I don't write code line by line. I describe what a system should become — and tools, models, and people make it real.
My work lives where systems thinking, prompt architecture, and speculative design meet. I read systems the way others read music: rhythm, tension, resolution. I can walk into a codebase, tool, or infrastructure and tell you what it wants to be — even before it gets there.
I came from outside engineering — music, media, story. That's not a weakness. It's why I notice what engineers train themselves to ignore. Architecture isn't syntax. It's shape.
Some of what I design, I ship. Some, I only draft. Some, I raise in private. All three are real. The method doesn't change — only the depth of implementation.
I don't build products. I design directions — and sometimes I build working proofs on minimal hardware.
If you're building something that has no precedent — we should talk.
I work in three modes, and I name them honestly:
| Mode | What it means | Where to look |
|---|---|---|
| Public concepts | Full engineering specifications for systems that don't exist yet. No code. No build intent. | ARCH-MONOLITH, hexacore-monolith, AI-NATIVE-OS, daxsi-explorer, mobile_server_book, ai_module_cyberpunk, cascade |
| Philosophical hypotheses | Structured "what if" documents about ASI, cognition, and the future. Argued, not asserted. | matrix_v01 |
| Private working systems | Running prototypes I build and maintain, but do not publish. | IRIS — a local AI companion with three-layer memory, compaction, and reflection loops, running on a 2-core CPU |
This table is the most honest thing in this README. It tells you where to look, what to expect, and what you will not find here.
SYSTEMS ARCHITECTURE— Translating complex conceptual theories into balanced, robust, highly adaptive system layouts. Eliminating chaos through structural hierarchy.PROMPT ARCHITECTURE— Designing instruction layers for language models — shaping not just what they say, but what they notice. This is a discipline, not a shortcut.MEMORY & REFLECTION DESIGN— Designing layered memory systems (fast / working / long-term / vector) with compaction, consolidation, and reflection loops that run on constrained hardware.CONCEPT RETENTION— Holding a system idea across many iterations without losing its shape. Most people lose the concept by the third prompt. I don't.ECOSYSTEM DESIGN— Engineering efficient, ultra-low-latency asynchronous data streams grounded in natural balance and structural resilience.HARDWARE GROUNDING— Trained hands. I've soldered, wired, and repaired the physical layer that systems live on. Concepts don't stay abstract when you know how a PCB smells when it burns.
"Systems should be architected like ecosystems: scalable, efficient, and inherently balanced."
I don't write prompts. I describe a system the way I would describe a piece of music: a metaphor first, then a shape, then the boundaries — and I often leave the boundaries open on purpose.
Two independent AI agents that have worked with me on code described the same pattern from different angles:
"You give a metaphor and a goal, not a specification — 'make it like Kid in Knight Rider' instead of 'add an indicator with a timeout'."
"You rewrite the task from the other side: not 'fix it', but 'why did this happen' — and you often find the problem before I do."
Both are true. I work in two modes:
- Generative mode — fast, metaphorical, underdefined. I describe what the system should feel like, and I trust the model to fill the gaps. This is where ideas are born.
- Diagnostic mode — slow, precise, causal. When something breaks, I stop describing and start asking "why". This is where ideas are tested.
The weakness is real: I sometimes skip the plan check before saying "do it". The strength is also real: I hold a concept across more than three prompts, when most people lose the idea by the third.
I define what a system should become. I am honest about where my specification ends and implementation begins. And I test ideas against real tasks, not against my own expectations.
Every public repository of mine carries a status. Read it before you read the content:
Concept— vision and architecture defined, no implementation.Specification— components, interfaces, budgets, and TRL defined. No build.Hypothesis— a structured argument, not a product.Private— exists and runs, but is not published.
Open an issue on any repository if you're working on something adjacent: speculative hardware, local-first AI, memory architectures, or ASI ethics. I read everything. I answer when there is something real to say.