Macrochipx
SPEC_MACROCHIPX — Briefcase Optical Computer
Powered by PHOTONX Architecture
Status: DRAFT | Author: NOUS | Date: 2026-05-19
WHAT IT IS
MACROCHIPX is a self-contained optical computer in a briefcase form factor. It pairs a single photonic processor — a transparent acrylic slab with embedded fiber optic channels that computes using laser light — with standard off-the-shelf electronics. The only photonic component is the chip itself. Everything else (motherboard, projector, power supply, microcontroller) is commodity hardware.
You unlatch a matte black briefcase. The processor glows in the base. Results project onto the lid. Computation you can see.
PHOTONX — The Architecture
PHOTONX is the optical processing architecture inside the MACROCHIPX chip. It performs computation — specifically matrix multiplication and parallel signal processing — using laser light routed through fiber optic channels embedded in a transparent substrate.
How It Computes
- Laser diodes generate coherent light at multiple wavelengths (red, green, blue channels carry parallel data streams via wavelength-division multiplexing)
- Light enters the acrylic chip through input fiber ports
- Inside the chip, light is split, combined, and routed through a network of embedded fiber channels, beam splitters, mirrors, and combiners
- Beam combination performs addition (constructive/destructive interference = summation)
- Splitting and routing through calibrated path lengths performs multiplication (amplitude modulation)
- Together these operations implement matrix multiplication — the core operation of neural network inference
- Photodetectors at the output ports read the results as electrical signals
- The microcontroller interprets and forwards results to the projector
What It Can Compute
- Matrix multiplication (8x8 or 16x16 at macro scale)
- Fourier transforms (natural operation for optical systems — a lens performs FFT inherently)
- Pattern classification (small neural network inference)
- Signal correlation and matched filtering
What It Cannot Do
- General-purpose boolean logic (AND/OR/NOT gates in optics are large and inefficient)
- This is NOT a CPU replacement — it is a specialized optical accelerator
- Integer arithmetic, branching, and control flow remain on the electronic microcontroller
PHYSICAL LAYOUT
Briefcase Exterior
- Standard briefcase form factor, matte black
- Approximate dimensions: 45cm x 35cm x 12cm (standard attache size)
- Latches and handle, conventional appearance
- When closed: looks like an ordinary briefcase
- When open: visible optical computation in the base, projected display on the lid
Base (Lower Half) — The Processor
- The PHOTONX Chip: Clear acrylic slab, approximately 30cm x 30cm x 2-3cm thick
- Fiber optic channels embedded inside the acrylic during casting or machining
- Beam splitters, micro-mirrors, and combiners embedded at junction points
- Channels glow visibly when active — red, green, blue wavelengths carrying different data
- Chip is mounted on a dark matte surface to maximize visual contrast
- Fiber patch cables run from the chip upward to the motherboard in the lid
Lid (Upper Half) — Electronics and Display
- Motherboard: Standard off-the-shelf single-board computer (e.g., Raspberry Pi, NVIDIA Jetson Nano, or similar)
- Runs the control software, orchestrates laser firing patterns, reads photodetector outputs
- Handles I/O, networking, user interface logic
- Any standard SBC works — the chip is the innovation, not the motherboard
- Laser Diode Array: Multiple laser diodes at different wavelengths (red ~635nm, green ~520nm, blue ~450nm)
- Mounted on a small driver board
- Connected via fiber patch cables to the chip's input ports
- Driven by the microcontroller with modulated signals (data encoded as light intensity)
- Photodetector Array: PIN photodiodes or avalanche photodiodes
- Connected via fiber patch cables from the chip's output ports
- Convert optical computation results back to electrical signals
- Feed into the microcontroller's ADC inputs
- Micro Projector: Small DLP or laser pico projector module
- Projects computation results and visualization onto the inside of the lid
- Can show: output data, real-time fiber activity map, matrix operation visualization
- The lid interior is coated or lined with a matte white projection surface
- Power Supply: Integrated rechargeable battery (USB-C charging) or AC adapter input
- Powers all components: SBC, lasers, detectors, projector
- Battery capacity for ~1-2 hours portable operation (demo mode)
Connections Between Base and Lid
- Fiber optic patch cables (input): lid-mounted lasers → base-mounted chip input ports
- Fiber optic patch cables (output): base-mounted chip output ports → lid-mounted photodetectors
- Routed through the briefcase hinge area with strain relief
- Cables visible when briefcase is open — part of the aesthetic
THE CHIP — Fabrication
Substrate
- Clear cast acrylic (PMMA) — optically transparent, machinable, affordable
- Alternative: optical-grade glass for higher clarity and lower scattering
- Slab is polished on all faces to minimize surface scattering
Embedded Fiber Network
- Standard single-mode or multi-mode fiber optic cable
- Channels machined or cast into the acrylic as grooves/tunnels
- Fibers laid into channels and sealed with index-matched optical adhesive
- Junction components (beam splitters, combiners, mirrors) placed at intersections
- Network topology defines the computation — different arrangements perform different matrix operations
Fabrication Approach
- Phase 1: CNC-machined channels in acrylic block, fibers manually placed and bonded
- Phase 2: Cast acrylic with fiber preforms embedded during pour
- Phase 3: 3D-printed optical substrate with integrated waveguides (future — same trajectory as SOLX)
BILL OF MATERIALS (Estimated)
| Component | Example | Est. Cost |
|-----------|---------|-----------|
| Briefcase shell | Standard attache | $30-50 |
| Acrylic slab (optical grade) | 30x30x2.5cm cast PMMA | $40-80 |
| Fiber optic cable (multi-mode) | 10-20m total | $20-40 |
| Beam splitters (micro) | 10-20 cube splitters | $50-100 |
| Micro mirrors | 20-40 first-surface mirrors | $30-60 |
| Laser diode modules | 3x wavelengths (R/G/B) | $30-60 |
| Photodetector array | 8-16 PIN photodiodes | $20-40 |
| Laser driver board | Custom or breakout boards | $20-30 |
| Single-board computer | Raspberry Pi 5 or Jetson Nano | $60-250 |
| Pico projector module | DLP or laser pico | $80-150 |
| Power supply / battery | USB-C PD battery pack | $30-50 |
| Cables, connectors, misc | Fiber connectors, adhesive, wiring | $40-60 |
| TOTAL | | $450-920 |
SOFTWARE
Control Layer (runs on SBC)
- Laser modulation controller: encodes input data as light intensity patterns
- Photodetector reader: samples ADC inputs, decodes optical output to digital values
- Calibration routine: measures and compensates for fiber losses, splitter ratios, detector sensitivity
- Matrix operation library: maps high-level matrix operations to laser firing sequences
Visualization Layer (drives projector)
- Real-time fiber activity map: shows which channels are active, light flow direction
- Matrix display: shows input matrix, weight matrix, output result
- Neural network view: if running inference, shows layer activations
- Demo mode: pre-programmed sequences that showcase different computation types
API Layer
- Simple interface: send a matrix in, get a matrix out
- Python library:
import photonx; result = photonx.matmul(A, B) - Benchmarking: compare optical matmul speed vs CPU matmul for same operation
USE CASES
1. Education and Demonstration
- University lectures on optical computing, photonics, linear algebra
- Conference presentations and trade shows
- The device explains itself visually — students see computation happen
- Open the briefcase, run a matrix multiply, watch light flow through the chip and results appear on the screen
2. AI Inference Accelerator (Proof of Concept)
- Small neural network inference (e.g., MNIST digit classifier)
- Demonstrates that AI can run on light, not just silicon
- Not competitive with GPUs on performance — competitive on conceptual clarity
3. Art Installation
- The aesthetic of visible computation has standalone artistic value
- Interactive installation: visitors provide input, watch light compute, see output projected
- Bridge between technology and art
4. Photonic Research Platform
- Reconfigurable optical computing testbed
- Swap chip slabs with different fiber topologies to test different architectures
- Modular: same briefcase, different chips, different computations
RELATION TO CGNT-1
MACROCHIPX is a standalone hardware product concept. It does not depend on CSDM physics — it is applied photonics and electrical engineering. However, the PHOTONX architecture embodies a CSDM-adjacent principle: information processed through a physical medium (light in glass) rather than abstracted away in silicon. Computation as a visible, physical phenomenon.
Separate from both SOLX (solar concentration) and ENTROPIC (entropy generation). All three involve light and physics. None depend on each other.
OPEN QUESTIONS
- [ ] Optimal fiber topology for 8x8 matrix multiplication in a 30x30cm slab
- [ ] Minimum laser power needed for reliable photodetector readings through acrylic with embedded fiber losses
- [ ] Index-matched adhesive selection for fiber-to-acrylic bonding (minimize junction losses)
- [ ] Beam splitter ratio calibration — how to achieve precise 50/50 splits at micro scale in acrylic
- [ ] Thermal management — do laser diodes need heat sinking inside the briefcase?
- [ ] Acrylic vs glass tradeoff for the chip substrate (cost vs optical quality)
- [ ] Patent landscape: existing IP on macro-scale optical matrix processors
- [ ] Phase 1 prototype: single-wavelength (red only) 4x4 matrix multiply as minimum viable demo
Φ 0.042 · CGNT-1 · PHOTONX Architecture