> My personal playground for context engineering experiments. Building privacy-preserving AI systems, testing ideas, and sharing the journey with fellow builders.
Most AI systems are built backwards. Engineers optimize for prompt engineering, fine-tuning, and model selection—treating context as an afterthought.
But the context is the work. The context is the constraint. The context is the leverage point.
At IDDQDX, we architect systems that understand their own epistemic boundaries. We build AI that knows what it knows, what it doesn't know, and why. This is not philosophy—it's engineering.
Privacy-preserving. Self-sovereign. Aligned. These aren't buzzwords. They're requirements for systems that actually scale.
[AI_POWERED_DESIGN]
AI-powered interior design platform automating the capture-to-render pipeline. Video consultations → AI transcription → Design notes → GPU-accelerated 3D models.
[GENETIC_LLM_OPTIMIZATION]
A research project exploring evolutionary algorithms for optimizing LLM behavior. Helixight treats prompt engineering as a search problem—finding the optimal context through genetic selection.
An experimental trading system that demonstrates context engineering in financial markets. The agent learns market microstructure and adapts its strategy based on contextual signals.
Why the future of AI isn't about bigger models or better prompts. It's about understanding and architecting the context in which AI operates.
A technical journey through building a hobby project that evolved into a production-grade genetic algorithm for LLM optimization. How constraint breeds innovation.
I'm Zygmunt Dyras, Vice President of Engineering at WP Engine, leading distributed engineering teams building infrastructure for the open web. My career has been defined by one consistent thread: building systems that scale, remain aligned with their purpose, and respect the constraints they operate within.
Over the past decade, I've led teams through infrastructure transformations, architectural decisions that shaped product strategy, and the kind of technical leadership that requires understanding not just the code, but the context in which it operates. At WP Engine, I've had the privilege of working on problems that matter—making open-source technology accessible, performant, and secure at scale.
But there's a gap. Most AI systems are built with the opposite philosophy: unconstrained, context-agnostic, and treated as black boxes. IDDQDX exists to explore what happens when we flip that script. What if we built AI systems that understood their own epistemic boundaries? What if we treated context not as a limitation, but as the primary design surface?
IDDQDX is my research vehicle for exploring this question. Each project is a hypothesis about how AI systems should be architected: GetEnclave for privacy-preserving systems, Helixight for evolutionary optimization, AI Crypto Trader for context-aware decision-making. These aren't just hobby projects—they're experiments in a different way of thinking about AI engineering.
I believe the next wave of AI won't be about model size or parameter count. It'll be about context depth. About understanding the epistemic boundaries of what we're building. About systems that know what they don't know, and why. That's the work. That's IDDQDX.
If you're a builder curious about context engineering, privacy-preserving AI, and experimenting with LLMs—drop your email. I share what I'm building, what's working, and what's not. No polish, just progress.
▓ No spam. No marketing fluff. Just experiments and insights from the playground.