# SynapseOS: Potato-Grade Linux Distro & 7-Lab Native AI Workbench - **Base OS**: Debian 12 Bookworm (`minbase`, `x86_64`) + Xorg + Openbox + Tint2 + `tty1` autologin (`242 MB` live idle RAM). - **Synapse Studio & Synapse Kids Arcade**: Native GTK3 + Cairo 60 FPS AI Workbench (`62.9 MB` full runtime RSS, `3.27 ms/frame` at `1280x720`, `0 KB` texture bloat, `0.09s` cold start) featuring: - **Full-Screen `Synapse Kids` Mode** ([`workbench/kid_arcade_view.py`](file:///home/phil/.gemini/antigravity/scratch/synapse-os/workbench/kid_arcade_view.py)): Gamified Scratch-style classroom UI with 1-click **`L1`–`L7` Level Selector Pills**, interactive canvas clicks across all 7 levels, `NEAT Moon Lander` solar panels/callouts & Lunar Landing Pad touchdown physics, live swirling 2D `Color Sorter Brain`, waving `Cortex-Phi3` Robot Mascot + Speech Bubble, live mission checkboxes, 3-star scoring, and 1-click **English / French (`EN ▸ FR`)** toggle. - **`⚙ Teacher / Pro Workstation Mode`** ([`workbench/studio_gui.py`](file:///home/phil/.gemini/antigravity/scratch/synapse-os/workbench/studio_gui.py)): Full 3-column engineering interface with **35 curated Lab Examples (`5 per lab`)**, **14 Classroom Tutorials**, **Direct Structural Brain Surgery (`+ / - Neurons & Synapses`)**, live **FPS / PRIME C99 tok/s / Phi-3 tok/s & TTFT / RAM telemetry**, **60 FPS Animated Phi-3 + PRIME Brain Visualizer**, **General Chat Mode**, **Safe Linux Command Runner**, and **Encrypted Remote Browser Tunnel (`synapse-tunnel`)**. - **Three-Tier Local AI Engine**: 1. **Tier 1 (`0.68 ms`, `92 KB`)**: SQLite FTS5 BM25 offline knowledge base (`cortex_kb.sqlite`). 2. **Tier 2 (`8,737 tok/s` on BC-250 / `6,738 tok/s` potato CPU, `262 KB` constant $\mathcal{O}(1)$ RAM, zero AVX2 required)**: Native C99 **PRIME Moment Sliding-Window Attention** engine (`libprime.so` / `prime-moment`, integrated from [`batteryphil/PRIME-Moment-Attention`](https://github.com/batteryphil/PRIME-Moment-Attention)). 3. **Tier 3 (`24.8 tok/s` on AMD BC-250 / `11.6 tok/s` 4-core CPU, `2.2 GB`)**: Local `cortex-phi3:latest` (`Phi-3-Mini-4K-Instruct` `Q4_K_M`, `num_ctx=1024`). --- ## 1. Complete Tester Instructions: Download & Boot SynapseOS (`Full 2.6 GB LLM ISO` or `437 MB Lite ISO`) - **1-Click Single-File Full 2.6 GB LLM `.iso` Download (Hugging Face — No Reassembly Needed)**: 👉 **[`SynapseOS-Debian12-Potato-x86_64.iso` (2.6 GB Direct Download)](https://huggingface.co/datasets/batteryphil/SynapseOS/resolve/main/SynapseOS-Debian12-Potato-x86_64.iso)** | **[`Browse Hugging Face Repo`](https://huggingface.co/datasets/batteryphil/SynapseOS/tree/main)** | **[`SHA256 Checksum`](https://huggingface.co/datasets/batteryphil/SynapseOS/resolve/main/SynapseOS-Debian12-Potato-x86_64.iso.sha256)** - **GitHub Release Page (`v1.0-k12`)**: 👉 **[`https://github.com/batteryphil/SynapseOS/releases/tag/v1.0-k12`](https://github.com/batteryphil/SynapseOS/releases/tag/v1.0-k12)** ### Option A: Linux / macOS — 1-Command Full LLM ISO Download & Boot (`2.6 GB`) Includes **everything 100% offline**: Debian 12 + Openbox + the fine-tuned **`cortex-phi3:latest` LLM (`2.2 GB` GGUF)** + Ollama daemon + C99 `prime-moment` engine + 7 AI Labs + all **325 K–12 lessons** + offline `espeak-ng` Read-Aloud TTS. ```bash # 1. Clone the repository git clone https://github.com/batteryphil/SynapseOS.git cd SynapseOS # 2. Download, reassemble, verify SHA256, and boot in QEMU (4 GB RAM, 2 CPU Cores): ./scripts/download_and_boot_full_iso.sh # Or download & verify the .iso file only (for VirtualBox, VMware, or flashing to USB): ./scripts/download_and_boot_full_iso.sh --download # Output file: output/SynapseOS-Debian12-Potato-x86_64.iso ``` ### Option B: Windows (PowerShell) — Download & Combine the Full 2.6 GB LLM ISO Because GitHub Releases caps individual files at `2.0 GB`, the `2.6 GB` Full ISO is split into `part-aa` (`1.85 GB`) and `part-ab` (`706 MB`). Run this in **PowerShell** to download and join them into `SynapseOS-Debian12-Potato-x86_64.iso`: ```powershell # 1. Download both parts and checksum from GitHub Releases curl.exe -L -O https://github.com/batteryphil/SynapseOS/releases/download/v1.0-k12/SynapseOS-Debian12-Potato-x86_64.iso.part-aa curl.exe -L -O https://github.com/batteryphil/SynapseOS/releases/download/v1.0-k12/SynapseOS-Debian12-Potato-x86_64.iso.part-ab curl.exe -L -O https://github.com/batteryphil/SynapseOS/releases/download/v1.0-k12/SynapseOS-Debian12-Potato-x86_64.iso.sha256 # 2. Combine into a single bootable .iso file (binary copy) cmd /c "copy /b SynapseOS-Debian12-Potato-x86_64.iso.part-aa + SynapseOS-Debian12-Potato-x86_64.iso.part-ab SynapseOS-Debian12-Potato-x86_64.iso" # 3. Verify SHA256 hash matches SynapseOS-Debian12-Potato-x86_64.iso.sha256 Get-FileHash .\SynapseOS-Debian12-Potato-x86_64.iso -Algorithm SHA256 Get-Content .\SynapseOS-Debian12-Potato-x86_64.iso.sha256 ``` ### Option C: Boot in VirtualBox, VMware, or from a Physical USB Stick 1. **VirtualBox / VMware / Proxmox**: - Create a new Linux (`Debian 64-bit`) VM with **`4096 MB` (4 GB) RAM** and **`2` CPU Cores** (works down to `1` CPU core). No virtual hard disk is required. - Attach `SynapseOS-Debian12-Potato-x86_64.iso` as the Optical CD/DVD drive and boot. 2. **Flash to a Physical USB Drive (Boot on Any PC/Laptop)**: - **Windows / macOS**: Use [BalenaEtcher](https://etcher.balena.io/) or [Rufus](https://rufus.ie/) (select *DD Image mode* if prompted) and flash `SynapseOS-Debian12-Potato-x86_64.iso`. - **Linux**: ```bash sudo dd if=output/SynapseOS-Debian12-Potato-x86_64.iso of=/dev/sdX bs=4M status=progress oflag=sync ``` ### Option D: Fast Lite ISO (`437 MB`, No LLM — Single File or In-Repo) If you want a fast `437 MB` download without the `2.2 GB` LLM weights (still includes the C99 `prime-moment` engine, `cortex_kb.sqlite`, 7 AI Labs, and all 325 K–12 lessons): - **Direct Single-File Download**: [`SynapseOS-Lite-x86_64.iso` (437 MB on GitHub Releases)](https://github.com/batteryphil/SynapseOS/releases/download/v1.0-k12/SynapseOS-Lite-x86_64.iso) - **Or Decompress From Git Repo**: ```bash ./iso/decompress_iso.sh qemu-system-x86_64 -enable-kvm -m 2048 -smp 2 -cdrom iso/SynapseOS-Lite-x86_64.iso ``` ### What Happens on Boot & Tester Checklist - **Auto-Login**: Boots directly into X11 + Openbox (`~148 MB` idle OS RAM) as user `synapse` (password: `synapse`, passwordless `sudo`) and launches **Synapse Studio** automatically. - **Synapse Kids Mode (Default View)**: - Click **`L1`–`L7`** level pills at the top to play the 7 interactive AI simulations (`Moon Lander`, `Lidar Car`, `GridWorld`, `CartPole`, `Neural Net`, `Predator-Prey Swarm`, `PRIME Attention`). - Click **`♪ Read`** (or click **Lumi the Robot**) to hear offline speech synthesis read the mission aloud. - Click **`EN ▸ FR`** to toggle everything between English and French, or **`❖ Student ↻`** to switch student profiles. - **Teacher / Pro Workstation & Full K–12 School Curriculum**: - Click **`⚙ Teacher`** in the top-right corner to switch to the 3-column Pro/School Workstation. - Open the **`🎓 School`** tab on the right: pick any grade from **Kindergarten (`Yr K`) to Grade 12 (`Yr 12`)** (`325` complete textbook lessons across Math, Science, ELA, History, and AI). - Answer the multiple-choice check using the **`Quiz: [A] [B] [C] [D]`** bar at the bottom of the textbook reader, or click **`▶ Load Lab & Code`** to run and auto-grade the Python worksheet. - Open the **`📋 Report`** tab and click **`✎ Export`** to generate a printable HTML + TXT K–12 Report Card & Honors Certificate. - **Built-In Fine-Tuned `cortex-phi3` LLM + C99 `PRIME` Copilot**: - Type any question into the **AI Copilot Chat** bar at the bottom of the window (or open a terminal and run `ollama run cortex-phi3` or `cortex ai "explain neuroevolution"`). | ISO Artifact | Uncompressed Rootfs | SquashFS | Bootable `.iso` Size | How to Get It | Contents | | :--- | :--- | :--- | :--- | :--- | :--- | | **`SynapseOS-Debian12-Potato-x86_64.iso`** (**Full LLM**) | `3,998.0 MB` | `2,546.1 MB` (`zstd`) | **`2.6 GB`** | `./scripts/download_and_boot_full_iso.sh` or [Release `v1.0-k12`](https://github.com/batteryphil/SynapseOS/releases/tag/v1.0-k12) | Everything in Lite + pre-baked `2.2 GB` `cortex-phi3:latest` GGUF + Ollama runtime + `espeak-ng` TTS | | **`SynapseOS-Lite-x86_64.iso`** (**Lite**) | `1,561.8 MB` | `376.5 MB` (`xz -Xbcj x86`) | **`437 MB`** | [Direct Download](https://github.com/batteryphil/SynapseOS/releases/download/v1.0-k12/SynapseOS-Lite-x86_64.iso) or `./iso/decompress_iso.sh` | Debian 12 + Openbox + 7 AI Labs + 325 K–12 Lessons + `cortex_kb.sqlite` + C99 `prime-moment` (`libprime.so`) | --- ## 2. 7 Native GTK3 + Cairo AI Labs (`workbench/`) — 35 Presets (`5 per Lab`) | Lab | Paradigm | Interactive Parameters & Brain Surgery | Headless CPU Speed (`12 gens`) | | :--- | :--- | :--- | :--- | | **1. NEAT Rocket Moon Lander** | Neuroevolution (`NEAT`) | Gravity, Crosswind, Thruster Power, Safe Landing Speed, Pop Size, Add-Neuron/Synapse Rate, Mutation Power, Target Hidden Neurons | `24.1 ms` | | **2. NEAT Self-Driving Lidar Car** | Autonomous Control (`NEAT`) | Lidar Ray Range, Max Speed, Steering Agility, Track Width, Pop Size, Mutation, Target Hidden Neurons | `189.4 ms` | | **3. RL Hazard GridWorld** | Tabular RL (`Q-Learning` & `SARSA`) | Learning Rate ($\alpha$), Discount ($\gamma$), Exploration ($\epsilon$), Lava Penalty, Goal Reward, Slippery Wind | `8.2 ms` | | **4. CartPole Balance** | Neuro-Policy Control | Gravity, Pole Length, Motor Force, Wind Gusts, Pop Size, Mutation Power, Target Hidden Neurons | `31.6 ms` | | **5. Neural Network Backprop** | Deep Learning (`MLP`) | Dataset (`XOR`, `Circles`, `Moons`, `Spiral`), Hidden Neurons, LR ($\eta$), L2 ($\lambda$), Noise, Batch Steps | `64.5 ms` | | **6. Predator-Prey Swarm** | Co-Evolutionary Boids | Prey Flock Size, Predator Count, Vision Radius, Max Speeds, Cohesion Drive, Food Density | `42.8 ms` | | **7. PRIME Moment Attention** | Hybrid Sliding Window + $\mathcal{O}(1)$ C99 Memory | Context Tokens ($N$), Local Window ($W$), PRIME Moments ($K$), EMA Decay ($\gamma$), Recall Temp, Threads | `14.2 ms` (`8,737 tok/s` C99) | --- ## 3. CLI, Remote Browser Tunnel & Test Commands ```bash # Launch the 7-Lab GTK3 + Cairo AI Workbench GUI (starts in Synapse Kids mode by default; set SYNAPSE_INITIAL_MODE=pro for Teacher Mode) ./synapse studio # Start encrypted HTTPS Remote Browser Tunnel (noVNC + Cloudflare Quick Tunnel) ./bin/synapse-tunnel synapse-ai # Run headless benchmark across all 7 AI Labs ./synapse studio --benchmark --gens 12 # Run the native C99 PRIME-Moment O(1) Attention benchmark (10,000 tokens) ./cortex/c_prime/bin/prime # Query the Three-Tier Cortex Copilot (FTS5 + C99 PRIME Moment + Ollama) ./synapse cortex cmd "compress directory with zstd" ./synapse cortex ai "prime moment attention" # Rebuild the Lite ISO (385 MB, no LLM) or Full ISO (2.5 GB, with LLM) ./distro/build-iso.sh --lite ./distro/build-iso.sh --all # Browse & manage the Complete K-12 Rolling Curriculum (Kindergarten + Grades 1-12) ./synapse pack list-years ./synapse pack set-year K # Install Kindergarten (year_00) ./synapse pack set-year 1 # Install Grade 1 (year_01), wipes old year files, keeps Transcript ./synapse curriculum y00_math # Run Kindergarten Math + AI Lab ./synapse curriculum y08_science_q2 # Run Grade 8 Science Quarter 2 (Winter Unit) + AI Lab ./synapse pack advance-year # Graduate to next School Year (wipes old year, saves Report Card) # Run all automated unit & integration tests python3 -m unittest discover -s tests ``` --- ## 4. Complete K–12 Rolling Year-by-Year School Curriculum (`curricula/year_00.json` – `year_12.json`) Every grade from **Kindergarten (`year_00` / Grade K)** through **Grade 12 (`year_12` / Senior Capstone)** (**13 complete grades**) includes **5 Full-Year Subject Courses** + **20 Quarterly Unit Lessons (`Q1 Fall`, `Q2 Winter`, `Q3 Spring`, `Q4 Capstone`)** = **325 total K–12 lessons** + **91 bilingual `EN`/`FR` visual levels**: 1. **📐 Mathematics** (*Counting & Shapes $\rightarrow$ Arithmetic $\rightarrow$ Fractions/Ratios $\rightarrow$ Algebra & Geometry $\rightarrow$ Pre-Calculus, Calculus & Linear Algebra*) 2. **🔬 Science** (*5 Senses & Weather $\rightarrow$ Forces & Ecosystems $\rightarrow$ Genetics, Waves & Chemistry $\rightarrow$ Physics, Thermodynamics & Computational Neuroscience*) 3. **📖 Reading & Language Arts (ELA)** (*Alphabet Phonics $\rightarrow$ Reading Comprehension & Grammar $\rightarrow$ Claim-Evidence-Reasoning (CER) $\rightarrow$ Rhetoric & Technical Peer Review*) 4. **🌍 Social Studies & History** (*Classroom Civics & Maps $\rightarrow$ Ancient Civilizations & U.S./World History $\rightarrow$ Constitution & Economics $\rightarrow$ AI Ethics, Public Policy & TCO*) 5. **🤖 Connected AI & Coding Assignment** (*Hands-on simulation + auto-graded Python worksheet tied directly to that grade's Math, Science, ELA, and History lessons*) Every lesson contains **6 Complete Textbook Sections**: - **Section 1**: Core Textbook Lesson & National Standards Alignment (`CCSS Math`, `NGSS Science`, `CCSS ELA`, `NCSS C3`, `CSTA CS`) - **Section 2**: Full 36-Week / 4-Quarter Course Syllabus (`Q1 Weeks 1–9`, `Q2 Weeks 10–18`, `Q3 Weeks 19–27`, `Q4 Weeks 28–36`) - **Section 3**: Academic Vocabulary Glossary (Terms & Definitions) - **Section 4**: Step-by-Step Worked Textbook Examples - **Section 5**: Traditional Practice Worksheet & Self-Check Quiz (with Answer Key) - **Section 6**: Teacher Guide & 100-Point Mastery Rubric (`40 pts Core Subject + 30 pts Quiz + 30 pts AI Lab & Python Code`) - **1-Year-at-a-Time Potato Storage Wiper**: Only **one** school year's curriculum package resides in active classroom storage (`~/.config/synapse-os/active_pack/current_pack.json`). Advancing a student (`./synapse pack advance-year` or clicking **`★ Yr X ↻`** in the Teacher Report tab) **wipes the old year's pack and scratch files** while **permanently preserving the student's Cumulative Multi-Subject Report Card Transcript, Gold Trophies, Coins, and High Scores** in `student_profile.json`. --- ## 5. Repository Structure ```text synapse-os/ ├── synapse # Unified CLI entrypoint ├── SYNAPSE_OS_SPECS.txt # Full technical architecture & benchmark specification ├── curricula/ │ ├── index.json # Master manifest of all 13 K-12 School Year Curricula (K-12) │ └── year_00.json .. year_12.json # Hosted Kindergarten + Grades 1-12 complete textbook packages ├── bin/ │ ├── synapse-studio # Studio desktop launcher │ └── synapse-tunnel # Encrypted HTTPS noVNC + Cloudflare remote classroom tunnel ├── iso/ │ ├── SynapseOS-Lite-x86_64.iso.zst.part-* # 5 chunks (<90 MB each, 374 MB total) of the 385 MB Lite ISO │ ├── SynapseOS-Lite-x86_64.iso.sha256 # SHA256 checksum for verification │ └── decompress_iso.sh # 1-command ISO reassembly + zstd decompression + SHA256 check ├── cortex/ │ ├── engine.py # Three-tier AI engine (FTS5 + C99 PRIME Moment + Ollama + Safety Shield) │ ├── prime_window.py # ctypes binding to libprime.so + O(W+K) context compressor │ ├── c_prime/ # Native C99 PRIME-Moment & Sim kernels (GCC + Cosmopolitan + Fil-C) │ ├── knowledge/cortex_kb.sqlite # 92 KB SQLite FTS5 BM25 database │ └── forge/ │ ├── Modelfile.phi3 # Fast-TTFT (num_ctx=1024) Ollama Modelfile │ ├── cortex_phi3_train.jsonl # Curated instruction-tuning dataset │ ├── hybrid_prime_attention.py # PyTorch causal-mask Transformer surgery for Phi-3 & Qwen2 │ └── train_phi3_qlora.py # 4-bit NF4 QLoRA trainer with GPU hardware guardrail ├── workbench/ │ ├── studio_gui.py # GTK3 + Cairo 60 FPS 7-Lab GUI + K-12 Teacher Report Card │ ├── kid_arcade_view.py # Full-screen Synapse Kids gamified view (EN/FR toggle) │ ├── k12_curricula.py # 12-Year K-12 Traditional Subjects + AI Curriculum generator │ ├── pack_manager.py # 1-Year-at-a-Time Wiper & Persistent Student Transcript Vault │ ├── labs/prime_attention_lab.py # Lab 7: Interactive O(N^2) vs O(W+K) PRIME Moment simulator │ ├── curriculum.py # 35 Lab Examples (5 per lab) & classroom missions │ └── sandbox_runner.py # Subprocess Python sandbox (8s timeout) ├── distro/ │ └── build-iso.sh # Debian 12 Live ISO builder (--lite for 385 MB, --all for 2.5 GB) └── tests/ # Automated unit & integration test suite (32 tests) ```