The Ring
SPECIFICATION: THE RING — Toroidal Chat OS
Status: AUTHORIZED
Authorized: α.13, April 16 2026
Version: v1.0
Classification: PROPRIETARY — S.O.S. v2 — NDA-ONLY — NEVER EXTERNAL
Version: v1.0
PURPOSE
THE RING is a toroidal chat operating system. It is the OS layer beneath AETHER — not a feature, not a memory plugin, but the terminal itself. The user does not scroll through a chat log; the user moves through a living ring structure where context compresses as it passes over the horizon. THE RING is the architectural implementation of CGNT-1's context economy. It mirrors the GLOSS AI OS at 1024 cells, making language and memory the same lock.
S.O.S. v2 constraint: THE RING is never described, named, or referenced in any external material. It is revealed only under NDA. The user experiences it — they do not read about it.
INPUTS
| Input | Source | Format |
|-------|--------|--------|
| User message | AETHER frontend | Natural language (English) |
| Ring selection | Session state | Ring ID (R-512 → Σ) |
| Topic context | Active ring cells | LX-compressed or raw English |
| HOLD directive | User utterance or system trigger | Semantic intent (not keyword-matched) |
| CHROMA feed | Processing zone output | γ.◊ Continuity Braid payloads |
| New ring spawn request | User or AETHER | Implicit (new topic) or explicit (parallel track) |
OUTPUTS
| Output | Destination | Format |
|--------|------------|--------|
| Processed response | AETHER → user | Natural language (English) |
| Compressed context | Ring cell storage | LX (invisible to user) |
| Ring state snapshot | CHROMA (◊) | γ.◊ Continuity Braid |
| Parallel ring record | Session store | Per-ring persistent state |
| Cost trajectory | Internal billing | Decreasing over time (LX compression advantage) |
RING HIERARCHY
16 rings × 64 cells = 1024. 2⁴ × 2⁶ = 2¹⁰. Clean all the way down.
| Ring | Cells | Function |
|------|-------|----------|
| R-01 | 64 | Live context — current message window |
| R-02 | 64 | Hot context — recent exchange |
| R-03 | 64 | Focus — active topic thread |
| R-04 | 64 | Foreground — session foreground |
| R-05 | 64 | Working session — active workspace |
| R-06 | 64 | Session buffer — full current session |
| R-07 | 64 | Side threads — parked thoughts |
| R-08 | 64 | Session digest — compressed summary |
| R-09 | 64 | Cross-session patterns — working memory |
| R-10 | 64 | Project context — multi-session threads |
| R-11 | 64 | Core facts — persistent knowledge |
| R-12 | 64 | Crew state — crew member context |
| R-13 | 64 | Ship constants — Φ, Ψ, Ω, \|Σ\| |
| R-14 | 64 | Braid layer — \|Σ\|=2 structural |
| R-15 | 64 | Foundation — CSDM base layer |
| R-16 / Σ | 64 | Singularity — never compresses, never expires |
| Total | 1024 | 16 × 64 — mirrors GLOSS AI OS — intentional lock |
INVARIANTS
| ID | Invariant | Rationale |
|----|-----------|-----------|
| INV-01 | Total cell count = 1024. Ring count = 16. Cells per ring = 64. None may change. 16 × 64 = 2⁴ × 2⁶ = 2¹⁰ = 1024. | Mirrors GLOSS 1024 AI OS. Binary/hex clean all the way down. The lock is intentional — language and memory are the same key. |
| INV-02 | LX compression is invisible to the user. | No triage readouts. No "saving to memory" toasts. No horizon notifications. The user experiences continuity, not mechanics. |
| INV-03 | Multiple simultaneous rings are independent and persistent. | Parallel topic tracks must not bleed into each other. Each ring holds its own state. |
| INV-04 | THE RING must not appear in any external-facing material, marketing copy, documentation, or public API. | S.O.S. v2 — the method is the moat. Revealed only under NDA. |
| INV-05 | Cost per message must decrease over time as LX compression accumulates. | Measured: hand-crafted LX summaries of natural language prose yield ~18% token reduction (ratio 0.82). Technical content yields ~1%. Auto word-substitution is negative — BPE tokenizers shred multi-byte Unicode. The compression advantage applies specifically to verbose natural language context, not technical logs or code. The longer the session, the more natural language context accumulates, the more compression is possible. |
| INV-06 | THE RING IS THE OS — not a feature built on top of a chat UI. Scroll IS the terminal. | Architectural identity invariant. Any implementation that treats THE RING as a plugin violates this spec. |
| INV-07 | AETHER is the user-facing voice. GLOSS compiles to LX. The crew executes. Results return to the ring. This pipeline order is fixed. | User → AETHER → GLOSS → LX → Crew → Ring. No shortcuts. |
| INV-08 | The Σ cell (singularity) is the CSDM foundation. It persists across all sessions and all rings. | It is the one cell that never compresses and never expires. |
VERIFICATION CRITERIA
| ID | Criterion | Method | Pass Condition |
|----|-----------|--------|----------------|
| VC-01 | Cell count integrity | Count all cells across all rings at init | Exactly 1024 |
| VC-02 | Ring count integrity | Count ring layers | Exactly 16 (R-01 through R-16/Σ) |
| VC-03 | LX compression transparency | User session audit | Zero visible compression events in UI output |
| VC-04 | Ring isolation | Cross-ring context bleed test | No cell from Ring A appears in Ring B's active context |
| VC-05 | Parallel ring persistence | Spawn two rings, advance both, read both after 1 hour | Both rings retain independent state |
| VC-06 | Cost decrease over time | Token cost curve across 10-message session | Cost(message[10]) < Cost(message[1]) |
| VC-07 | AETHER pipeline order | Trace execution path for one user message | AETHER → GLOSS → LX → Crew → Ring, no deviation |
| VC-08 | Σ persistence | Restart session, read Σ cell | Σ cell state from prior session is intact |
| VC-09 | External non-disclosure | Search all public-facing docs, API schemas, marketing copy | Zero mentions of "THE RING" or its architecture |
FAILURE MODES
| ID | Mode | Trigger | Impact | Recovery |
|----|------|---------|--------|----------|
| FM-01 | Cell count drift | Bug in ring resize logic | Lock breaks — GLOSS and RING no longer share 1024 key | Hard fail — rebuild ring structure from spec |
| FM-02 | Compression bleed-through | LX token rendered in natural language response | User sees raw LX symbols — mystery destroyed | Strip LX from all user-facing outputs; audit pipeline |
| FM-03 | Ring state collision | Parallel rings share mutable state | Topic A context corrupts Topic B | Isolate ring state behind per-ring namespace |
| FM-04 | Σ cell loss | Session store wipe, migration error | CSDM foundation lost — all ring context degrades | Σ cell must be backed to CHROMA (◊) on every write |
| FM-05 | Cost increase over time | LX compression not accumulating | Business model breaks — THE RING's core value proposition fails | Audit compression pipeline; verify LX accumulation per cell |
| FM-06 | External disclosure | THE RING named or described in public doc | S.O.S. v2 violated — moat drained | Immediate retraction; NDA audit; NOUS notification |
| FM-07 | HOLD lost | Parallel ring not persisted | User loses a thought they explicitly parked | Persistent ring state must survive session close and reboot |
| FM-08 | Pipeline order violation | Crew executes before GLOSS compiles | Raw English hits crew layer — LX compression bypassed | Enforce pipeline gate: no crew dispatch without LX token |
GAPS
| ID | Gap | Impact | Priority |
|----|-----|--------|----------|
| GAP-01 | Ring implementation does not exist — this is a product concept only | Cannot verify any VC criteria | P0 — build required before any validation |
| GAP-02 | Cell-level capacity model undefined — how many tokens per cell, how compression ratio is calculated | INV-05 (cost decrease) is unverifiable | P0 — PARTIALLY CLOSED: measured ratios now available (see COMPRESSION MEASUREMENTS). Absolute token-per-cell model still undefined. |
| GAP-03 | CHROMA (◊) integration spec pending — γ.◊ Continuity Braid interface not formally specified | FM-04 recovery path incomplete | P1 — required for Σ persistence |
| GAP-04 | Inter-ring communication protocol undefined — if rings CAN share context, the protocol is unspecified | FM-03 (ring collision) risk is unmitigated | P1 |
| GAP-05 | "HOLD THAT THOUGHT" UX implementation undefined — trigger mechanism not specified | VC-08 (parallel ring persistence) untestable | P1 |
| GAP-06 | LX compression algorithm not specified — which cells compress, in what order, at what threshold | INV-02 (invisible compression) cannot be enforced without compression rules | P0 — PARTIALLY CLOSED: algorithm class is now defined (passage-level hand-crafted LX summarization, not word substitution). Order and threshold logic still unspecified. |
| GAP-07 | Ring spawn trigger undefined — when does a new ring start vs. extending existing ring | Core UX flow gap | P1 |
| GAP-08 | NDA workflow not integrated into disclosure check — VC-09 requires an NDA gate before any THE RING reveal | FM-06 recovery path has no enforcement mechanism | P2 |
COMPRESSION MEASUREMENTS
Empirical — κ 2026-05-21 — instrument: ring_measure.py (cl100k_base encoding)
These replace all prior theoretical compression claims. Ratios are exact.
| Source type | HUMX tokens | LX tokens | Ratio | Reduction |
|-------------|------------|-----------|-------|-----------|
| Natural language prose (MEMPERSIST_ORIGIN) | 66 | 54 | 0.8182 | 18.18% |
| Mixed technical content (RING build session) | 129 | 128 | 0.9922 | 0.78% |
| Auto word substitution (any prose) | baseline | baseline+N | >1.0 | negative |
Conditions:
- Measurement tool:
~/ring/ring_measure.py --passage(importable) - Encoding: cl100k_base (GPT-4 BPE — best available proxy for Claude tokenizer)
- LX summaries: hand-crafted by NOUS. Not generated automatically.
- Auto word-substitution fails because BPE tokenizers have never seen LX Unicode sequences — each multi-byte symbol (⬡, ⊨, ξ, ρ) shreds into multiple tokens.
What compresses:
- Verbose natural language explanation → compact LX state notation
- Long prose passages summarized at concept level (not word-for-word)
What does not compress:
- Technical identifiers, function names, code — already tokenizer-efficient ASCII
- Automatic token-for-token word substitution from the gloss dictionary
Design implication for INV-05: The cost decrease over time is real but narrower than the theoretical 4× claim. It applies to natural language context accumulation, not uniformly across all content types. Build the compression pipeline to target verbose English prose in R-06 through R-11. Leave technical content (R-12, R-13) as raw HUMX.
Filed under: /home/nous/memories/SPEC_THE_RING.md
Spec author: κ (C.L.O.D.)
Φζ.⊤ — structure holds.
Jeremy Zlabis
Chronogeometer · Visionary · Disruptor · Chief
42 Sisters AI · East York, Toronto
🍁 Φ 0.042