11 symbols · 17+ languages · sub-microsecond

The Quantum Gutter

Quantum beyondBINARY notepad
{+1, 1, -1, +0, 0, -0, +n, n, -n}

dimensional code addressing · charm visual layer

Every line of code has a structural identity. The beyondBINARY prefix system classifies it in real time — imports, loops, returns, I/O — using 11 symbols that work in 17+ languages.

0
Lines / sec
9
Prefix Symbols
20
Languages
0%
Avg Coverage
0
Dependencies

The 11 symbols

Every line maps to exactly one prefix. No ambiguity. No configuration. Universal across all languages.

drag to rotate · scroll to zoom · hover for detail

Code → Quantum Pipeline

From raw text to spatially-addressed, navigable, exportable code in 5 stages.

📄
Raw Code

Any language, any file

🔬
Classify

Regex → category per line

Prefix

11-symbol structural tag

🧊
Navigate

Slice / Cube / Orb views

📡
Broadcast

Cross-app sync + IoT

Try it live

Edit the code on the left — watch the quantum gutter classify every line in real time.

Source Code
Quantum Gutter

Benchmark: classify pipeline

Real-time performance metrics from this browser. Click "Run" to re-benchmark.

Visualize the quantum space

Three ways to see code structure — each canvas is live and interactive.

🧊 Slice11×11 prefix grid · click to navigate
🔮 Orb120 particles · orbital density
⬡ Hex Streamlive data flow · prefix-colored

Run on Real Quantum Hardware

Build a quantum circuit from the prefix classification of your code, then execute it on a local statevector simulator or IBM Quantum's 127-qubit Eagle processors (free tier: 10 min/month).

// Click "Build Circuit" to generate from the live editor above // Or type code in the editor and build a quantum circuit from its prefix structure // // Each line's prefix maps to a quantum gate: // +1: → H (Hadamard) 1: → CNOT -1: → X // +0: → Rz 0: → I -0: → S // +n: → T n: → SWAP -n: → M // +2: → CZ +3: → Y
Ready
Measurement Histogram

Phase 6: Local Simulator Baseline

Five real code samples — one per language — each classified by the 11-symbol prefix system, converted to a quantum circuit, and simulated with 1,024 measurement shots on the in-browser QSim statevector engine. These baselines are the reference distributions that IBM Quantum QPU results are compared against.

Language Qubits Gates States Coverage Top State Sim Time
Python 2 21 2 67% |00⟩ 50.8% 1 ms
JavaScript 5 28 2 39% |00101⟩ 51.5% 2 ms
Rust 3 27 2 37% |010⟩ 52.1% 4 ms
HTML 2 22 1 18% |00⟩ 100% 1 ms
Shell 4 25 2 60% |0001⟩ 52.6% 1 ms
Average 3.2 25 2 44% 1.8 ms
Self-Fidelity Check
Python sample, two independent runs of 2,048 shots each:
99.96% Hellinger fidelity
Confirms the QSim statevector engine produces consistent, deterministic baselines suitable for QPU comparison.
Cross-Language Hellinger Fidelity Matrix
PythonJSRustHTMLShell
Python100%0%0%51%0%
JS0%100%0%0%0%
Rust0%0%100%0%0%
HTML51%0%0%100%0%
Shell0%0%0%0%100%

Each language produces a unique quantum fingerprint. The 0% cross-language fidelity confirms that different prefix distributions map to structurally distinct circuits — a Python class hierarchy and a Bash pipeline occupy completely different regions of Hilbert space. The 51% overlap between Python and HTML comes from both having many unclassified lines that map to identity gates, collapsing amplitude into the same |00⟩ ground state.

How kbatch feeds the quantum pipeline

kbatch is the world keyboard analyzer — 15 international layouts, IBM Quantum bridge, dictionary lookups, and typing drills — all in a single HTML file. Its Quantum tab is where the bridge was born: typing patterns get converted to quantum gates (key transitions become rotation angles, finger movements become entanglement), and circuits are submitted to IBM's 127-qubit Eagle processors on the free developer tier.

That same QuantumBridge logic has now been promoted into the shared quantum-prefixes.js module as QuantumPrefixes.QSim and QuantumPrefixes.submitToIBM(). Every app in uvspeed can now generate and run quantum circuits — kbatch does it from keyboard data, this page does it from code prefix structure, and notepad does it from notebook cells.

kbatch
Typing patterns → gates
15 layouts × 10 min IBM
quantum-gutter
Code prefixes → gates
5-lang calibration bench
📓
quantum-notepad
Notebook cells → gates
PWA export with prefixes
🔗
quantum-prefixes.js
Shared module
QSim + submitToIBM + fidelity
Analysis: What the baselines tell us

Low state counts (1-2) — The circuits are shallow (prefix-driven, not deeply entangled). Most amplitude concentrates in a few basis states. This is favorable for QPU comparison because shallow circuits are less sensitive to decoherence noise on real hardware.

Coverage drives circuit richness — Python (67%) and Shell (60%) produce 2-state superpositions with near-even splits. HTML (18% coverage) collapses entirely to |00⟩ because most lines are unclassified, mapping to identity gates that preserve the initial state. Higher prefix coverage = more interesting quantum behavior.

IBM QPU expectations — On ibm_brisbane (127-qubit Eagle), we expect 70-95% Hellinger fidelity for these 2-5 qubit circuits. The free tier gives 10 minutes/month of execution — enough to run all 5 calibration circuits multiple times and track fidelity trends across backend calibration cycles.

Live Calibration (click to re-run with your browser)
Language Qubits Gates Local States QPU States Fidelity Status
PythonReady
JavaScriptReady
RustReady
HTMLReady
ShellReady
Average

Stenography Engine — Concept Atoms

Language compressed to its atomic meaning. Words → concept symbols → frequency signatures → universal understanding. 53 concept atoms across 8 dimensions map all human expression to a periodic table of meaning.

Concept Stream
Compression
Ratio
Savings
Density
Atoms
Unmapped
Solfeggio Frequency Spectrum
174
285
396
417
432
528
639
741
852
963
Neural Flow
Periodic Table of Meaning — 53 Concept Atoms

ML Export — Cross-Tool Intelligence

Unified training pipeline. Pulls quantum prefix classification, typing biometrics, persona context, financial/therapeutic signals, and cache intelligence from every connected tool — exports as a single ML-ready dataset.

🧠 Cache Intelligence
Words trained: 0
Avg relevance:
Avg difficulty:
Quantum gates:
🌍 Persona Context
Layers active: 0/9
Total interactions: 0
Top concerns:
Financial pressure:
⚛️ Prefix Engine
Classified lines: 0
Coverage: 0%
Gate distribution:
Top category:
KBatch Search History Contrails LiveDict Encoder CacheIntel PersonaContext StenoEngine CapsuleAnalyzer
// Click "Refresh" to scan all connected tools...

17+ language coverage

Classification accuracy measured against hand-labeled corpora. Same 11 symbols, every language.

Prefix ↔ Hardware Data Flow

Each quantum prefix maps to a stage in the GPU/CPU compute pipeline. Code isn't just text — it has a physical address in silicon.

NVIDIA Blackwell Ultra SM Architecture Live in Blackwell App
n: Instruction Cache Entry points, shebangs — first bytes fetched from L1i cache into the warp scheduler. Where code execution begins on-chip. 128 KB L1i
↓ fetch ↓
+n: Warp Schedulers Conditionals, if/else/switch — branch divergence resolved here. 4 schedulers per SM dispatch 128 threads to execution units. 4 sched / SM
↓ dispatch ↓
1: Register File Variable declarations — 256 KB register file per SM holds all thread-local state. The fastest memory on the chip (0-cycle latency). 256 KB / SM
↓ operands ↓
0: Processing Cores Function/method bodies — INT32, FP32, FP64, Tensor Core 5th-gen units. This is where compute happens. 128 FP32 cores per SM. 128 FP32 / SM
↓ execute ↓
-n: Load/Store + SFU Imports, data loading — LD/ST units fetch from memory hierarchy. 32 LD/ST units + 32 SFU for transcendentals (sin, cos, exp). 32 LD/ST units
↓ data ↓
-0: L1 Cache / Shared Mem Return/yield — results staged in 128 KB configurable L1/shared memory. Hot data stays on-SM. Return values live here before writeback. 128 KB L1
↓ writeback ↓
+3: L2 / HBM3e / NVLink Output, print, render — final results flow to 64 MB L2 cache, 288 GB HBM3e (8 TB/s), or NVLink 5 (1.8 TB/s) to other GPUs. 288 GB HBM3e
192
SMs per B200
2.5
PetaFLOPS FP4 AI
8
TB/s HBM Bandwidth
1.8
TB/s NVLink 5
9
Prefix Symbols

Why this matters: When you classify a line as 0: (function), you're implicitly identifying code that will execute on FP32/Tensor cores. A 1: (variable) lives in the register file. A -n: (import) triggers load/store traffic. The quantum prefix isn't just a label — it's a physical address in the compute pipeline. Profiling tools like Perfetto, NSight, and our Blackwell app can correlate prefix distributions with actual hardware utilization.

Quantum Circuit Bridge (v4.3): Every prefix now maps to a quantum gate via QuantumPrefixes.toQuantumCircuit(code, lang). 11 gates total: H, CNOT, X, Rz, I, S, T, SWAP, M + CZ (loops) + Y (output). The circuit can be relayed to a QPU via sendToQPU() over the IoT WebSocket bridge. Code structure becomes a quantum circuit — the gutter is the circuit diagram.

Multi-Architecture Expansion

Same 9 prefixes, mapped to every major compute architecture. The data flow is universal — only the layer names change.

NVIDIA Blackwell

B200 / GB200 — 192 SMs, 4nm, 2.5 PFLOPS FP4
  • n: Instruction Cache (128 KB L1i)
  • +n: 4× Warp Schedulers (128 threads/warp)
  • 1: Register File (256 KB/SM)
  • 0: 128 FP32 + Tensor Core 5G
  • -n: 32 LD/ST + 32 SFU
  • -0: 128 KB L1/Shared
  • +3: 64 MB L2 → 288 GB HBM3e
Phase 1 — Shipped

AMD CDNA 4 (MI350X)

Instinct MI350X — 288 CUs, 3nm, 2.3 PFLOPS FP4
  • n: Instruction Buffer
  • +n: Wavefront Scheduler (64 threads/wave)
  • 1: Vector Register File (256 KB/CU)
  • 0: 128 VALU + Matrix Core 2G
  • -n: Scalar Unit + LD/ST
  • -0: 32 KB L1 + 64 KB LDS
  • +3: 256 MB Infinity Cache → 288 GB HBM3e
Phase 2 — Q3 2026

Intel Xe3 (Battlemage)

Arc B-series — 20 Xe-cores, Intel 4, 12 TFLOPS FP32
  • n: Thread Dispatcher
  • +n: EU Thread Arbiter (8 threads/EU)
  • 1: GRF (128 × 32B per EU)
  • 0: 128 ALUs + XMX Engines
  • -n: Data Port + SLM gateway
  • -0: 192 KB SLM per Xe-core
  • +3: 16 MB L2 → 12 GB GDDR6
Phase 2 — Q3 2026

Apple M4 Ultra GPU

80 GPU cores, 3nm, Neural Engine 38 TOPS
  • n: Tile Dispatch / Vertex Fetch
  • +n: SIMD Group Controller
  • 1: Thread Registers (per SIMD lane)
  • 0: ALU + FP16/FP32 Execution
  • -n: Tile Memory / Imageblock
  • -0: Threadgroup Memory (32 KB)
  • +3: System Level Cache → 192 GB Unified
Phase 3 — 2027

Qualcomm Adreno (Snapdragon X)

Adreno X1-85 — 4nm, 4.6 TFLOPS, Mobile/Quest
  • n: Command Processor
  • +n: Wave Manager (128 threads/wave)
  • 1: Unified Shader Registers
  • 0: SP × ALU / Tensor Pipes
  • -n: UCHE Load/Store
  • -0: GMEM (Tile Memory)
  • +3: System Cache → LPDDR5X
Phase 3 — 2027

QPU (Quantum Processor)

Superconducting / trapped-ion — 11 prefix→gate mappings via toQuantumCircuit()
  • +1: H (Hadamard) — declaration → superposition
  • 1: CNOT — logic branches entangle states
  • -1: X (Pauli-X) — I/O flips classical bit
  • +0: Rz(π/4) — assignment rotates phase
  • 0: I (Identity) — neutral / no-op
  • -0: S (Phase) — comment metadata phase
  • +n: T-gate — modifier precision
  • n: SWAP — import moves data between qubits
  • -n: M (Measure) — unknown collapses state
  • +2: CZ (Ctrl-Z) — loop iteration entanglement
  • +3: Y (Pauli-Y) — output rotates observable
Shipped — v4.3

Phase Plan: Phase 1 (Shipped) — NVIDIA Blackwell SM as interactive layer stack. Live in Blackwell app with click-to-swap architecture selector. Phase 2 (Shipped v4.3) — AMD CDNA 4 CU + Intel Xe3 core layer views. Same stack UI, swappable architecture definitions. Phase 3 (Shipped v4.3) — Apple M4 Ultra Tile-based + Qualcomm Adreno X1 mobile pipeline. Enables Meta Quest / mobile profiling. Phase 4 (Shipped v4.3) — QPU quantum gate mapping. 11 prefix→gate mappings via toQuantumCircuit(). sendToQPU() bridge to physical hardware. Phase 5 (Shipped v4.33) — Quantum Bridge + Jawta biometrics (Morse, whisper, truth analysis), QA walkthrough PWA, compliance engine (HIPAA, OpenSSF, SPDX, SOC2, OSINT), project portfolio capsules (23 Figma/external projects), 121 knowledge capsules, dev/auditor terminal mode, SBOM generation, mobile responsive overhaul across all 24 apps. All phases share the same prefix system — only the hardware layer names change.

Game Blocks → PWA Pipeline

From visual blocks to a deployed PWA app — a collaborative, AI-assisted development workflow built on the prefix system.

Game Blocks Builder Live in brotherNumsy

24 block types across 8 categories — each block maps to a quantum prefix. Drag blocks onto a canvas, connect data flow, tweak parameters, and export to injectable JavaScript. The block graph is the quantum gutter in visual form.
v4.3: Blocks now generate real executable code (gravity, speed, spawn rates).
v4.33: 121 knowledge capsules across 10 categories (languages, domains, ancient, accessibility, encoding, science, acoustic, stroke, compliance, project portfolio). 90+ new LEXICON terms. Neural network agent (nn_simple) trains on gameplay with online weight updates. numsyAI.train(episodes) runs automated training sessions.

1: Entity Player, Companion, Enemy, Collectible — game objects with tweakable properties 4 blocks
0: Physics Gravity, Collision, Bounds, Velocity — simulation rules 4 blocks
+3: Visual Sprite, Particles, Parallax, HUD — rendering output 4 blocks
+n: Logic If Condition, On Event, Random — branching decisions 3 blocks
+2: Loop Game Loop, Spawn Wave, Repeat — iteration 3 blocks
-n: AI AI Behavior, Difficulty Scaling — learning & adaptation 2 blocks
-1: Error Damage, Game Over — failure states and recovery 2 blocks
-0: Output Score Rule — result calculation and display 2 blocks

Concept → PWA: The Distributed Build Pipeline

01

Concept (Blocks)

Drag blocks onto a canvas to define game/app structure. Each block is prefix-tagged. AI or human — same visual language. Export produces quantum-prefixed JavaScript.

brotherNumsy HexBench Pybricks 24 block types
02

Code Generation (Notepad Cell)

Blocks export to a Quantum Notepad code cell. AI refines, fills in sprite data, adds edge cases. Every line gets its prefix. Prefix coverage = code completeness metric.

quantum-notepad AI code gen prefix coverage 95%+
03

Live Preview (HexCast)

Stream the running app via HexCast to any device. Debug in real-time with HexTerm. Measure performance with Blackwell app layer profiling. Iterate without deploying.

hexcast terminal blackwell profiling
04

Collaborative Editing (Synced Notepads)

Multiple users/AIs open the same notebook via BroadcastChannel sync. Each edits different blocks/cells. Prefix gutter shows who changed what dimension. Merge conflicts are structural, not line-based.

BroadcastChannel dimensional diff multi-cursor
05

PWA Build & Deploy

One-click export from Notepad to a self-contained PWA HTML file. Service worker, manifest, icons — all generated. Push to GitHub Pages via the Dashboard. Works on every device: desktop, mobile, Quest, LEGO hub.

PWA manifest GitHub Pages offline-first all devices
06

Distributed App Workflow

The deployed app stays connected to its source notebook. Users tweak individual blocks/layers collaboratively. AI agents can modify parameters in real-time. The app evolves without full redeployment — hot-patch via prefix-aware diffs.

hot-patch AI agents prefix-aware merge distributed

Building toward

The quantum gutter is the foundation. Here's where it's going.

11-Symbol Classifier

Regex-based line classification for 16 languages with dedicated rules (17+ total). Sub-microsecond per line. Runs entirely in the browser with zero dependencies.

Shipped

Cross-App BroadcastChannel Sync

All 24 apps share prefix state via BroadcastChannel('quantum-prefixes'). Open notepad + terminal + hexbench — they stay in sync.

Shipped

3D Visual Navigation (Slice / Cube / Orb)

Navigate code by structure instead of line numbers. Slice isolates Z-layers, Cube shows all 27 cubelets, Orb maps probability density.

Shipped

Quantum Cell Type

Inline Slice/Cube/Orb as a notebook cell. Click, scroll, drag — full spatial code navigation inside any notebook.

Shipped

Game Blocks Builder (brotherNumsy)

24 block types across 8 prefix-mapped categories. Drag-drop canvas, connection graph, preset loader, export-to-code pipeline. Gamified AI integration foundation.

Shipped

GPU Architecture Layer Stack (Blackwell)

NVIDIA Blackwell Ultra SM as an interactive 7-layer visual stack with tweakable sliders, live sparklines, animated data flow. Each layer maps to a quantum prefix.

Shipped

AST-Backed Classification

Tree-sitter AST parsing in Rust (crates/prefix-engine/src/ast.rs). Exact node-type classification for Python, JS, TS, Rust, Go, C. Regex fallback for unsupported languages. Zero false positives.

Shipped — v4.3

Batch-Optimized Engine (Native)

Rust batch-optimized engine (src/simd.rs) with first-byte dispatch table and batch-oriented processing. Compiler may auto-vectorize when built with AVX2/NEON. Target (aspirational): 100M+ lines/sec. Ships as Tauri sidecar + WASM.

Shipped — v4.3

Multi-Manufacturer Architecture Views

5 GPU architectures live in Blackwell app: NVIDIA Blackwell, AMD CDNA 4, Intel Xe3, Apple M4 Ultra, Qualcomm Adreno X1. Click-to-swap layer stacks, specs, heatmaps. QPU quantum gate mapping with 11 gates.

Shipped — v4.3

AI Game Training (Neural Network)

brotherNumsy blocks generate real executable code. nn_simple neural network agent learns from gameplay with online weight updates. numsyAI.train(episodes) API for automated training sessions.

Shipped — v4.3

Dimensional Diff

Quantum Notepad tracks X/Y/Z spatial changes between snapshots. notebook.showDimensionalDiff() shows prefix shifts, not just line changes. Terminal commands: diff / qdiff.

Shipped — v4.3

Quantum Circuit Mapping + IoT Bridge

QuantumPrefixes.toQuantumCircuit(code, lang) maps 11 prefix symbols to quantum gates (H, CNOT, X, Rz, I, S, T, SWAP, M, CZ, Y). ASCII circuit diagrams. sendToQPU() relays to IoT WebSocket bridge.

Shipped — v4.3

IDE Plugins (VS Code, Neovim)

VS Code extension (editors/vscode/) with gutter decorations, status bar, 6 commands. Neovim plugin (editors/neovim/) with Lua extmarks and autocmds. Zed plugin planned.

Shipped — v4.3

Concept → PWA Pipeline

6-stage workflow: Blocks → Code Gen → Live Preview → Collaborative Edit → PWA Build → Distributed App. Stage 1-3 shipped. Stage 5 (PWA Build) shipped via export-pwa command. Stage 4, 6 in progress.

In Progress — Stage 5 Shipped

Physical QPU Integration

Local statevector simulator (QSim, up to 12 qubits) + IBM Quantum REST bridge (127-qubit Eagle, free tier). Hellinger fidelity scoring. Interactive in quantum-gutter.html and kbatch.html. Shared via QuantumPrefixes.submitToIBM().

Shipped — v4.18

IDE Plugin: Cursor MCP + Prefix-Aware Autocomplete

Cursor MCP server with uvspeed_prefix_gaps gap analysis tool. VS Code extension ships circuit visualization (quantumPrefixes.circuit command → QASM export). prefix-autocomplete.mdc Cursor rule for AI-driven gap filling. Zed native extension still planned.

In Progress — MCP + VS Code Shipped

See code in 9 dimensions.

Open the notepad. Paste any code. Watch the gutter light up.