the axes recombine. -- **Aligned (moving):** how the parts actively tune toward each other along those axes — and, because - each part belongs to its own larger whole, how the object is really the *discovered compatibility - of several larger rhythms*, sustained under a stamina budget (align well **and** align long). +## Orthogonal to classical reduction -To describe an object-under-a-cut, five questions (the cut — boundary × timescale × scope — having -been declared first): +This runs perpendicular to the habit classical physics trained into us. The reductive default +isolates a system at a **static object-boundary** — this mass, that charge, this cell — treats the +object as a persisting thing with fixed identity, and explains behaviour by **cause and effect +between such objects**: A strikes B, B moves. The object comes first and is real; behaviours are its +properties; causation runs between objects. -1. **What behavior** does it constitute, read along time, space, quantity, or their compounds? -2. **Which part owns which axis**, and at what site do the axes recombine (the meeting-site that owns - none of them)? -3. **Over what timescales at once** — the object is verified concurrently at several grains, each a - store the faster fills and the slower reads. -4. **As what alignment project** — what are the parts tuning toward, within which larger wholes, and - under what stamina budget? -5. **How does it turn day to night** — day constitutes the object from behavior; night re-evokes it - as a probe to decide what structure to keep. +Here the order is inverted. **Behaviours come first** — read at cuts — and the object comes last, as +a name over some of them. There is no privileged object, not because we must choose among many +objects, but because objects were never on the mechanism's side at all: only behaviours are, and +"object" is our bookkeeping. Ask "what is the neuron trying to achieve," and any answer is partial — +the neuron is part of an assembly it cannot see, part of an organ, part of an organism, with no top +where the question closes, and made of synapses and channels, with no bottom either. The question +presumes a privileged object; there is none. This is the descriptive face of the model's founding +principle (logic_principles, Part I): as there is no global state and no privileged actor *inside* +the system, there is no privileged *object* from which to *describe* it. A name-over-cuts is to +description what one history is to simulation — the only tractable thing, necessarily partial, +honestly chosen. -And, throughout: **what does this cut push into context** — what it cannot see, which a larger or -smaller cut would. Naming the blind spot is part of describing the object honestly. +Two riders. First, ordinary cause and effect still holds *within a single cut* (at the ms synapse +cut, this release causes that response); what has no clean analogue is cause/effect *between the +named objects* — the synapse does not cause its parts, it is a name over their behaviours. Second, +the useful cuts are not arbitrary: they fall at the **joints**, where the system's own coupling is +denser inside than across (a synapse's three components interact more with each other than with the +next synapse). We prefer these because they carve where the coupling already is — but they remain +cuts, still level-choices, still partial. There are affinities here with the scale-relative frontier +of physics — the renormalization group, non-equilibrium thermodynamics — which also make description +depend on the scale of the cut; but as the simulation argument showed, those point in the direction +without solving this system. They tell us cut-relative description is legitimate physics; they do not +hand us the object. -The objects below are the ones where these have interesting, non-obvious answers. +*The pseudocode is itself the finest cut* — it reads behaviours at the single-component level and +treats every larger whole as context arriving at the boundary (dopamine, the day/night context, the +renormalization are the organism and the hypothalamus reduced to inputs). This document lays names +over *coarser* collections of cuts in the same web. +## How each object is written +Given the machinery, each object below has one shape: **the name, then a series of `cut → behaviours` +entries, then the note that the name is simply the collection.** The compound observables — a +**frequency** (events per time), a **flow** (quantity per time), an **elapsed interval**, an **amount +at a moment**, a **spatial extent**, a **coincidence** — are just behaviours read along one or more +axes at a stated cut. Each entry names its cut (the levels on time, space, quantity, and the scope) +and reads the behaviours that cut makes verifiable. The object is nothing over and above the set. --- ## Object 1 — The Synapse (constituted by PRE ⇄ POST ⇄ ASTROSYNAPSE) -*The cut. **Boundary:** three components — one bouton, one spine, one perisynaptic process — treated -as the object; everything larger (both neurons, the astrocyte, the organism) enters only as arriving -signals (glutamate schedule from PRE's neuron, depolarisation from POST's, alpha and spike from the -astrocyte, dopamine and day/night from above). **Timescale:** read across all its native grains -(ms coincidence to overnight structure), which is itself part of what makes it interesting. -**Scope:** both day and night, since the object's identity rotates between them. **Pushed into -context:** why the two neurons fire when they do (that lives in the assembly-cut and the neuron-cut); -what the alignment is ultimately for (that lives in the organism-cut). This cut sees the synapse -trying to align; it cannot see what the alignment serves.* +"The synapse" is a name over the behaviours read across the cuts below. Each cut names its levels — +time, space, quantity, scope — and reads what becomes verifiable there. None of these is more "the +synapse" than another; the word is the collection. Note in passing how much *one* observable, say +frequency, fractures across cuts: frequency of what, at what grain, in which scope — each a different +behaviour, all called "the synapse's." -The synapse is the first unexpressed object: nowhere in the mechanism, everywhere in the behaviour. -Three components constitute it, and it is real only as what they verify together. +**Cut — ms · single-component space · per-quantum · day.** At the finest time grain, cutting the +space at one component and the quantity at single events, the readable behaviours are the raw rates: +the **frequency of NT emission** (PRE's release events, read as PRE's fast-trace level), the +**frequency of channel opening** in POST (its fast trace), the **elapsed interval** since the last +release (how far a trace has decayed), the **quantity** of a single release (occupancy × drive). Each +is one component's behaviour; nothing joint yet. -### Decomposed — the coincidence, separated into three owned axes +**Cut — ms · cleft space · concentration · day.** Widen the spatial level from one component to the +cleft, keep time at ms. Now the readable behaviour is the **flow of NT out of the cleft** and its +**persistence** — the dwell time, set by the astrosynapse's clearance. This is a behaviour no single +component has; it lives at the cleft level, and the astrosynapse is where its shadow is held. "How +long transmitter stays" is a synapse behaviour only at *this* cut. -A coincidence is several things at one time and place. In a two-party contact it would be an -undifferentiated fact — release met response, or it did not. The third party decomposes it, because -each party owns one axis: -- **PRE owns quantity** — how much is released, set by its own occupancy × drive. -- **The astrosynapse owns timing and space** — its clearance sets how long transmitter dwells (the - temporal window); its coverage sets whether release stays contained or spills (spatial isolation). -- **POST owns the recombination** — its NMDA receptor is where quantity (glutamate), the postsynaptic - contribution (depolarisation), and the astrosynapse's gain (D-serine) meet; it reads the - coincidence *as scaled by* the timing and gain the others set. +**Cut — ms · three-component space · joint · day (the third-party cut).** Cut the space to include +all three components at once and read their *joint* behaviour. Now a new behaviour is verifiable that +none of the previous cuts could reach: the **coincidence** — glutamate and depolarisation and +D-serine present together — read at POST's NMDA, which holds none of the three inputs but is where +their shadows overlap. At this cut the three axes have three owners (PRE: quantity; astrosynapse: +timing and space; POST: recombination), so the coincidence reads as *how much, how sharply timed, how +contained*. This is the cut at which "the synapse detects coincidence" is true — and it is true only +here, because coincidence is a joint behaviour and only a three-component space cut makes it +verifiable. (This is the tripartite decomposition of logic_principles §2, now located as one cut +among the synapse's many.) -So the synapse verifies not "did they coincide" but "how much, how sharply timed, how contained" — -three separable reads because three owners. This is why the synapse is tripartite and not bipartite -(logic_principles §2, here made concrete as a division of who-holds-what). And the sites are plural: -quantity is integrated in PRE's release, timing/space in the astrosynapse's clearance and coverage, -the coincidence *event* at a fourth site — POST's NMDA — which holds none of the three axis-stores -but is where their shadows overlap. The astrosynapse is what makes the coincidence witnessable under -locality: it supplies the third input neither coinciding party owns, so the event registers without -any party reading another's interior. +**Cut — tens-to-hundreds of ms · three-component space · burst · day.** Coarsen the time level from +single spikes to bursts. The readable behaviour is now **train-to-train alignment** — does PRE's +*burst* fall in POST's depolarised *window* — and the **frequency of pre-post coincidence over a +train**, not spike by spike. Short-term plasticity is the synapse's behaviour at this cut: the gain +adjusting so the next burst lands better. Synchronisation here is over multiples of spikes; the +single-spike cut could not see it. -### Aligned — the same three axes, in motion, as a sustained project +**Cut — minutes · three-component space · running average · day.** Coarsen time further. The +behaviour is **participation** — is this synapse consistently in the co-active set — read as the +occupancy and tag trajectories accumulating. "Is this synapse reliably aligned" is verifiable only at +the minutes cut; below it there are only instantaneous events, no reliability. -The decomposition is the object at rest. In motion, the synapse is not detecting a coincidence but -**working toward one, repeatedly, along all three axes** — the two sides tuning toward each other so -the next attempt lands better: -- **quantity alignment** — PRE and POST tune the match between how much is released and how strongly - POST responds (release capacity against receptor sensitivity); -- **timing alignment** — the astrosynapse tunes the window (clearance → dwell) so release and - response fall within the same instant; -- **space alignment** — the astrosynapse tunes coverage so the contact is contained, not bleeding - into neighbours. +**Cut — overnight · three-component space · structural · night.** Switch scope to night and the time +level to the consolidation cycle. The behaviours are **build ⇄ release of structure** — PRE's active +zone, POST's receptor field, the astrosynapse's coverage — read against the tag and the replayed +participation. "The synapse consolidates" or "is pruned" is a behaviour of this cut alone: the same +name, now naming an overnight restructuring rather than a millisecond coincidence. -The synapse is the *project of aligning along three axes at once*. And it is never bilateral: each -party belongs to its own larger whole — PRE to its neuron, POST to its neuron, the astrosynapse to -the astrocyte's territory (with its alpha rhythm and territory-wide calcium spike). So the two -synaptic partners are not free agents agreeing to meet; each is already committed to a larger rhythm. -The synapse therefore verifies the **discovered compatibility of larger rhythms** — do PRE's neuron -and POST's neuron (and the astrocyte's territory) turn out to match in *when*, in *how much*, and in -*where*, often enough and sustainably enough to be worth cementing. Strengthening happens where three -larger wholes prove temporally, quantitatively, and spatially compatible at one contact point. The -astrosynapse's "indirect assistance" is precisely its folding of a *third* larger rhythm (the -territory's) into the two-neuron alignment. +**One cut worth dwelling on — the alignment reading (any of the day cuts, read as motion).** Read the +day cuts not as snapshots but as a process, and a single compound behaviour appears across them: the +synapse **aligning along three axes** — PRE and POST tuning the quantity match, the astrosynapse +tuning the timing and space match — and doing so never bilaterally, since each party belongs to a +larger whole (PRE's neuron, POST's neuron, the astrocyte's territory). At this reading the synapse +verifies the *discovered compatibility of three larger rhythms* in when, how much, and where. And +because alignment costs fuel, it carries two success-conditions at once — **precision** (did the +alignment land: the strength tag) and **stamina** (could it be held: the endurance need) — the two +consolidation pathways seen as the two dimensions of one project: be good at coincidence, for as long +as it takes. This is not a different object; it is the day cuts read as a sustained project rather +than a series of instants. -### Concurrent — several alignments at once, at different grains, coupled through stores - -The alignment is not one process but a stack of them, running **at the same time** at different -temporal grains, and this is the hard thing to say plainly: at any instant the synapse is -simultaneously -- attempting a **spike-to-spike** alignment (does this release meet this depolarisation? — ms), -- tracking a **train-to-train** alignment (does the *burst* fall in the depolarised *window*? — tens - to hundreds of ms; synchronisation is over multiples of spikes, not single ones — short-term - plasticity lives here), -- accumulating a **participation** alignment (is this synapse *consistently* in the co-active set? — - minutes; the running average), -- and holding a **structural** alignment (is this worth permanent structure? — the overnight tag). - -They run concurrently but communicate only through stores, and the coupling has a direction: **the -fast attempts deposit into stores the slower processes integrate, and the slow decisions set the -configuration the fast attempts run within.** Last night's structural alignment is this morning's -starting bias (the standing ceilings); the participation average reads the train-level's success; the -train-level tunes the gain the next spike-attempt uses; each spike deposits the fast trace the -train-level reads. So one does not describe them in sequence — one describes a stack of servos, each -closing its loop at its own rate, nested so that slow sets the frame and fast fills the evidence. This -is the timescale ladder (logic_principles §4) read as concurrent alignment rather than as static -decay rates. - -### Under a budget — precision and stamina are the two success-conditions of one project - -Alignment is not free; it costs fuel, and the budget limits how long the synapse can keep trying. So -the project has *two* success-conditions, not one: **align well** (land the coincidence hard enough -to matter) **and align long** (hold the alignment as long as the task demands). These are exactly the -two consolidation pathways the mechanism separates — strength and endurance — now revealed as the two -dimensions of a single project: *be good at coincidence, for as long as it takes*. Strength is -**precision** (did the alignment land?); endurance is **stamina** (could you sustain it?). A memory -must be both — well-aligned and sustainable — which is why the model carries both a dopamine-gated -strength tag (precision, at the significance-deciding sites) and a homeostatic endurance need -(stamina, everywhere). The synapse is not just trying to align; it is trying to align *and hold*, -within a budget that says how long the holding can last. - -### Day to night — constitute, then re-verify - -By day the synapse constitutes the coincidence from behaviour and deposits evidence (the tag at -PRE/POST, the coverage-need at the astrosynapse). By night the same three parties re-run the -coincidence as a probe — replay release, replay response, replay clearance — not to transmit but to -verify whether the alignment still holds when the pattern is re-evoked without the world driving it. -The participation this re-verification measures decides whether each party keeps its structure. So the -synapse's day-role (constitute the alignment) and night-role (re-verify it to consolidate) are the -identical three-party project read for two purposes — the rotation of logic_principles §3, at the -level of the object. - -**The synapse in one line:** *three larger rhythms, meeting at one contact, trying to align along -when / how-much / where — concurrently at four grains, under a stamina budget — so that a coincidence -is decomposed into three owned axes by day and re-verified as a sustained alignment by night.* +**The name is the collection.** Synapse-at-ms, synapse-at-the-cleft, synapse-as-third-party, +synapse-over-a-train, synapse-at-participation, synapse-at-night, synapse-as-alignment — these are +not different objects, and none is the real one. They are one name laid over behaviours read at +different cuts. It is always the synapse; it is always the synapse *at a cut*. What the word buys us +is the ability to move between these cuts without losing the thread — to say "the synapse" and then +choose, by the cut, which of its behaviours we mean. ---