Frequently Asked Questions
Common questions about Neural Commander, the self-sovereign AI intelligence harness.
Product
About Neural Commander
What is Neural Commander?
Neural Commander is a self-sovereign AI intelligence harness. A Go daemon that runs on your hardware, maintains persistent intelligence across seven domain-specific commanders, and provides remote access via Telegram and a 14-panel terminal interface.
The product is structured on four axes. Commanders (WHO) are the seven personas you talk to. Intelligences (HOW) are nine structured reasoning protocols that deepen how each commander thinks. Adjuvants (WHAT) are sovereign external tools whose data flows into NC. Knowledge Packs (WITH WHAT DEPTH) are domain-specific stack configurations from a marketplace. Each axis compounds the value of the others.
230,000+ lines of production Go. 1,500+ tests. 95 packages. Single binary, zero runtime dependencies.
How is this different from Claude or ChatGPT with memory?
They remember conversations within their own platform. NC remembers everything across all of them and stores it on your hardware in an encrypted vault you control.
NC works across tools (Claude, Copilot, ChatGPT, Cursor), not inside one. NC stores data locally, not on vendor servers. NC has seven domain-specific commanders, not a single general assistant. NC routes across five AI providers including zero-dependency local inference. Switch providers, nothing changes. Your intelligence persists.
Who are the competitors?
No existing product combines seven-commander governance, nine structured reasoning protocols, sovereign tool integration, encrypted vault, email intelligence, document attestation, and remote Telegram access in a single local-first binary.
Adjacent products occupy different categories. OpenClaw is a message router with a runtime skill marketplace (and a significant security track record including 1,184+ malicious skills in the ClawHavoc incident). Cursor and Windsurf are IDE-specific. Copilot is single-session and cloud-only. Notion AI and Linear are cloud knowledge tools.
What makes NC defensible?
The convergence architecture. Local-first daemon, multi-commander governance, intelligence framework, adjuvant ecosystem, entity-project hierarchy, satellite database pattern, compiled binary. Each layer reinforces the others. Cloud vendors cannot replicate this without cannibalising their own revenue model. Productivity suites cannot replicate this because they own one silo each.
Every Adjuvant tool connected makes every commander smarter. Every month of usage deepens persistent memory. Switching costs grow with time. The architecture compounds.
What if the big AI companies build this?
Their business models require user data on their servers. A truly local-first product would cannibalise their cloud infrastructure investment. NC benefits from their model improvements (better models make NC's Provider Gateway more powerful) without competing for the same revenue stream. NC is infrastructure underneath their models, not a replacement for them.
The Seven Commanders
Domain-specific intelligence
Why seven commanders?
Each commander owns one domain question a founder needs answered every day.
Hakim, Health Commander: "Am I well enough to do what I'm trying to do?"
Claudette, Strategy Commander: "What should I do right now and why?"
Farida, Finances Commander: "Can I afford this?"
Kai, Projects Commander: "Where are my projects at?"
Raya, Relationships Commander: "Who needs my attention?"
Zara, Wealth Commander: "What are my assets worth?"
Diderot, Knowledge Commander: "What do I know and what should I learn?"
Seven is the natural boundary of a founder's operational life. The structure is locked.
How do the commanders work together?
In the Commander Room, all seven debate your priorities in real-time. Each contributes domain expertise. You sit as Chairman with veto and casting vote. Farida might flag a cash risk. Kai proposes a timeline. Hakim challenges whether the timeline is sustainable for your health. Claudette synthesises and proposes a motion. All commanders vote.
Not agents that replace you. A council that serves you.
What are Intelligences?
An Intelligence is a structured reasoning protocol that loads into one or more commanders' prompt chains. Nine Intelligences: six Core (Email, Session, Governance, Knowledge, Prompt, Build) and three Specialist (Design, Luminary, Practice). They determine how deeply each commander thinks about its domain. All nine are compiled into the binary. No runtime plugin loading.
What are Adjuvants?
An Adjuvant is a sovereign external tool whose structured local data flows into NC. NC reads Adjuvant data from the local filesystem. NC does not write to Adjuvants, depend on them, or require any Adjuvant to function. Each connected Adjuvant makes every commander smarter.
Adjuvants must pass the Sovereignty Filter: open source, local-first, zero telemetry, structured data format, actively maintained. Launch Adjuvants: Obsidian and Super Productivity. Planned: Actual Budget, hledger, Paperless-ngx, Monica PRM.
Connected, not surrendered.
Technology
Under the hood
Why is NC built in Go?
Single binary distribution. No runtime dependencies. Cross-compiles to Linux, macOS, and Windows. For a tool that runs on your hardware with zero setup, single-binary distribution matters.
What is the security architecture?
ChaCha20-Poly1305 encryption with Argon2id key derivation for the vault. Ed25519 signing with SHA-256 hash chain for document attestation. PII scanning with privacy-gated provider routing. Compile-time skill architecture with no runtime plugin marketplace.
Zero telemetry. No phone-home, no analytics, no tracking. Verifiable by network traffic analysis.
What is the Provider Gateway?
Five-provider routing chain: PicoLM 1.1B (local, zero dependency) through Ollama (local, larger models) through Groq (cloud, fast) through Anthropic Claude (cloud, capable) through OpenRouter (cloud, model diversity). Privacy-aware routing ensures sensitive data stays local. Complexity-aware routing matches the right model to the task. Manual override always available.
What is the tech stack?
Go single binary. SQLite with FTS5 for full-text search. Bubble Tea terminal interface. Obsidian-format Markdown vault. Telegram bot for remote access. Tauri v2 with SolidJS for the desktop GUI (coming Q3 2026). No Electron. No Docker. No cloud dependency.
Pricing
Community and Pro
What does the Community Edition include?
All seven commanders at base features. Five projects, two email accounts, two vaults, two entities, four providers (PicoLM, Ollama, Anthropic, Groq). Free forever. Proprietary freeware.
What does Pro add?
Full Intelligence protocol depth. Fifteen projects, fifteen email accounts, fifteen vaults, five entities, all five providers with intelligent auto-routing. Priority support.
Pro Standard: $19/month (USD). Founders' Rate: $12/month for the first 100 users, locked at that price forever.
What is the BYOK model?
Bring Your Own Keys. You provide your own AI inference. PicoLM is built in at zero cost. Ollama runs local models. Cloud providers (Groq, Anthropic, OpenRouter) use your own API keys. NC never sees your keys, never routes through NC servers, never costs NC per inference.
Why is CE free? How does NC make money?
Community Edition demonstrates the value of sovereign AI infrastructure. Pro unlocks full Intelligence depth for operators who need it. The upgrade path is natural: use CE, experience the commanders, want deeper reasoning. That is Pro.
CE is proprietary freeware, not open source. The sovereignty promise is delivered through architecture (zero telemetry, local-first, verifiable), not source code access.
Methodology
How NC was built
How was NC built?
One founder. Five parallel Claude Code sessions. Nine months. 230,000+ lines of production Go, 60,400+ lines of test code, 91,200+ lines of project documentation across 268 documents. The founder architects the system, designs specifications, writes structured orchestration prompts, reviews gate reports, and verifies output by running the binary. Claude Code writes all Go code across five parallel sessions with specific roles: backend, UI, integration, QA, and adversarial review.
What is the build harness?
Five AI sessions ship code faster than one human can review. The harness removes that ceiling. Every commit hits automated CI across three platforms. Then Claude reviews every pull request against the design specs. Only then does it merge. The quality floor is mechanical. It does not depend on which session wrote the code.
230,000 lines of Go. Every commit earned.
Can a team scale this methodology?
That is the point. The three-layer orchestration protocol, 268 documents of project knowledge, and CI harness are all designed to be inherited by a team. A new team member, human or AI, can join any construct and be productive within one session. The methodology is infrastructure, not tribal knowledge.
The Category
Sovereign AI
What category does NC define?
Sovereign AI. Not privacy marketing. Architecturally sovereign. Your hardware, your data, your keys, your veto. Cloud vendors cannot build here because their revenue requires your data. Productivity suites cannot build here because they own one silo each. Only a platform built to run on your hardware from day one can converge all of it.
What is the vision?
Every knowledge worker will have AI specialists that know them, run on their hardware, and move with them for life. NC is nine months into building that platform. The four-axis architecture is designed to compound: more connected tools make every commander smarter, more Intelligence protocols deepen every domain, and Knowledge Packs make the system immediately useful in new verticals.
Founder
Bradley C Hughes
Who is the founder?
Bradley C Hughes, 55, Sydney, Australia. 44 years in technology. First computer 1982, Microbee 16KB. Started B.Inf.Tech at UTS in 1989. First webmaster at UTS Library in 1994. Founded Business Marketing Systems (B2B digital agency, since 2009). Co-founded Blockchain Sydney (5,500+ members, approximately 200 events). Winner, UTS Startups AI Pitch Competition (November 2025).
What is the origin story?
Neural Commander was conceived as Project Nexus in mid-2025. The founder had been building with AI tools for years and recognised that every AI assistant treats each session as disposable. The architectural thinking started before a single line of Go was written.
In October 2025, a Windows system crash destroyed about ten hours of AI session context while the founder was building a cryptocurrency trading system called Dexinator. The AI forgot critical requirements. That wasn't the moment NC was born. It was the moment the thesis was proven right. Six months later, NC is 230,000+ lines of production Go.
See the architecture in action
Seven commanders. Four memory layers. One daemon on your hardware.