Financial & Risk Reality
INVESTMENT FAILUREV
95% of GenAI initiatives fail to impact P&L (MIT 2025).
ALGORITHMIC LOSSV
Zillow Collapse: $881 Million lost to algorithmic over-trust.
OPERATIONAL COLLAPSEV
Volkswagen (CARIAD): €7.5 Billion in cumulative software-unit operating losses, 2022–2024.
DEEPFAKE FRAUDV
Security Breach: $25 Million lost to a single AI deepfake executive impersonation.
PRODUCTIVITY ILLUSIONV
40% of time "saved" by AI is lost fixing low-quality outputs & hallucinations.
LEGAL SURGEV
GenAI Litigation growing by 137% year-over-year (2024).
A SOLO AI STUDIO
VERTEXPACE
Building deterministic, sovereign AI tools for individuals — runs on your hardware, owns its own answers.
Available now · one-time purchase
What You Can Buy Today.
Three products are live and instantly downloadable. No subscription, no cloud, no account required.
Baby AI Lite — $29
Private AI chat app that runs entirely on your Mac via Ollama. Unplug Wi-Fi mid-chat — it keeps working.
Buy Baby AI Lite → NovaGlyph Kit — $79
Compiler + 29 templates. Build your own ~35 KB native AI. Every answer traces back to a fact you wrote.
Buy NovaGlyph Kit → Light Code Cheatsheet — $9
The full 64-concept programming language on one A4 page. Print it, pin it, ship faster.
Buy Cheatsheet → The Ghost in the Grid
Beyond the Black Box.
Cloud AI works—until it doesn't. Outages, hallucinations, API price hikes, and data leaks aren't edge cases; they're built into the business model. VertexPace builds deterministic expert systems for when you need answers, not venture-funded optimism.
GROUNDED. DETERMINISTIC. AUDITABLE.
We have replaced the probabilistic "token" with a documented 75-concept gridConcept GridA 75-concept semantic encoding where every operation maps to a single byte.. VertexPace was born from this technical clarity—choosing to rebuild AI to be as auditable as a spreadsheet.
Current State · May 2026
Where We Stand Now.
Vertexpace is the business. Lumina is the family of sovereign products. Light is the underlying language and 75-concept grid that compiles them. Three layers, one mission — sovereign intelligence that runs on hardware you already own.
Out in the world: Baby AI Lite, the NovaGlyph Kit, and the Light Code Cheatsheet are available to anyone who wants them. LightNet Chat — end-to-end Curve25519 + AES-GCM mesh chat — runs as a real installed app on macOS and iOS.
Running internally: Lumina OS v0.4.0 boots on Raspberry Pi 5, including a bare-metal kernel that runs LuminaVM directly with no Linux beneath it. Lumina STONE — ESP32-S3 firmware with on-device LuminaVM, a 64-LED concept grid, and haptics — builds today (~18 KB code). Lumina Phone — a Swift port of the runtime, 6/6 tests green; macOS app runs, iOS deploy is one Xcode config away.
The reasoning planes (research runtimes): POLYGRID is the architecture that lets the same bytes mean different things in different contexts. PREDICT (predictive coding), DECIDE (7-step decision loop), and LEARN (which reads the others' logs) all run inside the reasoning research runtime with passing tests.
Source for the live products is in private repos for now; the OSS pieces (glyphshrink, NovaHoney) are on GitHub. Anything you see described above either runs today or is honestly labelled as not yet running. No vapor.
Evolutionary Timeline
How We Got Here.
Every product on this site grew out of the one before it. The ecosystem was not designed top-down — it accreted. What started as a 75-concept language inside a sovereign OS has become a family of working products, an encrypted mesh, and hardware you can hold.
Late 2025
Origin
Origin
75-concept language born inside Lumina OS
A deterministic expert system using a 75-concept compressed semantic language. Every operation is a spatial coordinate lookup—no probabilistic tokens.
March 2026
LightNet extends the grid to 64 bytes
The Light compiler matures — Light compiles to native ARM machine code. The core runtime rebuilt around the grid; tests green.
April 2026
Going Full-Time
Going Full-Time
The studio takes shape
Decision to focus on this work full-time. POLYGRID specified — same 64 bytes, different planes. NovaGlyph emerges as the integration coding language above Light.
April 23, 2026
LightNet Chat v0 shipsSHIPPED
Mac ↔ iPhone E2E encrypted chat working. Curve25519 → HKDF → AES-GCM. A purple ? button queries the peer's Baby directly.
April 28, 2026
Baby AI Lite releasedRELEASED
First public release. Same week a significant compiler milestone goes green on Apple ARM64.
April 30, 2026
First hardware image built
Lumina Mesh Router Pi SD-card image built (481 MB). The ecosystem gets its first piece of bootable hardware.
May 2–3, 2026
The Lumina hardware family takes shape
OS v0.4.0 binary
MigratePacket. Bare-metal Pi 5 kernel boots LuminaVM directly. Phone Swift port: 6/6 green.May 5, 2026
NovaGlyph Kit releasedRELEASED
Second public release — v1.3.0. 29 templates that compile to ~35 KB native binaries. The grid moves into other people's hands.
May 10, 2026
Today
Today
Where we are nowNOW
Site rewritten to describe what actually runs. Lumina STONE is the next piece of hardware to land. Cheatsheet queued for release.
Ongoing
Roadmap
Roadmap
What comes next
CREATE plane spec. STONE prototype → production. Lumina Home on ESP32-C3. Same grid, every step.
← Scroll horizontally to follow the timeline →
The AI OS Paradigm
Beyond the Application.
The bet behind Vertexpace is that AI stops being a "feature" inside apps and becomes the KernelKernelThe core part of an operating system that manages hardware resources and provides fundamental services for all other parts of the system. — the layer specialised agents use to coordinate hardware and workflows on behalf of one user. Lumina OS is one solo attempt at that picture. It is not the only one being tried, and we are not claiming to lead an industry — we are building the version we want to use ourselves.
AI-Native Architecture
Built with AI as the primary orchestrator of the system, not a bolted-on interface. The machine logic is derived from intent, not just binary code.
Agent-Centric Interface
Natural language is the primary interface. Our agents reduce the friction between intent and execution, interacting with software autonomously to deliver results.
Unified Lifecycle
Lumina manages the entire intelligence lifecycle—from data preparation to autonomous reflection (Dream StateDream StateAn autonomous reflection cycle where the system synthesizes logs and self-corrects during idle periods.), ensuring a living, honest system.
Absolute Sovereignty
Operating locally by design. No cloud dependency. Your AI OS stays on your hardware, protecting your privacy and digital freedom.
Technical Strength
The Core Advantages.
Our current builds demonstrate that Lumina OS and LightNet are functional realities — a compact proprietary concept language compiling to lightweight native binaries, with a verified core runtime, native compilation pipeline, and a clear roadmap toward a broader concept coverage.
Ultra-Compact Binaries
Programs compile to small native binaries, with new capabilities layered in without inflating binary size. Exact specification is proprietary.
Native Compilation
Source compiles to native machine code via a mature backend. A deterministic inference engine and interactive REPL are proven in internal builds.
Verified Core Operations
String manipulation, arithmetic, and common primitives are implemented and verified as part of the core runtime. Additional capability layers follow the same compact encoding.
Broad Capability Roadmap
The language is designed to scale across file I/O, networking, graphics, cryptography, hardware, and AI/ML primitives — all within the same compact encoding. Most of those rows are still in development.
Real-World Impact
Beyond the Compute.
The VertexPace ecosystem is designed to solve the two greatest crises in modern computing: the soaring energy cost of intelligence and the loss of individual sovereignty. By reducing intent to its geometric essence, we unlock a new future for the user and the world.
Absolute Sovereignty
No subscriptions. No corporate logging. You own the hardware and the intelligence. Your data never leaves your device.
A Green Network
Moving away from power-hungry GPU farms to sustainable micro-compute. Lumina runs reasoning on less energy than a lightbulb.
Traceable answers
Specialists are deterministic lookups over a 64-concept grid. A Baby's answer is always traceable to a fact you wrote — no probabilistic invention inside its own knowledge set.
LightNet (in progress)
Specialist Babies compile to ~35 KB native binaries via the NovaGlyph Kit (live). The peer-to-peer Bluetooth/Wi-Fi mesh that lets devices share answers without a server is on the roadmap — not running today.
Built for People
For the People.
VertexPace is built on honesty. The work moves away from corporate "gatekeeper" subscriptions toward sovereign tools that you own and run yourself. Free open-source releases sit alongside small, self-contained apps. If you want to see everything in one place, the full catalogue covers it.
Source policy
Open where it counts.
The OSS pieces — glyphshrink and NovaHoney — are public on GitHub today. Source for the live paid products (Baby AI Lite, NovaGlyph Kit, the single-file specialist apps, the Cheatsheet) sits in private repos. Same for the in-development runtimes (Lumina OS, the Phone port, the Mesh Router). The plan is to open more over time as the products earn enough to keep going, not to keep things permanently closed.
Core Values
What We Stand For.
Sovereignty
Your data, your keys, your hardware. Every product we build operates locally by default. No cloud dependency. No corporate logging. No backdoors.
Transparency
All performance claims are empirical. All benchmarks are reproducible. We disclose how our systems work, how they were tested, and what they cannot yet do.
Verifiability
Governance without verification is theatre. Every output from a VertexPace system can be audited. Every decision has a traceable, deterministic logic chain.
Honesty
We will not overstate what we have built. We will not hide what is unfinished. The AI industry is saturated with hype. VertexPace is not contributing to it.
Our Products
The Ecosystem.
Local Intelligence
BABY AI LITE
The cornerstone of the Lumina product family. A private, sovereign AI chat app that runs entirely on your Mac via Ollama. No cloud. No account. No telemetry. Unplug your Wi-Fi mid-conversation; the app keeps working. That is the only privacy proof that matters. macOS Apple Silicon and Intel, Windows and Linux builds available.
Read more about Baby AI Lite →Build Your Own AI
NOVAGLYPH KIT
A compiler and 29 ready-to-edit templates that let you build tiny, private reasoning AIs — "Babies" — that run entirely on your Mac. Each one is a knowledge file plus a built-in reasoning engine, compiling to ~35 KB native binaries with zero runtime dependencies. Every answer traces back to a fact you wrote. No hallucinations by construction.
Read more about the NovaGlyph Kit →One-Page Language Reference
LIGHT CODE CHEATSHEET
A whole programming language on one page. Sixty-four words, eight rows of eight — the full base vocabulary of Light Code, the language behind LightNet and the Lumina OS family — printed to A4. A second page covers the POLYGRID extension rows for file I/O, networking, data, functional, system, and crypto. Print it, pin it, ship faster.
Read more about the Cheatsheet →Private preview · AI prompt review
PROMPTCOACH
A reviewer that audits AI interactions for sycophancy, hedging, hallucination, and structural failure — using deterministic failure-logic checks, a multi-dimensional scorer, and a cross-model verification pass to surface where the model agreed without delivering. Currently in private preview; email Zac for access.
Explore PromptCoach →Running internally · v0.4.0 · Not yet for sale
LUMINA OS
A deterministic AI runtime that uses Light Code — a 64-concept programming language — for exact lookup over generative inference. Inside its own knowledge set, a Baby AI's answer is always traceable to a fact you wrote. The agent persists knowledge across sessions and can execute real system commands; the specialist architecture keeps each agent small and auditable. Boots on macOS and Raspberry Pi 5 today, including a bare-metal kernel with no Linux beneath it. Source for Lumina OS is private for now.
Explore Lumina OS →Vision · LightNet Chat works · mesh routing in progress
LIGHTNET
The picture: a mesh of small, topic-specialised Babies (first aid, weather, local plants, recipes) running on personal devices, finding the right specialist for each question without a central model. What runs today: LightNet Chat — a P2P mesh chat app with end-to-end Curve25519 + AES-GCM, working on Zac's own Mac and iPhone. Wider peer routing, specialist discovery, and offline-mesh Baby answers are on the roadmap, not running yet.
Explore LightNet →Visual Symbol Engine · 64-Concept Grid
NOVAGLYPH
The integration coding language built on the same 64-concept grid that powers Light Code. NovaGlyph compresses complex concepts into compact glyph representations, and POLYGRID — its plane architecture — lets the same 64 bytes mean different things in different contexts (PREDICT, DECIDE, LEARN, CREATE). NovaGlyph Studio is a five-tab IDE with a live grid coverage view; NovaGlyph Face is the cross-platform GUI for the Kit.
Explore NovaGlyph →Hardware in development · Not yet for sale
LUMINA STONE
The Lumina family's intended hero hardware device. ESP32-S3 firmware with on-device LuminaVM, a 64-LED concept grid that lights up as the runtime executes, and haptic feedback. Builds today at ~18 KB code. No production hardware unit yet — this is firmware running on a dev board.
See the Lumina hardware family →Working locally · Not yet publicly distributed
LIGHTNET CHAT
A P2P mesh chat app. Auto-discovers peers over Wi-Fi and Bluetooth, runs a per-connection Curve25519 handshake into AES-GCM, and never touches a server. A purple ? button queries the peer's local Baby AI directly. Currently installed on Zac's own Mac and iPhone — tested end-to-end, but not yet available to download.
Explore LightNet Chat →GET IN TOUCH.
Interested in the Prompt Architect, Lumina OS, LightNet, or any other VertexPace project? Send me an email and I'll personally help you get set up with the right build for your hardware.