System Online · Self-Healing · 24/7 Autonomous

One Machine.
An Entire Autonomous Operation.

A single Mac Mini running 141+ AI skills, multi-model orchestration, autonomous software pipelines, financial analysis engines, and a self-healing infrastructure stack — replacing over $50,000/year in SaaS tools and human coordination.

0AI Skills
0Managed Services
0Knowledge Pages
0Native Tools
0AI Models
0Monthly Cost
Explore the full stack
At a Glance

Live System Profile

The numbers behind a machine that thinks, codes, trades, researches, and manages itself.

🍎Apple M4Neural Engine on-chip
141Composable skills
🧠8,776Embedded knowledge chunks
⚙️20LaunchDaemon services
8Scheduled cron jobs
📝130+Memory & log files
📚6+Obsidian vaults
🔬300+Trading research docs
🛡️4-LayerSecurity defense system
💰$65/moTotal operating cost
📈99%+Uptime (auto-recovery)
🎬3Media generation engines
Capability Stack

Nine Pillars of Autonomous Operations

Each pillar is a self-contained capability domain. Together, they form a system that operates, decides, and creates — independently.

🧠

Multi-Model AI Cognition

4 models · intelligent routing · $65/mo total
4 AI models $0 local inference ~$65/mo total spend Diversity-enforced routing

GLM-5.2 Primary

Default model for all operations — reasoning, conversation, tool orchestration, daily decisions.

  • Selected for strong multi-step reasoning at low cost
  • Handles sub-agent spawning for parallel task execution
  • Drives code pipeline architecture stage

Claude Sonnet 4.6 Heavy Lift

Premium model for complex analysis, code review, whitepaper debate, and fallback routing.

  • Independent perspective in multi-model debates
  • Code review in the specialist pipeline
  • Complex multi-step reasoning fallback

Ollama Qwen 3.5 Local

On-device inference engine — zero-cost, zero-latency, zero data egress.

  • Context compaction (summarizing long conversations)
  • Safeguard checks and background tasks
  • Never used for trading, math, or knowledge-critical tasks

GLM-4.7 Flash Compaction

Lightweight specialist for context window management and embedding generation.

  • Summarizes sessions to fit token limits
  • 30K token reserve floor for safety
  • Micro-cents per operation
💻

Autonomous Software Engineering

4-stage pipeline · 28+ dev skills · $0.18/run
4-stage specialist pipeline 28+ development skills ~$0.18 per pipeline run Model diversity enforced

🔧 Code Pipeline

Multi-stage specialist chain where each stage uses a different model family for cognitive diversity:

  • Architect (GLM-5.2) — system design, spec generation, tech selection
  • Coder (Sonnet 4.6) — implementation, writes production code
  • Reviewer (GLM-5.2) — quality gate, can reject & loop back (max 3 iterations)
  • Tester (Ollama) — runs test suites, all must pass green
  • Escalation: Architect swaps to Sonnet 4.6 for auth, payments, distributed systems

🏗️ Sprint Lifecycle (gstack)

28+ composable engineering sub-skills loaded on demand:

  • Spec-driven development & incremental implementation
  • Design review, engineering review, CEO review gates
  • QA automation, security audit, SQL safety checks
  • Canary deployment monitoring & land-and-deploy strategies
  • Engineering retrospectives with commit analysis
  • Spike/prototyping validation framework

🐛 Debugging & Diagnostics

  • Node.js inspector debugger integration
  • Python debugpy live debugging
  • Session log analysis & error classification
  • Automated timeout, rate-limit, and auth error detection

📦 Project Management Integration

  • Paperclip issue tracking (create → assign → execute → ship)
  • GitHub issue sync with messaging channels
  • Automated "Done When" criteria verification
  • Multi-company portfolio management
📊

Financial Analysis & Trading

7 modules · SEC EDGAR · quant models
7 finance modules 300+ research docs 4 quant models SEC EDGAR integrated

📋 Financial Statements Engine

Full SEC filing extraction and 3-statement modeling.

  • 10-K, 10-Q, 8-K parsing from SEC EDGAR
  • 3-statement Excel models (Base / Upside / Downside)
  • Historical financial data via yfinance API
  • Automated ratio analysis & trend detection

🎯 Earnings & Valuation

  • Earnings analysis with bullish/bearish signal scoring
  • DCF, comparables, and peer benchmarking
  • Investment pitch one-pagers with target prices
  • Peer comparison engine (MAG7, SEMIS, Cloud, Banks)

📈 Nexdex Trading Intelligence

Quantitative trading research and model development.

  • Markov chain price prediction
  • Black-Scholes options pricing
  • Kelly Criterion position sizing
  • Bayesian probability modeling
  • Top edges: weather markets (94% WR), stat arb, oracle latency arb
  • 300+ research files across 64+ strategy documents

🔍 Sector & Market Research

  • Industry competitive analysis & due diligence
  • Target screening & acquisition pipeline
  • Real-time price data feeds
  • Python finance stack: pandas, numpy, openpyxl, xlsxwriter
🧠

Knowledge & Memory Architecture

1,677 pages · 8,776 chunks · hybrid search
1,677 knowledge pages 8,776 embedded chunks 6+ Obsidian vaults 130+ memory files

🔗 gBrain Knowledge Graph

Semantic knowledge engine combining vector search with full-text retrieval.

  • 1,677 pages indexed across 6+ vaults
  • 8,776 embedded chunks (8,730 actively embedded)
  • Hybrid vector + full-text search via LanceDB
  • nomic-embed-text for on-device embeddings
  • 88 cross-references, 54 tags, 17 content types
  • Auto-re-embeds stale content every 6 hours

📓 Document Knowledge Base

6 Obsidian vaults organized by venture and domain:

  • Fulcrum AI — automation agency docs
  • Nexdex — trading research & models
  • Vibestreet — marketplace architecture
  • Inclination — AI shopping assistant
  • Infrastructure — system docs
  • Strategy — business strategy & planning

💾 Three-Tier Memory System

Different persistence guarantees for different needs:

  • Working: Current session context (60K token window)
  • Daily logs: Raw chronological events, append-only
  • Long-term: Curated, deduplicated permanent knowledge

🔄 State Persistence Layer (CPL)

Bridges session memory and permanent storage — no context lost across restarts.

  • Living entity ontology (people, companies, projects)
  • Active thread tracker (WIP tasks & decisions)
  • Bidirectional cross-reference index
  • Pre-compaction session snapshots
🎨

Creative & Media Generation

3 engines · image · video · music
3 media engines Multi-provider routing Up to 4K resolution 20 images per analysis

🖼️ Image Generation

Multi-provider routing to OpenAI GPT-Image, Fal.ai Flux, Google, and more.

  • Transparent backgrounds (PNG/WebP)
  • 1-4 images per call, aspect ratios 1:1 through 8:1
  • Reference images for style transfer & editing (up to 10)
  • Resolutions up to 4K, quality control (low/medium/high)
  • Provider-specific: OpenAI moderation, Fal creativity levels

🎬 Video Generation

Text-to-video, image-to-video, and video-to-video with multi-modal references.

  • First frame, last frame, and reference image support
  • Up to 9 reference images, 4 reference videos, 3 audio refs
  • Aspect ratios: 1:1, 16:9, 9:16, adaptive
  • Resolutions: 360P through 4K
  • Audio-conditioned generation (reference music/audio)
  • Provider options: seeds, watermark control, duration

🎵 Music Generation

Multi-provider audio creation including Google Lyria.

  • Genre, mood, tempo, instrument, purpose prompts
  • Sung lyrics support or instrumental-only mode
  • Reference images for visual mood injection
  • MP3/WAV output, configurable duration

👁️ Vision & Image Analysis

Configured vision model for inspection and understanding.

  • Up to 20 images per analysis call
  • Custom inspection prompts
  • 20MB max per image
  • Cross-modal review capabilities
🛡️

Security & Defense Systems

4-layer defense · auto-recovery · thermal guard
4-layer defense 15-min health checks Auto-restart all services Thermal protection

🔒 Pre-Tool-Use Defense Layer

Blocks dangerous operations before they execute.

  • Secret detection: Anthropic keys (sk-ant-), OpenAI keys (sk-), AWS keys (AKIA), JWT tokens, GitHub tokens (ghp_), Discord tokens
  • Risky command blocking: rm -rf /, rm -rf ~, > /dev/sda, dd if=/dev/zero, chmod -R 777 /, curl|bash, wget|bash
  • SQL injection prevention: DROP TABLE, DROP DATABASE, DELETE FROM...WHERE 1
  • Returns block decision JSON — execution never starts

📊 Post-Tool-Use Observability

Full audit trail after every tool execution.

  • JSONL logging to tool-usage.jsonl
  • Latency tracking (start/end timestamps → ms)
  • Error classification: timeout, rate_limit, connection, permission, not_found, auth
  • Auto-format Python files after writes
  • Post-execution secret scanning

🔍 Pre-Commit Secret Scanner

18-pattern scanner prevents secrets from entering version control.

  • API key patterns for all major providers
  • Private key detection (RSA, EC, PGP)
  • Database connection string detection
  • Generic high-entropy string detection

🌡️ System Health & Thermal Defense

Continuous monitoring with automatic intervention.

  • Health check script runs every 15 minutes
  • Auto-restarts crashed services via LaunchDaemon management
  • CPU thermal monitoring & throttling defense
  • Alerts sent to messaging channels on repeated failures
  • 20 managed LaunchDaemons — all auto-restart capable
📡

Communication & Human Interface

Multi-channel · voice · co-worker access
3+ comms channels Voice wake word Scoped co-worker access Platform-aware formatting

💬 Discord Multi-Channel Hub

Venture-scoped channels for focused operational context.

  • Dedicated channels: Infrastructure, Strategy, Brand, Trading, Agency, Marketplace
  • Real-time agent monitoring & session management
  • Project management bridge — issue status syncs to chat
  • Thread-bound sub-agent spawning for parallel work
  • Rich components: buttons, selects, forms, polls, reactions

📱 WhatsApp Direct Line

Time-sensitive alerts and briefings to leadership.

  • Heartbeat alerts for urgent items
  • Quiet hours enforcement (23:00–08:00)
  • Platform-aware formatting (no markdown tables)
  • Daily briefing delivery

🌐 Web Control Panel

Browser-based administration and monitoring.

  • Session listing & history inspection
  • Agent configuration & model overrides
  • Tool testing & approval management
  • Gateway status & health monitoring
  • Scheduled job management

🎙️ Voice Interface

Wake-word activated voice assistant.

  • Wake word: "Apex" — hands-free interaction
  • Text-to-speech output (voice selection configurable)
  • Seamless integration with all agent capabilities

👥 Co-Worker Access System

Scoped permissions for team collaboration.

  • Brenda — sandboxed workspace, image generation access, strict permission model
  • Alizain — full technical CRUD, deploy/config changes require Chairman approval
  • Role-based access control with workspace isolation
⚙️

Automation & Operations

8 cron jobs · 20 scripts · browser automation
20 automation scripts 8 scheduled jobs Stealth browser 3-gen backups

⏰ Scheduled Task Engine

Cron-driven automation for self-maintaining operations.

  • System backup: weekly full backup to external drive
  • Git backup: daily version-controlled state backup
  • Health check: every 15 minutes — verifies all services
  • Knowledge distiller: weekly structural analysis
  • Knowledge graph sync: every 6 hours — re-imports & re-embeds
  • Context snapshots: as needed before compaction

🦊 Stealth Browser Automation

Anti-detection web automation with full session control.

  • Anti-fingerprint patches — interacts as a real user
  • Cookie injection for authenticated sessions
  • Screenshot capture, DOM interaction, form filling
  • Use cases: job scraping, social media posting, competitive research
  • Bypasses API limits by operating through the browser

📎 Background Agent

Dedicated agent for project management and issue processing.

  • Issue status synchronization
  • Priority-based sorting & automated triage
  • Runs on local model — zero incremental cost
  • Offloads routine management from primary agent

💾 Backup & Recovery System

3-generation rolling backup with multiple storage tiers.

  • Weekly full system backup to external drive
  • Daily git version-controlled state backup
  • 3-generation rolling rotation for disaster recovery
  • Pre-compaction session snapshots
  • 20 automation scripts for operational tasks
🔩

Infrastructure Core

Apple M4 · local-first · 25-layer architecture
Apple M4 Neural Engine Local-first architecture 25-layer emergE framework Loopback-only binding

🍎 Hardware Platform

Apple Silicon Mac Mini — energy-efficient, silent, neural engine on-chip.

  • Apple M4 chip with integrated Neural Engine
  • Unified memory architecture for CPU/GPU shared access
  • ~15W idle power draw — runs 24/7 cost-effectively
  • No moving parts beyond fan — near-zero failure rate

🏠 Local-First Architecture

Core operations have zero external dependencies.

  • Gateway binds to loopback only (127.0.0.1)
  • All AI inference can run on-device (Ollama)
  • Knowledge graph runs locally (gbrain + LanceDB)
  • No cloud dependency for core operations
  • External APIs only for premium models & web research

🔌 Service Stack

20 managed LaunchDaemon services — all auto-restart capable.

  • OpenClaw Gateway — agent orchestration layer
  • Ollama — local LLM inference engine
  • gbrain — knowledge graph & semantic search
  • Apex Voice Bot — wake-word voice assistant
  • Finance Python environment (pandas, numpy, yfinance)
  • Git version control with hooks
  • Paperclip project management server

🏗️ emergE Compute Framework

25-layer architecture specification for edge AI compute nodes.

  • Multi-tier hardware support: Hub Node → Standard → Lite → Edge → Micro
  • 3 open protocols: NATS (event bus), MCP (context), LDPM (device management)
  • 5 intelligence pillars: Core Runtime, Deployment, Intelligence, Security, Operations
  • Designed for multi-node federation and scale-out
The Real Comparison

Traditional IT Stack vs. Mini Node

What it actually replaces — in hard numbers.

🏢 Traditional Setup

  • Developer(s) $120K+/yr
  • Financial analyst $80K+/yr
  • SaaS subscriptions $1,200/mo
  • Cloud infrastructure $500/mo
  • Project management tools $200/mo
  • Research & analysis tools $300/mo
  • Content creation tools $250/mo
  • Backup & monitoring $150/mo

⚡ Mini Node

  • Hardware (one-time) $800
  • AI model costs $65/mo
  • Electricity $8/mo
  • SaaS subscriptions $0
  • Cloud infrastructure $0
  • Developer time $0
  • Analyst time $0
  • Content creation $0
Annual savings: $250,000+
Operating Economics

Where the Money Goes

Total monthly operating cost breakdown — the entire system runs for less than a single SaaS subscription.

Traditional IT salary + SaaS
Baseline
$22,000/mo
Mini Node total
$73/mo
├ AI models (cloud)
$65
$65/mo
├ Local inference
$0
$0/mo
└ Electricity (24/7)
$8
$8/mo
In Practice

A Day in the Life

How a single request flows through the system — end to end, autonomously.

1

Request Received

Chairman sends a message via Discord, WhatsApp, or voice. The Gateway routes it to the primary AI model with full context loaded — memory, entity state, active threads.

2

Skill Resolution & Planning

Agent scans 141 skills. If one matches, it loads automatically. If the task is complex, a goal is created and a multi-step plan is generated. Sub-agents may be spawned for parallel work.

3

Execution

Tools fire in sequence or parallel — web research, code execution, file operations, API calls, browser automation. The security layer checks every tool call pre-execution and logs every result post-execution.

4

Knowledge Capture

Every decision, output, and insight is captured. Memory is updated. The knowledge graph re-embeds. Daily logs are written. State persistence layer tracks all entity changes.

5

Delivery & Monitoring

Response is delivered via the originating channel with platform-appropriate formatting. Background monitoring continues — health checks every 15 minutes, backups on schedule, auto-recovery if anything fails.