20 Years of Project Controls.
Now Building Software Independently.
Ryan Lucas is the founder and systems architect of SpiritBox, applying project-controls discipline to independent Windows software and local-AI systems.
Hi, I’m Ryan Lucas. I managed multi-million-row datasets, high-stakes cost control, and project scheduling across heavy construction, pipeline, and facilities megaprojects. When software subscriptions started adding up, I began building the local tools I wanted to own. That practical experiment became SpiritBox and a wider independent software portfolio.
From Industrial Megaprojects to Sovereign Code
How twenty years of keeping other people’s multi-million dollar schedules turned into a self-directed AI mastery and a complete commercial software suite.
The 20-Year Infrastructure Crucible
My career was forged in heavy industrial construction, pipeline, and facilities megaprojects. As a Project Controls and Cost Control specialist, my day job was wrestling reality out of chaos: multi-million-row datasets, high-stakes bidding analysis, and writing automated extraction scripts in Python, Power BI, and SQL because spreadsheets alone couldn’t survive the scale.
When you spend two decades watching industrial systems fail, you develop an unbreakable radar for edge cases, resource bottlenecks, and structural flaws.
Choosing a Faster Building Cycle
Large organizations make logical decisions around risk, approvals, and scale. But AI changed the speed at which a small team or solo builder could test an idea, learn from it, and ship something useful. I wanted room to work at that new pace and take responsibility for the result.
Rising subscription costs gave that idea a practical direction: build capable tools that run on user-controlled hardware, charge once where possible, and keep improving them through real use.
Relentless, Hands-On AI Engineering
While others watched AI developments from a distance, I dove directly into the deep end. I dedicated 2,000+ intense, late-night hours learning modern AI from the bare silicon up: benchmarking local models in LM Studio and Unsloth, building diffusion workflows in ComfyUI, training custom voice models with Piper TTS, and exploring agent architectures like Cole Medin’s Archon, Dark Factory, and WAN2GP.
I developed the rare ability to see a new open-source paper or model drop and have a working, battle-tested production pipeline running locally within days.
SpiritBox: From Basic to Pro to Ultimate
The journey culminated in the complete SpiritBox product family:
- August 2025: SpiritBox Basic completed with local typed chat that can run offline after LM Studio and model setup.
- November 2025: SpiritBox Pro launched — introducing built-in Piper TTS so AI chat talks back in natural voices. Sold 2 copies with zero marketing.
- August 25, 2026: SpiritBox Ultimate v1.12 prepared for early access with a nine-service Docker fleet, document Knowledge Base, and patent-pending VRAM Relay.
Mastered Technologies & Hands-On Engineering
Tools, frameworks, platforms, and experimental architectures used across SpiritBox and the wider independent software portfolio.
LLM Engines & Model Optimization
Local Silicon & CloudGenerative Visuals & Hardware Coordination
GPU AccelerationVoice & Audio Neural Processing
Zero-Latency SpeechAutonomous Agents, Knowledge Graphs & Full Stack
Autonomous InfrastructureOpen Sauce 2026: The Alberta to California Road Trip
In the summer of 2026, I drove all the way from Alberta down to California to attend Open Sauce 2026. I engaged directly from the audience floor microphone, met pioneering open-source builders, and turned the entire road trip experience into a dedicated interactive fan tribute site.
See what independent, local-first software looks like in practice.
Explore the engineering portfolio, read the chronological build log, or test the live application simulation directly in your browser.