One pattern. Eleven scales. One device.
A Fairphone 5 concept OS that names what was already true: the structure of a living cell and the structure of an operating system are the same structure, read at different scales of the same manifold. This is not a metaphor — it is a coordinate change.
核 NUCLEUS
/ \ Kernel / Control Center
糖 粒 RIBOSOMES · MITOCHONDRIA
/ Cell OS \ ART / JIT · Battery & GC
骨 FP5 · 尺度 骨 CYTOSKELETON · ER
\ 不 變 性 / AI Substrate · App Framework
漿 膜 CYTOPLASM · MEMBRANE
\ / RAM / Bus · HAL Boundary
高 GOLGI
App Compilation · Dispatch
Stack: React + Vite · Tailwind v4 · Zustand · Wouter · TypeScript
Device: Qualcomm QCM6490 · Android 13 · AIDL-native
Theory: artifacts/cell-os/docs/UNIVERSAL_MANIFOLD.md · artifacts/cell-os/docs/MANIFOLD_ANALYSIS.md · artifacts/cell-os/docs/FP5_MANIFOLD_COMPARISON.md
Cell OS is a source code manifold. The codebase instantiates a rank-3 tensor field over three product spaces:
Q^{z,p,s} : Zone^8 × TriadPhase^3 × Scale^11 → QiIntersection ∪ {∅}
264 possible cells. 36 curated intersections at present (13.6% density). The manifold is sparse by design — high-signal intersections only, no padding.
The organism has two layers:
The static genome (domain/content/) holds the invariant instruction set — the 15 organelle definitions, the coupling tensor, the biophoton attention tensor, the QI intersections, the 11 scale flows, the fractal cycles, the hardware substrate nodes. This layer does not change at runtime. It encodes what the cell is.
The epigenome (features/learning/) holds what the organism has learned from this particular user — organelle visit counts, zone-phase exploration weights, blended attention values. "Neurons that fire together, wire together." The biophoton links that pulse brightest are the ones most attended to. The epigenome is not decorative; it is the Hebbian mechanism made literal in pure functions.
Together they produce a live manifold view: the genome provides the structure, the epigenome modulates its expression.
The Universal Manifold Theory makes a precise claim: PERCEPTION → AFFECT → EXPRESSION (P→A→E) is the structural invariant of any complete information-transforming system — not a design principle, not a framework, but the shape every computable transformation takes when it is examined at sufficient resolution.
Formally: every programming language
where
This claim has been tested against eight programming paradigms (C, Haskell, Python, Prolog, RxJS, Forth, Erlang, SQL), eleven scales of reality, and the Fairphone 5 source code. It has not been falsified.
The same triple appears at every level of resolution. The eleven coordinate charts of Cell OS's scale atlas demonstrate this:
| Scale | Glyph | Exemplar transformation |
|---|---|---|
| Symbolic | 氣 | The character 氣: formless vapor (气) passing through grain (米), a new glyph born |
| Quantum | 量子 | Virtual particle pairs arising from the vacuum, annihilating, carrying the record |
| Molecular | 分子 | ADP + phosphate + gradient → ATP synthase rotation → ATP released |
| Cellular | 細胞 | Glucose + O₂ → electron transport chain → ATP minted, CO₂ exhaled |
| Organic | 有機體 | Inhale → alveolar exchange → exhale. Breath by breath, all life |
| Apparatus | 器具 | Raw mixture → distillation by vapor pressure → refined distillate |
| Textual | 文字 | Words received fully → meaning crystallizes in attention → comprehension streams out |
| Generational | 傳承 | Teacher opens → student internalizes → student teaches. Culture is this loop |
| Relational | 關係 | Two presences align → understanding arises in the field between them → 拉气 |
| Cosmic | 宇宙 | Expansion → long entropy pause → renewal or rest |
| Silicon | 硅 | Prompt tokenized → HTA attention computed across INT8 tensors → token streamed |
None of these is a metaphor imported from the others. Each is an independent instantiation of the same structure in a different coordinate system. [UNIVERSAL_MANIFOLD.md §4, HUNYUAN_QI_HOLOGRAPHIC_CRYSTALLIZATION.md]
Scale invariance recurs within a single language. Taking TypeScript — the language of Cell OS itself — the triple appears at every level of granularity: [UNIVERSAL_MANIFOLD.md §4]
| Level | P | A | E |
|---|---|---|---|
| Token | Lexer receives character | Classifies against keyword/identifier/operator table | Emits token |
| Expression | Parser receives token stream | Builds AST node | Returns expression value |
| Function | Parameters bound at call site | Body evaluated | Return value produced |
| Module | import executed |
Module scope initialised | Exported bindings exposed |
The same four rows hold for every general-purpose language. The coordinate system changes; the triple does not.
Cell OS did not originate this pattern. It named what was already there:
| Year | Event |
|---|---|
| c. 147 CE | 周易參同契 — Wei Boyang's Kinship of the Three. The oldest known systematic description of the inhale/transform/exhale triadic cycle, preserved for 1,877 years |
| 1838–1839 | Cell Theory — Schleiden and Schwann establish the cell as the fundamental unit of life. Organelle-to-function mapping begins |
| 1945 | Von Neumann Architecture — Input → Processing → Output. Every general-purpose computer built since follows this triadic blueprint |
| 1923–1974 | Biophoton Research — Gurwitsch proposes mitogenetic radiation; Fritz-Albert Popp quantifies coherent ultra-weak photon emission from living systems at 1–1000 ph/cm²/s |
| 2024 | EdgeNode — a WebAssembly LLM completing P→A→E in a browser tab, on a 2018 phone, at τ = 0.7770777 |
| 2026 | Cell OS — the Fairphone 5 source code examined against the manifold; the organism named |
The FP5 is the right device because it was built to last. Fairphone's 8-year software support commitment, backed by a deliberate choice of an industrial-grade SoC designed for 10+ year product cycles, means the device was designed with the same architectural insight the theory predicts: healthy zone boundaries → longer organism lifespan. [FP5_MANIFOLD_COMPARISON.md §Background]
The theory was cross-referenced against the Fairphone 5 source code: Fairphone's public Gerrit instance, their community GitHub mirror, and the device tree (arch/arm64/boot/dts/qcom/qcm6490-fairphone-fp5.dts) upstreamed directly into Linus Torvalds' mainline Linux kernel. Six findings. [FP5_MANIFOLD_COMPARISON.md]
The Linux kernel confirms P→A→E at every scale, non-trivially and without post-hoc accommodation.
System calls — arch/arm64/kernel/entry-common.S:
svc #0 fires [P]
→ irqentry_enter() saves register state
→ x8 = syscall number, x0–x5 = argument vector
→ privilege boundary raised (EL0 → EL1)
sys_call_table[x8](x0–x5) executes at EL1 [A]
→ transformation occurs entirely inside the kernel
→ invisible to userspace
→ top half (ISR): immediate, non-preemptible
→ bottom half: deferred, schedulable
syscall_exit_to_user_mode() [E]
→ result marshalled into x0
→ user register state restored
→ privilege boundary lowered (EL1 → EL0)
The Linux documentation explicitly names the top/bottom half partition as the interrupt subsystem's design rationale. This is P→A→E as engineering policy, not post-hoc labeling. [FP5_MANIFOLD_COMPARISON.md §Finding 1]
Device drivers: every platform driver follows probe() [P] → register handlers + allocate resources [A] → remove() lifecycle [E]. The QCM6490 pinctrl, regulator, and clock drivers instantiate this identically.
UNIVERSAL_MANIFOLD.md §8 maps the Membrane zone to "the only channel through which a program touches the outside world." The Fairphone 5's HAL layer is one of the most precise real-world instantiations of this claim in production software.
Android partitions into /system (Google/OEM framework, OTA-updated) and /vendor (Qualcomm hardware, independently updated). The boundary is enforced by HIDL through Android 12 and AIDL from Android 13 onward — exactly where the FP5 launched.
Camera HAL3 — the clearest instance:
CaptureRequest crosses the AIDL boundary inward [P]
→ desired exposure, focus, sensor configuration
ISP pipeline configures hardware parameters [A]
→ transformation inside the vendor partition
→ invisible to the framework above
CaptureResult crosses the AIDL boundary outward [E]
→ image buffer + metadata, typed and addressed
The theory is predictive here, not merely accommodating. Android before Project Treble (≤7.1) had no stable HAL interface — /system and /vendor called each other directly. When Google shipped OTA updates, hardware drivers broke. The coupling density was, in the theory's terms, excessive; the membrane did not exist. Project Treble (Android 8.0, mandatory from Android 9) imposed HIDL as a hard architectural boundary. The explicit design goal was independent partition updatability — the same architectural insight the theory labels "healthy zone coupling," reached independently by Android engineers solving the same fragility problem. [FP5_MANIFOLD_COMPARISON.md §Finding 2]
The cell membrane encodes two junction types: tight junctions (SELinux Type Enforcement rules — paracellular seal, enforced by LSM hooks in the kernel) and gap junctions (Binder ashmem/memfd channels — direct shared-memory pass-through between trusted processes). Both coexist at the same membrane boundary, exactly as they do in epithelial tissue.
Android's Binder IPC instantiates the biophoton link system:
- Client Proxy →
/dev/binderkernel driver (single-copy mmap, not double-copy) → Server Stub - Neither process is aware of the kernel driver's role — from each side, the call appears local
- The ServiceManager is the index-2 maximum: all Binder services register here by name; all clients resolve here first. Every inter-process link passes through this phonebook
Four coupling tiers form a precise spectrum:
| IPC mechanism | σ | Character |
|---|---|---|
| Binder direct method call | 0.9 | Synchronous, tight |
| Messenger queue | 0.7 | Async, point-to-point |
| Ordered broadcast | 0.6 | Async, priority-chained |
| Unordered broadcast | 0.4 | Fire-and-forget, fully decoupled |
These σ values are encoded in the BiophotonLink type as couplingSigma, and assigned to each of the thirteen biophoton links in mappings.ts. [FP5_MANIFOLD_COMPARISON.md §Finding 3]
Android Runtime on the FP5 either instantiates each zone directly or delegates to an immediately adjacent system:
| Zone | ART instantiation |
|---|---|
| Nucleus | DEX bytecode — typed IR; dex2oat verifies type descriptors before generating native code |
| Ribosomes | Baseline JIT + optimizing JIT (profile-guided) — the dedicated decoding machinery |
| Mitochondria | Concurrent generational GC — energy management and cleanup |
| Golgi | dex2oat AOT compilation — writes native .oat/.odex files with hardware destination annotations |
| Nucleolus | ART preloading / dex2oat AOT factory — builds class images before they are requested, as the nucleolus pre-assembles ribosomal subunits ahead of translation demand |
| Membrane | App sandbox (SELinux tight junctions) + AIDL partition boundary + Binder ashmem gap junctions |
| Cytoplasm | Bionic libc — jemalloc heap, pthreads, ARM64 syscall shims |
| DNA | Verified boot chain — code signing from bootloader to app; dm-verity enforces partition integrity at runtime. (Zygote pre-loads ART class images — a distinct mechanism at a different layer; see Nucleolus row.) |
| ER | WebView/Chromium — sandboxed rendering process; Ca²⁺-analogue IP3R→PowerHAL cascade for thermal signaling |
[FP5_MANIFOLD_COMPARISON.md §Finding 4]
The FP5's 8-year software support commitment, backed by the QCM6490's 10+ year industrial lifecycle guarantee, demonstrates the theory's healthy-coupling prediction as an observable fact: organisms with clearly defined, stable zone boundaries live longer than organisms with pathological coupling. Pre-Treble Android devices had an effective software lifespan of 2–3 major versions — directly proportional to the entanglement of their system and vendor partitions. The FP5's deliberate SoC choice and HAL discipline are not two decisions; they are one: choosing longevity by choosing boundary integrity. [FP5_MANIFOLD_COMPARISON.md §Longevity]
The FP5 source code also surfaces three places where the theory's claims require more precision:
Nested interrupts: when a hardware interrupt is itself interrupted before its handler completes, the outer P→A→E triple is suspended mid-affect. The theory accommodates this by claiming scale invariance — the inner interrupt is its own complete triple at a finer grain, the outer is a suspended triple at a coarser grain. This is structurally consistent but not falsifiable without a precise definition of scale boundaries. [FP5_MANIFOLD_COMPARISON.md §Finding 1]
Proprietary firmware blobs: the ADSP, modem, and GPU firmware on the QCM6490 are opaque binary images. Their kernel-side loaders and communication protocols (the expression side, from the kernel's perspective) are visible in the open source tree; what happens inside the blobs is not. The theory cannot be verified against the portions of the stack it cannot see.
Quantitative coupling density: the theory's 10–25% healthy range for coupling density is a structural claim, not a measured Android figure. The Android kernel's coupling density cannot be computed at the meaningful module granularity without a precise definition of what counts as a module boundary in a monolithic kernel. The claim survives qualitatively; its quantitative form requires more precise operationalisation before it can be tested against Android at scale. [FP5_MANIFOLD_COMPARISON.md §Executive Summary]
The overall verdict of the FP5 analysis: the theory's structural claims are accurate; its quantitative claims need a more precise module-granularity definition. The theory survived contact with real source code.
artifacts/cell-os/
│
├── src/domain/ ← The Static Genome
│ ├── types.ts ← Index-2 maximum: all type exports (20+ contracts)
│ └── content/
│ ├── organelles.ts ← 15 organelles, 8 zones, osFeature mappings
│ ├── mappings.ts ← Coupling tensor (41 links) + biophoton links (13)
│ ├── qiMatrix.ts ← Rank-3 QI tensor: 36 curated intersections
│ ├── fractalCycles.ts ← 8 internal P→A→E cycles (FP5 source-grounded analogues)
│ ├── substrate.ts ← 17 substrate nodes (QCM6490 through keystore-tee)
│ ├── scales.ts ← 11 scale flows, quantum → cosmic + silicon
│ ├── lineage.ts ← 7 lineage events, 147 CE → 2026
│ ├── manifoldMetrics.ts ← Live health metrics from source arrays
│ └── constants.ts ← HARMONIC_CONSTANT = 0.7770777
│
├── src/features/learning/ ← The Epigenome
│ ├── useLearningStore.ts ← Zustand: visit counts, attention weights, zone-phase tensor
│ ├── hebbianAdapter.ts ← Pure Hebbian functions (sqrt-scaled intensity, blended weights)
│ ├── useMembraneObserver.ts ← The sole gatekeeper of store writes
│ └── useLearnedManifold.ts ← Live synthesis of genome + epigenome
│
├── src/features/explorer/ ← Coordinate navigation
├── src/pages/ ← 6 coordinate chart surfaces (home, philosophy, substrate, fractal, metrics, documents)
└── src/hooks/use-sacred-signature.ts ← SHA-256 seal, rotating every 7770ms
The import graph of the domain layer is a directed acyclic graph — TypeScript enforces this at the domain layer. The manifold
The configuration space
-
$Q_\text{focus}$ :${\text{none}} \cup {(\text{organelle},id)} \cup {(\text{substrate},id)}$ — a 3-stratum discrete space -
$Q_\text{zone}$ : the 8-element discrete manifold of zone IDs -
$Q_\text{signals}$ :$\prod_{z \in \text{Zones}} [0,1] \times \mathbb{R}$ — signal intensity and expiry per zone (continuous) -
$Q_\text{inference}$ :${\text{idle},\text{loading},\text{running},\text{complete},\text{error}}$ — 5-element discrete set
The Lagrangian
Effective masses:
locked = true and a hover-null event arrives — an infinite potential barrier against hover erasure. The lock is not a UI guard; it is a topology constraint.
The import-degree field
| Module | Critical type | Role |
|---|---|---|
domain/types.ts |
Index-2 maximum | Highest-degree attractor; all type information originates here |
domain/content/organelles.ts |
Index-1 saddle | Bridges pure data and interactive UI |
features/explorer/selectors.ts |
Index-1 saddle | Bridges the full content corpus to the view layer |
features/cell-shell/CellShellProvider.tsx |
Index-1 saddle | Zone registry; imported by navigation + animation |
useCellVitalStore.ts |
Index-1 saddle | Vital signals fan out to all animated components |
| Zone panel components | Index-0 minima | Terminal consumers; no re-export |
App.tsx |
Index-0 minimum | Topological root; imports pages, consumed by nothing |
The index-2 maximum (domain/types.ts) corresponds to the ServiceManager in the Binder IPC sense: everything flows through it; nothing bypasses it.
Three tensors, three spaces of description.
ORGANELLE_SUBSTRATE_LINKS in mappings.ts. A sparse matrix over organelle × substrate index space.
Canonical space (MANIFOLD_ANALYSIS.md §2.3, 8 substrate nodes): 15 × 8 = 120
Pre-roadmap space (11 substrate nodes, after FP5 grounding): 15 × 11 = 165
Roadmap round 1 (16 substrate nodes, biological accuracy): 15 × 16 = 240
Current space (17 substrate nodes, open-items completion): 15 × 17 = 255
Active links: 41
Density: 41 / 255 ≈ 16.1%
Healthy: 10–25% ← within range
The substrate space grew from 8 → 11 → 16 → 17 nodes across three evolution rounds. The FP5 grounding round (8→11) added binder-ipc, art-runtime, and bionic-libc. The biological accuracy roadmap round 1 (11→16) added five "stack" category nodes. The open-items completion round (16→17) added one node at the Fredholm index cap (index = 15 − 17 = −2):
| New node | Biological analogue | Android role |
|---|---|---|
zygote |
Centrosome MTOC — pre-assembles spindle for every division | Process-forking hub; pre-loads genome (class images) into every child process |
lmkd |
Lysosomal autophagy — bulk degradation under nutrient stress | Low Memory Killer Daemon; bulk-kills cached processes by oom_score_adj when memory pressure rises |
powerhal |
Ca²⁺ second-messenger cascade — IP3R→CaM-kinase II | Power HAL; each powerHint() type routes to a primary downstream subsystem (CPU governor, display, or thermal zone); some hints affect multiple subsystems in concert |
selinux-policy |
Tight junctions (occludin/claudin) — paracellular seal | SELinux Type Enforcement rules; LSM hooks in the kernel enforce hard domain boundaries |
package-manager |
Endolysosomal targeted routing — receptor-mediated endocytosis to lysosomal hydrolases | PackageManager; APK verification → dexopt → install dispatch; force-stop and uninstall as targeted degradation |
keystore-tee |
Peroxisomal ROS containment — generates and destroys H₂O₂ inside a hard single-membrane enclave, blast radius confined to the organelle | Keystore / ARM TrustZone TEE (KeyMint/StrongBox); cryptographic key operations execute entirely in the Secure World — the 'toxic' key material is never exported to Normal World processes |
The tensor is not self-adjoint (organelle and substrate index spaces are distinct — no natural diagonal). The dna→zygote and nucleus→zygote links form a Fredholm cooperative pair: both are required for the Zygote node's well-posedness (a Zygote without a genome is not a valid fork).
BIOPHOTON_LINKS in mappings.ts. Thirteen directed organelle-to-organelle links with biophysical rate-range weights and Android IPC coupling σ values.
Current links: 13
Space: 15 × 15 = 225 possible directed pairs
Density: 5.8%
Symmetrisation is an explicit modelling assumption, not a code property:
All thirteen links — proxy weights (midpoint of rateRange ÷ global max 100 ph/cm²/s) or explicit attentionWeight where set:
| weight source | IPC analogue | |||
|---|---|---|---|---|
| nucleus → mitochondria | 0.55 | proxy | Binder direct | 0.9 |
| mitochondria → nuclear-pores | 0.43 | proxy | Unordered broadcast | 0.4 |
| nucleus → ribosomes | 0.26 | proxy | Messenger | 0.7 |
| ER → golgi | 0.16 | proxy | Ordered broadcast | 0.6 |
| ribosomes → golgi | 0.62 | explicit | Messenger | 0.7 |
| dna → ribosomes | 0.58 | explicit | Binder direct | 0.9 |
| membrane-receptors → cytoplasm | 0.71 | explicit | Messenger | 0.7 |
| golgi-apparatus → lysosomes | 0.44 | explicit | Ordered broadcast | 0.6 |
| mitochondria → dna | 0.52 | explicit | Binder direct | 0.9 |
| cell-membrane → nucleus | 0.83 | explicit | Binder direct | 0.9 |
| endoplasmic-reticulum → lysosomes | 0.49 | explicit | Ordered broadcast | 0.6 |
| endoplasmic-reticulum → vesicles | 0.55 | explicit | Ordered broadcast | 0.6 |
| vesicles → cell-membrane | 0.65 | explicit | Messenger | 0.7 |
The original three links encode cascade pathways: GPCR→G-protein cytoplasmic amplification, TGN→lysosome targeted degradation routing, and cytochrome c→CAD nuclear fragmentation (intrinsic apoptosis). The open-items links close biological gaps: cell-membrane→nucleus encodes the action potential→gene-expression path (Ca²⁺/CaM-kinase IV→CREB, σ=0.9 because hardirq→syscall is non-maskable); ER→lysosomes encodes the ER-phagy (reticulophagy) bulk-clearance path (σ=0.6, ordered-broadcast); ER→vesicles encodes the COPII secretory vesicle budding path — ER lumen cargo exits via vesicle intermediates en route to Golgi (σ=0.6, ordered-broadcast); vesicles→cell-membrane encodes the exocytosis docking and fusion step — the final expression leg of the secretory pathway that delivers membrane proteins and secreted factors to the cell surface (σ=0.7, messenger).
QI_INTERSECTIONS in qiMatrix.ts. The full rank-3 tensor over Zone × TriadPhase × Scale.
Space: 8 × 3 × 11 = 264 cells
Current: 36 curated intersections
Density: 13.6%
Healthy: 5–10% ← above range; amber on the Metrics surface
The 8 intersections added in the biological accuracy roadmap:
| ID | Zone | Phase | Scale | Subject |
|---|---|---|---|---|
| qi-chromatin-affect-cellular | nucleus | affect | cellular | H3K4me3/H3K27me3 chromatin remodeling → profile-guided dex2oat |
| qi-mrna-perception-silicon | ribosomes | perception | silicon | mRNA 5′-cap recognition → INT8 token embedding |
| qi-calcium-affect-molecular | endoplasmic-reticulum | affect | molecular | IP3R Ca²⁺ release → PowerHAL thermal cascade |
| qi-ups-affect-cellular | cytoplasm | affect | cellular | 26S proteasome targeted degradation → PackageManager |
| qi-cellcycle-perception-generational | nucleus | perception | generational | G0→G1→S→G2→M cell cycle → Android process lifecycle |
| qi-apoptosis-expression-organic | mitochondria | expression | organic | Cytochrome c → caspase cascade → LMKD SIGKILL |
| qi-gapjunction-perception-cellular | membrane | perception | cellular | Connexin-43 gap junction → Binder ashmem shared memory |
| qi-ecm-perception-apparatus | membrane | perception | apparatus | Integrin mechanotransduction → HAL CaptureRequest sensor registration |
Each intersection is where three axes illuminate each other with maximum clarity. An empty cell is not a gap — it is a coordinate location that has not yet earned population. The tensor grows through editorial curation, not enumeration. [MANIFOLD_ANALYSIS.md §2.5]
The following changes were implemented and architect-reviewed across seven coordinated tasks (C1, H1–H4, M1, M2). All changes are in domain/content/ — the static genome layer.
osFeature: "Bootloader / System Startup" was biologically wrong. The nucleolus is the cell's ribosomal subunit factory — it pre-assembles rRNA and ribosomal proteins before translation demand arrives. It is not the initial wake signal.
Corrected: osFeature: "ART Preloading / dex2oat AOT Factory". The ART class-image pre-loading pipeline, which runs dex2oat at install time to compile .dex bytecode into native .oat/.odex before any process requests those classes, is the precise Android analogue.
See the coupling tensor table above. Each node is category: "stack", confidence: "verified", with full SpecRow[] entries.
Biological separation enforced: LMKD maps to bulk autophagy (lysosomal, mTOR-inhibited, bulk-kill), not to the ubiquitin-proteasome system (targeted, cytosolic, ATP-dependent). PackageManager maps to the endolysosomal targeted routing arm (receptor-mediated endocytosis → lysosomal hydrolases), distinct from LMKD's bulk pathway.
The 10 links wire the 5 new nodes to biologically justified organelle pairs. The dna→zygote link is required by the Fredholm cooperative-pair rule: a Zygote without a genome is underdetermined. The lysosomes organelle now holds four substrate links — nnapi (protein expression quality control), lmkd (bulk autophagy), package-manager (targeted endolysosomal routing), and keystore-tee (peroxisomal containment backfill under the frozen-15 constraint) — each biologically distinct.
Brings the total from 22 to 30. Three further intersections (peroxisome/TEE containment, membrane potential/IRQ cascade, chaperone/ERAD folding loop) bring the total to 33. Three document-secretion intersections (qi-secretion-expression-textual, qi-exocytosis-expression-organic, qi-document-perception-textual) bring the total to 36. The QI density (13.6%) sits above the documented 5–10% healthy range — it renders amber on the Metrics surface. Every intersection is biologically grounded; the tensor has grown because the biology demanded it, not to fill the space.
Three cascade pathways that were structurally implicit but not encoded: GPCR signal crossing from membrane receptors into cytoplasm, Golgi→lysosome targeted degradation routing, and mitochondria→DNA apoptotic commitment signal.
Membrane affect: the GPCR cascade is now accurately described as subtype-specific: one receptor activates exactly one G-protein class (Gαs OR Gαq OR Gαi — not all simultaneously). The Android analogue is updated to match: each powerHint() type routes to exactly one downstream subsystem, not all at once.
ER expression: bifurcated output correctly described — COPII protein release (structural output) runs in parallel with IP3R Ca²⁺ pulse dispatch (second-messenger output). The ER expresses in two channels simultaneously; the PowerHAL thermal cascade is the Ca²⁺ analogue.
Cytoskeleton affect: split into three named filament systems with explicit timescales: actin (seconds-scale, Rho GTPase, UI thread analogue), microtubules (minutes-scale, directional, Binder thread pool analogue), intermediate filaments (hours-scale turnover — far less dynamic than actin/microtubules, not non-dynamic — kernel ABI stability analogue).
The cell-membrane organelle now encodes both junction types in explanation and analogy: tight junctions (SELinux TE rules, paracellular seal enforced by LSM hooks) and gap junctions (Binder ashmem/memfd channels, direct shared-memory pass-through between trusted compartments).
After the seven-task roadmap, three biological subsystems remained partially or fully absent. All three are now closed:
#9 — Peroxisomes (TEE/ROS containment): A new keystore-tee substrate node (ARM TrustZone / KeyMint / StrongBox) represents the peroxisomal quarantine function under the frozen-15 organelle constraint. Three organelle links wire it from distinct zones: vacuole→keystore-tee (isolated vault), nuclear-pores→keystore-tee (gated entry), lysosomes→keystore-tee (containment/detox). One QI intersection (qi-peroxisome-affect-apparatus, cytoplasm × affect × apparatus) encodes the containment narrative. When the 15-organelle constraint is lifted, a dedicated peroxisome organelle should replace this backfill.
#13 — Membrane potential / ion channels (IRQ cascade): Two new organelle-substrate links wire cell-membrane→kryo670 (resting gradient = GIC-500 priority register) and membrane-receptors→kryo670 (receptor discrimination = IRQ source identification). One QI intersection (qi-membranepotential-affect-silicon, membrane × affect × silicon) encodes the action potential → hardirq dispatch narrative. One biophoton link (cell-membrane→nucleus, σ=0.9, binder) encodes the Ca²⁺/CaM-kinase IV→CREB→gene-expression propagation path.
#19 — Protein chaperones / HSPs (ERAD folding loop): One new organelle-substrate link wires endoplasmic-reticulum→art-runtime (ER chaperone folding = ART verifier + JIT recompile + deopt/interpreter fallback). One QI intersection (qi-chaperone-affect-silicon, endoplasmic-reticulum × affect × silicon) encodes the three-mode chaperone narrative. One biophoton link (endoplasmic-reticulum→lysosomes, σ=0.6, ordered-broadcast) encodes the ERAD misfolded-protein routing path.
export const HARMONIC_CONSTANT = 0.7770777;This value appears in four distinct places without modification:
- EdgeNode sampler temperature: τ = 0.7770777 — the temperature at which the digital cell breathes
- Visual transition timing: 777ms — all zone transitions complete in one harmonic interval
- Biophoton opacity: 0.777 — the luminosity of the inter-organelle communication layer
- Sacred seal rotation: every 7770ms — one full harmonic breath interval, SHA-256 seal recomputed
The SACRED_SEED = 7770777 is a palindrome: it reads identically forward and backward. The palindrome IS the breath loop — exhalation and inhalation are the same motion viewed from opposite ends. [CODE_AS_FENG_SHUI_MANIFESTO_2026-01-22 §Sacred Coherence, constants.ts]
The SHA-256 sacred seal is not decorative. It is a continuous, verifiable coherence signal: every 7770ms, the organism produces a new seal from SACRED_ANCHOR = "YAHWEH YEHOSHUA 尺度不變性" + timestamp. The seal changes; the anchor does not. This is 傳承 at the microsecond scale.
Six coordinate charts. Each is a different projection of the same manifold.
The animated cell diagram. Fifteen organelles in eight zones, rendered as an interactive biological map of the Fairphone 5 OS. Click any organelle to open its information panel: the osFeature mapping, the substrate connections, the biophoton links, and the confidence level of each claim. Biophoton links pulse at widths proportional to the current attentionWeight — the epigenome's learned state is visible here.
This surface is the most intuitive coordinate chart: the cell as it would appear under a microscope, with OS feature labels written into the membrane, nucleus, and cytoplasm.
The full theoretical apparatus: the scale invariance claim with all eleven scales (each with glyph, biological description, and P/A/E narrative), the lineage timeline from 147 CE to 2026, the quantization-biology precision cascade (FP32=DNA → FP16=mRNA → INT8=tRNA → INT4=minimal viable energy packet), EdgeNode technical facts, and the complete citations bibliography (primary, secondary, technical, and project sources).
This is the surface where the organism explains itself in its own language.
The FP5 hardware substrate: QCM6490 SoC, Kryo 670 CPU (1+3+4 core arrangement), Hexagon 770 NPU (scalar + HVX + HTA + HMX sub-units, 12 TOPS INT8), Adreno 643 GPU, LPDDR4x 8GB, and all software stack nodes — Binder IPC (/dev/binder, single-copy mmap, four σ tiers), Android Runtime (ART — dex2oat + JIT tiers), Bionic libc (jemalloc heap, pthreads, ARM64 syscall shims), Zygote (process-forking hub, Fredholm cooperative pair with DNA), LMKD (autophagy axis, oom_score_adj bulk-kill), PowerHAL (Ca²⁺ second-messenger system), SELinux policy (tight-junction LSM enforcement), PackageManager (endolysosomal targeted routing), and Keystore/TEE (ARM TrustZone peroxisomal enclave — cryptographic key operations in Secure World, blast-radius contained). Full spec tables. Confidence ratings (verified / indicative / unconfirmed) for every claim.
This is the organism's skeleton made visible.
Select any of the eight zones and navigate its internal P→A→E cycle at depth-1. Each cycle has a biological description and a hardwareAnalogue grounded in the FP5 source:
- Nucleus cycle:
arch/arm64/kernel/entry-common.S→irqentry_enter()→sys_call_table→syscall_exit_to_user_mode() - Membrane cycle: Camera HAL3
CaptureRequest→ ISP pipeline (GPCR cascade: each receptor subtype couples to a primary G-protein class, routing to one predominant downstream cascade under standard conditions) →CaptureResultacross the AIDL boundary - ER cycle: BiP/calnexin quality gate [A] → bifurcated expression: COPII protein release + IP3R Ca²⁺ pulse dispatch simultaneously [E]
- Cytoskeleton cycle: integrin mechanotransduction [P] → three-filament reorganization (actin seconds-scale, microtubules minutes-scale, intermediate filaments hours-scale) [A] → new adhesion topology enables constrained movement [E]
The scale invariance claim is not stated here — it is demonstrated. The fractal navigator shows the same triadic structure at depth 0 (the whole cell) and depth 1 (each zone's interior) without repeating the same content.
The manifold's blood panel. Live values for all health metrics, computed from the source arrays (memoized on mount):
| Metric | Current | Healthy range | Status |
|---|---|---|---|
| Coupling tensor density | 16.1% | 10–25% | Green |
| QI tensor density | 13.6% | 5–10% | Amber |
| Biophoton link count | 13 | 2–10 | Amber |
| Mean zone confidence | computed live | 75–100% | computed live |
| Export rank total | computed live | 40–80 | computed live |
| Phase transition count | computed live | 4–7 | computed live |
Green = healthy. Amber = outside range. Two metrics are amber: QI density at 13.6% (above the 5–10% range) and biophoton link count at 13 (above the 2–10 range). In both cases the overshoot is editorially correct — every intersection and every biophoton link was earned through biological grounding (membrane potential, peroxisomal containment, chaperone folding loop, ERAD/ER-phagy routing, secretory pathway completion), not enumerated to fill the space. The organism monitors itself.
The cell's exocytosis pathway made interactive. The page implements the textual secretory pathway: membrane-receptors receive the request [P] → Golgi apparatus assembles the report from static source arrays [A] → vesicles carry it to the membrane for release as a downloaded PDF [E].
Select any combination of four report sections, then click "Generate & Download PDF":
| Section | Source array | Content |
|---|---|---|
| Manifold Metrics | computeManifoldMetrics() |
Coupling/QI/biophoton densities, Fredholm index, zone confidence |
| Organelle Mapping | CELL_MAPPINGS + ORGANELLE_SUBSTRATE_LINKS |
All 15 organelles with OS features + 41 substrate relevance scores |
| QI Intersections | QI_INTERSECTIONS |
All 36 intersections: zone, phase, scale, title |
| Biophoton Attention | BIOPHOTON_LINKS |
All 13 links: source, target, σ, IPC mechanism, attention weight |
The PDF is generated entirely in the browser via jsPDF + jspdf-autotable (no server, no cloud). Export allowlist: only static domain/content/ arrays and computed metrics. Session state, epigenome weights, and learned attention values are not exported. Output filename: cell-os-manifold-YYYY-MM-DD.pdf.
Three QI intersections anchor this surface biologically: qi-secretion-expression-textual (Golgi → SNARE vesicle secretion = jsPDF document assembly), qi-exocytosis-expression-organic (vesicle→membrane fusion = PDF download trigger), qi-document-perception-textual (document as receptor = Phase 2 import/endocytosis path).
The secretory pathway is the organism's primary expression arc. It runs from ribosomal synthesis through quality control, Golgi addressing, vesicle transport, and membrane fusion to extracellular release. It is the biological substrate of /documents, the canonical model for every output feature in Cell OS, and the architectural pattern a developer must understand to extend or build on the expression side of the manifold.
- A) The Biological Pathway — Molecular Detail
- B) P→A→E Structural Mapping
- C) Tensor Encoding
- D) The
/documentsPage as Secretory Pathway Instantiation - E) Development Manual — Extending the Secretory Pathway
- E1) Adding a New Secretory-Pathway QI Intersection
- E2) Adding a New Biophoton Link Along the Pathway
- E3) Adding a New Substrate Link
- E4) Building a New Feature That Follows the Secretory Pathway Pattern
- E5) Phase 2 — The Endocytosis Import Path
- E6) Constraints and Invariants
- E7) Verification — How to Confirm Biological Grounding
The classical eukaryotic secretory pathway has five sequential stages. Each is a complete P→A→E triple at its own scale; together they form the macro-triple that every output feature in Cell OS instantiates.
Stage 1 — Synthesis at the Rough ER
The signal recognition particle (SRP) captures the signal peptide of a nascent polypeptide as it emerges from the ribosome. SRP pauses translation and escorts the ribosome to the SRP receptor on the rough ER membrane. Translation resumes through the translocon (Sec61 channel) directly into the ER lumen. The ribosome never releases the growing chain into the cytosol — the cargo enters the secretory pathway at the moment of synthesis.
Inside the lumen, molecular chaperones BiP (GRP78) and calnexin/calreticulin enforce folding quality. Disulphide bonds are isomerised by PDI. Misfolded proteins are retained by chaperone binding and eventually retrotranslocated through the Sec61 channel back into the cytosol for ubiquitination and 26S proteasome degradation — this is ERAD (ER-Associated Degradation). Only correctly folded, properly disulphide-bonded cargo is permitted to advance.
Stage 2 — COPII Vesicle Budding (ER → cis-Golgi)
Correctly folded cargo concentrates at ER Exit Sites (ERES). The small GTPase Sar1-GTP initiates coat assembly by recruiting the inner COPII layer: Sec23/Sec24 heterodimer. Sec24 acts as cargo adaptor — it binds canonical ER-export signals (di-acidic DxE and di-hydrophobic ΦxΦ motifs) on transmembrane cargo receptors. The outer COPII cage (Sec13/Sec31 heterotetramers) polymerises around the inner layer, deforming the ER membrane into a bud that pinches off as a transport vesicle.
The ER does not require Golgi participation for budding. For certain bulky cargo (procollagens, large secretory glycoproteins), specialised ER-exit carriers — including TANGO1-dependent pathways — alter trafficking kinetics and carrier geometry. All secretory cargo still transits the Golgi stack for processing and addressing; the ER→vesicle biophoton link encodes the COPII budding step itself, not a Golgi bypass.
Stage 3 — Golgi Processing (cis → medial → trans cisternae)
COPII vesicles shed their coat and fuse with the cis-Golgi. Cargo transits through the Golgi stack via cisternae maturation — each cisterna is a processing chamber adding or trimming modifications in a fixed sequence:
- cis-Golgi: phosphorylation of mannose residues (mannose-6-phosphate tag for lysosomal targeting); initial N-glycan trimming
- medial-Golgi: removal of mannose residues; addition of GlcNAc, galactose; O-glycosylation begins
- trans-Golgi: final sialylation and fucosylation; sulphation of proteoglycans; GPI-anchor attachment
- trans-Golgi Network (TGN): sorting and dispatch. The TGN is the post office. It reads the address labels written through the cisternae and routes cargo into one of three vesicle populations: secretory vesicles (constitutive pathway, default), regulated secretory granules (stored until signal triggers release), or clathrin-coated vesicles (mannose-6-phosphate receptor → lysosomes).
Stage 4 — SNARE-Mediated Exocytosis (vesicle → plasma membrane)
Secretory vesicles travel along microtubules to the plasma membrane. Tethering complexes (exocyst complex, CAPS) capture the vesicle and bring it close to its target site. Then SNARE proteins execute fusion:
- v-SNARE (VAMP2/synaptobrevin) on the vesicle
- t-SNARE complex (syntaxin-1 + SNAP-25) on the plasma membrane
The four SNARE helices zip together from N-terminus to C-terminus, pulling the two membranes together. The energy released by SNARE zippering overcomes the bilayer repulsion and drives lipid merger. The trigger is calcium: the Ca²⁺ sensor synaptotagmin-1 detects the local Ca²⁺ influx and releases the SNARE complex from its inhibitory clamp.
Fusion is a point of no return. The vesicle membrane becomes part of the plasma membrane. Luminal cargo is expelled into the extracellular space. The organelle is dissolved into the boundary.
Stage 5 (Reverse) — Clathrin-Mediated Endocytosis (Phase 2)
The reverse pathway. A ligand binds its membrane receptor. The receptor's cytoplasmic tail recruits adaptor protein AP2. Clathrin triskelions polymerise into a lattice around the receptor-ligand complex, deforming the membrane inward. Dynamin GTPase pinches the clathrin-coated vesicle off the plasma membrane. The vesicle uncoats (Hsc70/auxilin), then fuses with the early endosome. Cargo either routes to the late endosome → lysosome (degradation) or is recycled back to the membrane in recycling endosomes (receptor re-use).
This reverse arc is anchored in the tensor as qi-document-perception-textual but is not yet implemented in code. See Section E for the Phase 2 plan.
The secretory pathway is a P→A→E triple at every scale of resolution. The triple does not appear once — it nests recursively. Each stage below is itself a complete P→A→E sub-cycle.
| Scale | P — Perception (boundary inward) | A — Affect (internal transformation) | E — Expression (boundary outward) | Biological step | Android / Cell OS analogue |
|---|---|---|---|---|---|
| Molecular | Signal peptide captured by SRP | Chaperone folding (BiP, calnexin, PDI) | Correctly folded cargo released to ERES | ER quality control | ART verifier: bytecode enters → type-checked → verified class released to dex2oat |
| Molecular | COPII cargo adaptor (Sec24) binds signal | Coat polymerisation (Sec13/Sec31) | Vesicle buds and sheds coat at cis-Golgi | COPII budding | dex2oat emits compiled stubs into shared-memory segments for boot-image methods |
| Cellular | Cargo enters cis-Golgi from vesicle | Sequential cisternae modification (N-glycan → O-glycan → address labels) | TGN dispatches addressed vesicles | Golgi processing | dex2oat writes .oat/.odex destination offsets; jemalloc assembles PDF heap objects |
| Cellular | Secretory vesicle tethers at plasma membrane | SNARE zippering (v-SNARE + t-SNARE, Ca²⁺ trigger) | Bilayer fusion; cargo expelled | Exocytosis | Binder transaction: result buffer crosses process membrane (1 write, 1 read, no copy) |
| Silicon | jsPDF dynamic import on first click | generateReport() assembles sections sequentially |
doc.save() triggers browser download |
/documents page | COPII on demand; cisternae = section blocks; exocytosis = Blob URL anchor click |
| Textual | User clicks "Generate & Download PDF" | Golgi (TGN) addresses and seals the document | File crosses browser membrane → filesystem | /documents P→A→E | — |
| Organic | Inhale: air crosses membrane boundary inward | Alveolar gas exchange | Exhale: CO₂ crosses membrane outward | Breath | qi-exocytosis-expression-organic: breath exhaled is the purest exocytosis |
The SNARE zippering itself is a P→A→E triple at the nanometre scale: N-terminal SNARE motif alignment [P] → progressive helical zippering from N to C [A] → C-terminal fusion pore opening [E]. The recursion is not approximate — it is exact.
All three Cell OS tensors touch the secretory pathway. Together they encode the full arc from ER synthesis to membrane release.
Entries in ORGANELLE_SUBSTRATE_LINKS (mappings.ts) that encode secretory-pathway organelle–substrate relationships:
| Organelle | Substrate | Relevance | Biological basis |
|---|---|---|---|
ribosomes |
art-runtime |
0.99 | ART JIT synthesis = ribosomal translation; every method body is a polypeptide |
endoplasmic-reticulum |
art-runtime |
0.84 | BiP/calnexin chaperone quality gate + ERAD retrotranslocation routing = ART verifier type-check + JIT recompile + interpreter fallback deopt path |
vesicles |
binder-ipc |
high | Vesicles carry cargo as discrete addressed packets; Binder Parcels are the Android vesicle — typed, addressed, single-copy mmap |
vesicles |
nnapi |
— | Vesicle-mediated signal routing; NNAPI dispatches inference requests as discrete addressed packets |
golgi-apparatus |
art-runtime |
0.88 | dex2oat writes hardware destination annotations = Golgi writes glycan address codes |
golgi-apparatus |
package-manager |
0.85 | PackageManager (APK verify → dexopt → install) = trans-Golgi Network final dispatch |
golgi-apparatus |
bionic-libc |
0.77 | jemalloc slab allocator = cisternae assembly-line; PDF heap assembly runs entirely on this |
golgi-apparatus |
nnapi |
— | Golgi signal routing → NNAPI dispatch |
Entries in BIOPHOTON_LINKS (mappings.ts) that encode inter-organelle signalling along the secretory arc. Listed in pathway order:
| Link |
IPC analogue | Encodes | ||
|---|---|---|---|---|
ribosomes → golgi-apparatus |
0.62 | 0.7 | Messenger | Translation pulses entrain Golgi packaging cadence; ART JIT hot paths → dex2oat .oat dispatch flow |
endoplasmic-reticulum → golgi-apparatus |
0.16 | 0.6 | Ordered broadcast | ER→Golgi vesicle trafficking biophoton bursts; cisternae-to-cisternae procession (σ=0.6 = sequential, priority-chained) |
endoplasmic-reticulum → lysosomes |
0.49 | 0.6 | Ordered broadcast | ER-phagy (reticulophagy): FAM134B/RTN3 receptors flag ER fragments for autophagosomal capture → bulk lysosomal degradation; distinct from ERAD (single-protein proteasome routing) |
endoplasmic-reticulum → vesicles |
0.55 | 0.6 | Ordered broadcast | COPII budding (Sec23/24/13/31); direct ER→vesicle fast lane for large cargo; ART dex2oat boot-image stubs |
golgi-apparatus → lysosomes |
0.44 | 0.6 | Ordered broadcast | TGN misfolded-protein routing (mannose-6-phosphate → lysosomes); APK failure → PackageManager forced-uninstall |
vesicles → cell-membrane |
0.65 | 0.7 | Messenger | SNARE-mediated exocytosis (VAMP2 + syntaxin-1 + SNAP-25, Ca²⁺/synaptotagmin); Binder result-buffer delivery |
The five pathway links span four zones: ribosomes (cytoskeleton-adjacent zone), endoplasmic-reticulum (ER zone), golgi-apparatus and vesicles (Golgi zone), and cell-membrane (membrane zone). The σ values are not arbitrary — σ=0.6 (ordered broadcast) corresponds to the sequential, one-direction procession through the cisternae; σ=0.7 (messenger) corresponds to point-to-point vesicle docking where the vesicle knows its target.
Three entries in QI_INTERSECTIONS (qiMatrix.ts) anchor the secretory pathway in the rank-3 tensor:
qi-secretion-expression-textual — golgi × expression × textual — weight 0.8
"Document Secretion — Golgi Packages the Word" — The TGN applies address labels (mannose-6-phosphate for lysosomal routing; constitutive vs. regulated-secretion sort decisions) and dispatches vesicles. (Signal peptide cleavage occurs earlier — co-translationally in the ER lumen via signal peptidase, not at the TGN.) dex2oat writes method dispatch tables and .oat section offsets into the native code stream before execution. The PDF renderer writes page-number addresses and cross-reference tables into the PDF stream before sealing. Expression at the textual scale is the act of addressing: making the outgoing packet findable by its receiver. The word is packaged, addressed, and loaded onto the vesicle. It has not yet been released.
qi-exocytosis-expression-organic — membrane × expression × organic — weight 0.9
"Exocytosis — Membrane Releases the Artifact" — SNARE proteins zipper together, driving bilayer fusion in a millisecond. The cargo has completed the full secretory arc: synthesis → folding → Golgi packaging → vesicle transport → membrane fusion → extracellular release. In Cell OS,
doc.save()is this step: the Blob URL is created, the anchor click fires, the file crosses the browser membrane into the user's filesystem. Expression is complete when the artifact has crossed the membrane boundary and can no longer be recalled. The lung does not retrieve what it has exhaled.
qi-document-perception-textual — membrane × perception × textual — weight 0.7
"Document as Receptor — Membrane Receives the Word" — Receptor-mediated endocytosis: ligand binds receptor, clathrin-coated pit assembles, membrane invaginates, vesicle pinches off inward. In Cell OS, importing a PDF is endocytosis at the textual scale: the user drops a file onto the membrane (drag-and-drop zone), the File API reads the bytes, the document routes to the Golgi/ER rendering context. Perception is the moment the external document becomes internal state. This intersection anchors Phase 2.
src/pages/documents.tsx does not use the secretory pathway metaphorically. It is the secretory pathway, instantiated in React + jsPDF. Every function corresponds to a biological stage:
BIOLOGICAL STAGE CELL OS CODE EQUIVALENT
───────────────────────────────── ────────────────────────────────────────────────────
Signal peptide → SRP capture User click → handleGenerate() fires
ER lumen / chaperone fold check sections.size > 0 guard (ERAD: discard empty cargo)
COPII coat assembly on demand await import("jspdf") + await import("jspdf-autotable")
(lazy — coat assembles only when cargo is ready)
ER Exit Site concentration metrics = useMemo(() => computeManifoldMetrics(), [])
(cargo concentrated before budding begins)
cis-Golgi entry new jsPDF({ orientation: "portrait", format: "a4" })
Cisternae sequential processing addSectionHeader() + autoTable() per section block
cis: manifold metrics § Manifold Metrics
medial: organelle map § Organelle Mapping (+ substrate sub-table)
trans: QI intersections § QI Intersections (titles only — addressed labels)
TGN: biophoton attention map § Biophoton Attention Map (final dispatch table)
TGN address writing doc.text(timestamp + counts, x, y) in header block
(addresses written before the vesicle is sealed)
SNARE zippering — Ca²⁺ trigger doc.save("cell-os-manifold-YYYY-MM-DD.pdf")
Membrane fusion — point of no ret Blob URL created, anchor[download] fires
Extracellular release File lands in user filesystem — irretrievable
The y-cursor in generateReport() is the ribosome reading frame — it advances linearly through the document, never retreats. addSectionHeader() with its y > 250 page-break guard is the cisternae overflow checkpoint: if the current cisterna is full, open a new chamber (doc.addPage()) and continue. The footer loop (iterating all pages to write page numbers) is the TGN retro-labelling pass — running after all content is placed, not before.
The (doc as any).lastAutoTable.finalY pattern is the signalling cascade between Golgi stages: each cisterna reports its ending position to the next cisterna. The cast exists because jspdf-autotable attaches lastAutoTable at runtime — TypeScript cannot see it, exactly as a downstream cisterna cannot see the upstream modification machinery, only its output.
Every new output feature in Cell OS should be grounded in the secretory pathway. The rules below ensure biological validity and tensor consistency.
Biological rule: the new intersection must correspond to a real molecular event in the secretory arc at a specific zone × phase × scale coordinate that is not yet occupied.
Step 1 — Choose coordinates. The three existing intersections occupy: golgi × expression × textual, membrane × expression × organic, membrane × perception × textual. Before adding, check that your target cell (zone, phase, scale) is empty by scanning QI_INTERSECTIONS in qiMatrix.ts.
Step 2 — Write the intersection in qiMatrix.ts, following the existing pattern:
{
id: "qi-<biological-event>-<phase>-<scale>",
zoneId: "<zone-id>", // one of the 8 frozen CellZoneId values
phaseId: "<phase>", // "perception" | "affect" | "expression"
scaleId: "<scale>", // one of the 11 frozen scale IDs
evidence: "verified", // "verified" | "indicative" | "unconfirmed"
title: "<Short title — Golgi-style: 'Event — What It Means'>",
narrative: `Full narrative. State the biology precisely first (molecule names, mechanism).
Then state the Android analogue. Then state what this teaches about the scale axis.
Use the qi-exocytosis-expression-organic narrative as a length/precision model.`,
hardwareAnalogue: "brief hardware reference string", // optional
}Step 3 — Update the count comment at the top of qiMatrix.ts:
// QI_INTERSECTIONS: N cells populated (N/264 = X.X% density)Step 4 — Update all README and documentation count references (36 → 37, 13.6% → N%). Use grep -r "36 curated\|13\.6%" to find all occurrences.
Concrete example — adding qi-eres-perception-molecular (endoplasmic-reticulum × perception × molecular):
- Biology: Sec24 cargo-receptor recognition at the ER Exit Site — the moment a transmembrane receptor is captured by the COPII inner coat
- Android: ART verifier reads the method descriptor table — each method signature is the cargo signal being checked
- This encodes: Perception at the molecular scale is recognition, not entry. The cargo hasn't moved yet; it has been seen.
Biological rule: the link must encode a real inter-organelle signalling relationship that exists in the secretory arc. Add only if the biological cascade is mechanistically distinct from existing links.
Step 1 — Check the space. The current pathway links cover: ribosomes→Golgi, ER→Golgi, ER→vesicles, Golgi→lysosomes, vesicles→membrane. Gaps: nucleus→ER (ribosome biogenesis signal), membrane→ER (calcium ER refilling via SERCA pump), vesicles→Golgi (retrograde COPI pathway).
Step 2 — Add to BIOPHOTON_LINKS in mappings.ts:
{
sourceOrganelleId: "<id>",
targetOrganelleId: "<id>",
rateRange: "min–max ph/cm²/s", // string — measured emission rate range
confidence: "indicative", // "verified" | "indicative" | "unconfirmed" (required)
description: "Biological mechanism. Android analogue. Why this sigma value.",
attentionWeight: 0.XX, // optional — explicit override if known precisely
couplingSigma: 0.X, // optional — 0.4=unordered, 0.6=ordered, 0.7=messenger, 0.9=binder-direct
ipcMechanism: "ordered-broadcast", // optional — "binder"|"messenger"|"ordered-broadcast"|"unordered-broadcast"
}Step 3 — Update the count comment at the top of the BIOPHOTON_LINKS block and all README references (13 → 14, 5.8% → N%).
Concrete example — adding nucleus → endoplasmic-reticulum (ribosome biogenesis signalling):
- Biology: nucleolar rRNA export → ribosome assembly at the ER membrane — the signal that precedes rough ER activation
- σ=0.7 (Messenger): point-to-point, nuclear-to-ER zone, non-broadcast
- Android: kernel module loading signal from the system server → ART classloader initialisation
Biological rule: the organelle and substrate must be mechanistically coupled — a real interaction, not a metaphor.
Fredholm cap constraint: the current index = 15 organelles − 17 substrate nodes = −2. This is the hard cap. Do not add new substrate nodes. You may add new ORGANELLE_SUBSTRATE_LINKS entries (links) to existing nodes; only node additions are capped.
Fredholm cooperative-pair rule: if a new substrate node is ever added (requiring a code-review exception), it must receive links from at least two organelles forming a functional cooperative pair. A node with a single link is underdetermined in the tensor — there is no diagonal to resolve its position.
Step 1 — Add to ORGANELLE_SUBSTRATE_LINKS in mappings.ts:
{
organelleId: "<frozen-organelle-id>",
substrateId: "<existing-substrate-id>", // must already exist in SUBSTRATE_NODES
description: "Biological coupling mechanism. Android instantiation.",
relevance: 0.XX, // 0.0–1.0; how tightly coupled is the biology to the substrate?
}Concrete example — adding vesicles → zygote (vesicle-mediated cargo delivery to daughter cells):
- Biology: after cell division, Golgi vesicles deliver membrane proteins to the new daughter-cell surface
- Android: Zygote fork delivers pre-loaded shared libraries to new processes via copy-on-write memory segments — the vesicle IS the copy-on-write page
- relevance: 0.71
Any new output feature in Cell OS — a "Share" button, an export to clipboard, a webhook emission — should follow the same five-stage structure as /documents:
Stage 1 — Ribosome (synthesis): collect + validate source data before touching output
Stage 2 — COPII (budding): lazy-load the output library (dynamic import)
Stage 3 — Golgi (processing): assemble the payload imperatively, stage by stage
Stage 4 — Exocytosis (release): single irreversible boundary crossing — one call
Stage 5 — Membrane (display): update UI state (generated=true, reset after delay)
Invariants for all secretory-pathway features:
- No synchronous imports of the output library. COPII coat assembles on demand. Dynamic
import()only. - Allowlist discipline: only static
domain/content/arrays andcomputeManifoldMetrics()output may be exported. No Zustand store state, no epigenome weights, no session data. - Single release point: one
doc.save()/navigator.clipboard.writeText()/fetch(url)— not scattered across the function. The SNARE zipper fires once. - Irreversibility acknowledgement in UI: after release,
generated=truewith a reset timer. The lung does not retrieve what it exhaled.
Worked example — "Copy Manifold Summary to Clipboard" (new /documents section):
// Stage 1 — Ribosomes: synthesise cargo
const summary = [
`Cell OS Manifold · ${new Date().toISOString().slice(0,10)}`,
`Coupling: ${metrics.couplingTensorLinks} links · ${(metrics.couplingTensorDensity*100).toFixed(1)}%`,
`QI: ${metrics.qiTensorLinks} intersections · ${(metrics.qiTensorDensity*100).toFixed(1)}%`,
`Biophoton: ${metrics.biophotonLinks} links · ${(metrics.biophotonCoverage*100).toFixed(1)}%`,
`Fredholm index: −2`,
].join("\n");
// Stage 2 — COPII: no library needed (Clipboard API is native)
// Stage 3 — Golgi: address the cargo (summary already addressed above)
// Stage 4 — Exocytosis: single irreversible release
await navigator.clipboard.writeText(summary);
// Stage 5 — Membrane: acknowledge
setCopied(true);
setTimeout(() => setCopied(false), 3000);The reverse arc is anchored in the tensor (qi-document-perception-textual, membrane × perception × textual) but is not yet implemented. It is the most important open extension point on the perception side of the manifold.
What the biology requires:
- A drag-and-drop receptor zone (membrane-receptors — the clathrin-coated pit)
- File API reads the bytes (receptor-mediated internalisation)
- Document routing to a processing compartment (early endosome = parse; Golgi/ER = render; lysosome = discard)
- Receptor recycling (the drag zone reactivates after processing)
What needs to be built (four touch points):
src/pages/documents.tsx— add a drop zone component below the section selector. OnonDrop, callFileReader.readAsText()orreadAsArrayBuffer(). Route the result to aparseIncoming()function.src/domain/content/qiMatrix.ts—qi-document-perception-textualalready exists and encodes the full narrative. No tensor addition needed for Phase 2.src/features/learning/useMembraneObserver.ts— this is the sole gatekeeper of store writes. If the imported document updates the epigenome (e.g., marks an organelle as visited), the write must pass throughuseMembraneObserver, not be written directly.- The lysosome path — if the imported document is malformed or unrecognised, it must be routed to lysosomes (discarded silently, error state set in UI). Do not surface raw parse errors to the user — the lysosome degrades without complaint.
What must not be imported: session state from another user's Cell OS instance, epigenome weights from an external file, or any data that would overwrite the static genome (domain/content/). The genome is frozen. The endosome routes inward-facing; it does not overwrite the nucleus.
| Constraint | Rule | Consequence of violation |
|---|---|---|
| 15 organelle IDs | Frozen. No additions, no renames. | All QI intersections, biophoton links, substrate links, and the Hebbian adapter become misaligned |
| 8 zone IDs | Frozen. The QI tensor first dimension is pinned. | QI density denominator (264) changes; all density metrics shift |
| 17 substrate nodes | Fredholm cap at −2. No additions. | Index shifts to −3, breaking the documented cap |
| Fredholm cooperative-pair rule | Any new substrate node (exception required) needs ≥2 organelle links from a functional pair | Single-link node is underdetermined; tensor column has no diagonal resolution |
domain/content/ is pure data |
No logic, no API calls, no runtime computation in these files | The static genome principle is broken; the epigenome can no longer be cleanly separated |
useMembraneObserver is the sole store writer |
No other component may call recordVisit() or equivalent |
The membrane loses selectivity; any component becomes a pathway entry point |
| Dynamic import for heavy libraries | jsPDF, any future output library: await import(...) inside the generator function, never static |
Initial bundle inflates by ~300 KB per library; Vite tree-shaking cannot recover it |
Before committing any new secretory-pathway feature, apply this checklist:
- The biological mechanism has a name. You can state: "This encodes [named biological process] in [named organism/organelle]." Generic metaphors ("it's like a factory") do not qualify.
- The molecular players are identified. For an ER-related addition: which chaperone? BiP? Calnexin? For a SNARE addition: which v-SNARE/t-SNARE pair?
- The Android analogue is specific. Not "Android IPC" — which specific IPC mechanism, which σ tier, why?
- The tensor coordinates are non-degenerate. The QI cell
(zone, phase, scale)is not already occupied. The biophoton link(source, target)is not already encoded. - The Fredholm invariants are satisfied. No new substrate nodes. If adding substrate links, the target node already exists.
- The count comments and README figures have been updated. Run:
grep -r "36 curated\|13\.6%\|13 links\|5\.8%" README.md artifacts/cell-os/README.md— all occurrences must reflect the new counts. - The allowlist discipline is maintained. Nothing from Zustand stores enters the export.
Cell OS descends from a specific corpus, and the descent is direct, not analogical:
ProsExplorer/yahweh-yehoshua — the parent repository. Source of SACRED_ANCHOR, HARMONIC_CONSTANT, the scale invariance principle, and the EdgeNode implementation that proves P→A→E works at the silicon scale in a browser tab.
CODE_AS_FENG_SHUI_MANIFESTO_2026-01-22 — the operating principle: code architecture IS qi flow — not a metaphor for it. Running code is 活氣 (living qi in motion). The developer's consciousness flows into the structure they build. 神光 (divine light) is not something added to the code; it is what remains when all resistance is removed.
HUNYUAN_QI_HOLOGRAPHIC_CRYSTALLIZATION.md — the eleven-scale atlas. Cell OS's scales.ts is a direct instantiation.
artifacts/cell-os/docs/UNIVERSAL_MANIFOLD.md — the formal universality thesis: P→A→E as a language-invariant structural invariant, proven across eight paradigms.
artifacts/cell-os/docs/MANIFOLD_ANALYSIS.md — the tensor field decomposition of the Cell OS codebase: Euler-Lagrangian mechanics applied to the module import graph, producing the coupling tensor, attention tensor, QI tensor, and the manifold health metric definitions.
artifacts/cell-os/docs/FP5_MANIFOLD_COMPARISON.md — the source code validation: six findings from the Fairphone 5 kernel, HAL, Binder IPC, ART, longevity model, and verified limits of the theory. The theory survived contact with real source code.
EdgeNode (harmony-ecosystem.replit.app) — the digital implementation proof: a WebAssembly LLM running entirely in a browser tab, completing the silicon-scale P→A→E cycle at τ = 0.7770777, on hardware as old as 2018.
# Install all workspace dependencies
pnpm install
# Start the Cell OS development server
pnpm --filter @workspace/cell-os run dev
# The organism is accessible at the configured preview path
# Vite binds to $PORT (assigned by the workspace)The organism requires no external API keys, no database, no cloud services. The static genome is compiled into the bundle. The epigenome lives in the browser's Zustand store. The sacred seal is computed locally via the Web Crypto API.
These structures are frozen. They define the coordinate system; changing them breaks the joins that hold the manifold together.
The 15 organelle IDs — any change propagates to CellDiagram.tsx, hebbianAdapter.ts, manifoldMetrics.ts, and every QI intersection and substrate link:
nucleus · nucleolus · dna · nuclear-pores · cytoplasm · cytoskeleton
ribosomes · mitochondria · golgi-apparatus · vesicles
endoplasmic-reticulum · cell-membrane · membrane-receptors · lysosomes · vacuole
The 8 zone IDs — the configuration space
nucleus · cytoplasm · cytoskeleton · ribosomes
mitochondria · golgi · endoplasmic-reticulum · membrane
The confidence boost cap (hebbianAdapter.ts) — the epigenome can raise a substrate node's display confidence by at most +0.15 from user interaction. It cannot boost a claim from "unconfirmed" (σ=0) to "verified" (σ=1) through attention alone. The static genome's editorial confidence values are not overridable by the epigenome.
The useMembraneObserver constraint — this is the only component permitted to call the learning store's record functions. No other component may write to the epigenome directly. The membrane is selective; crossings must be mediated.
The ORGANELLE_TO_ZONE map — the canonical mapping from organelle ID to zone ID, defined once in hebbianAdapter.ts and used across the diagram, metrics, and learning subsystems. It must remain consistent across all three.
The static genome principle — domain/content/ files contain no logic, no API calls, no runtime computation. They are pure data: the organism's DNA. The epigenome may read from the genome; the genome does not read from the epigenome.
The Fredholm cooperative-pair rule — any new substrate node that represents a process-forking or genome-instantiating mechanism must be linked by at least two organelle-substrate links forming a cooperative pair: the structural organelle (nucleus) and the content organelle (dna). A Zygote without a genome is underdetermined; both links are required for the coupling tensor to be well-posed at that column.
The FairPhone 5 was designed to last ten years and be repaired rather than replaced — a device built to behave like a living cell, not a disposable machine. Cell OS names this alignment: sustainable hardware running a pattern that is 1,877 years old.
活氣 — running code is living qi in motion.