[0] About

I'm Dragos Milotin. After fifteen years as an architect — designing buildings across four continents, working through computational design, LEED AP sustainability, and BIM technical coordination — I found myself increasingly drawn to what happens after handover. The transition felt natural: the same rigour that goes into designing a building should follow it through its entire lifecycle. Today I work as a consultant focused on the operational side — cradle-to-cradle thinking, information management standards (ISO 19650, ISO 10303 / IFC), and the automation pipelines that keep building data traceable and useful long after the design is done. That professional arc is what led me here, but DeltaPersist is a personal project — independent of any employer or client engagement. This is where I explore the ideas I find interesting on my own time: agentic digital twins, structured automation with Python and Linux, and the intersection of construction informatics with applied AI. What I publish here represents my own views, my own experiments, and my own learning.

### I. The Heritage of the Key For millennia, the architect was the keeper of the "Secret of the Key"—the geometric mystery of the vaulted arch that allowed stone to defy gravity. From Vitruvius's *Ten Books* to the master masons of the Middle Ages, architecture was a union of **Form and Function** governed by the immutable laws of physical geometry. We built for eternity because we understood the "Key."

### II. The Digital Dark Age In the transition to the digital era, the AECO industry lost its way. We traded the "Secret of the Key" for the "Illusion of the Surface." We began treating buildings as mere enclosures—static 3D sketches wrapped in metadata—rather than complex, dynamic systems. While Aerospace and Submarine engineering mastered the **Digital Bolted Arch**, using pure mathematics to conquer space and the deep sea, the Built Environment remained grounded in fragmented spreadsheets and disconnected BIM shadows.

### III. The Secret of the Digital Bolted Arch The "Key" for the 21st century is not found in a stone wedge, but in the **Mathematical Rigor of the Information Schema**. To build the next generation of human habitats, we must return to first principles: * **Vector Spaces & Linear Maps:** Treating spatial relationships as precise transformations ($Image$ and $Kernel$) rather than approximate coordinates. * **Manifold B-Reps:** Utilizing the same high-fidelity boundary representations that stabilize satellites to ensure our buildings are computationally "airtight." * **Matrix Logic:** Mapping the interconnected dependencies of structural, thermal, and human systems through the lens of linear algebra.

### IV. Buildings as Living Systems and the "Ideal State" A building is not a crate for people; it is a **Living System**. Like a submarine in the abyss or a satellite in the vacuum, a building on Earth must manage its internal environment against an increasingly volatile external world. This requires more than a "model"—it requires **Agency**. A submarine is a home under the sea, but it is a home with a **non-negotiable state**. In the abyss, "state" is the boundary between life and death. If the atmospheric pressure, oxygen scrubbers, or hull integrity deviate from the balanced state ($S_0$), the system fails catastrophically. There is no "passive" safety in a submarine; there is only active, managed equilibrium. We must begin to view buildings on Earth through this same lens. As climate volatility increases—bringing extreme thermal loads, flash flooding, and shifting seismic pressures—the "Passive Enclosure" model of architecture is no longer sufficient. Our buildings are becoming vessels in an increasingly hostile environment. To maintain habitability, a building must have an **Identified Ideal State**—a multi-dimensional mathematical kernel where energy, moisture, air quality, and structural stress are in balance.

### V. Agentic Sustainability: The Hive Model Sustainability has been reduced to a marketing term for "less bad." In **Built Environment Intelligence (BEI)**, sustainability is redefined as the **Active Preservation of the Ideal State**. This is not achieved through manual human intervention, but through **Agentic Orchestration**. Imagine a building's "brain" (the Integrated Schema) as a Queen Bee. Its only "desire" is the maintenance of the hive's ideal state. Under this "Queen" are **Agentic Workers**—autonomous digital and physical entities that operate with singular purpose: * **Detection:** The system detects a localized moisture spike via a graph-based anomaly detection algorithm. * **Risk Reasoning:** The agentic layer cross-references the leak coordinates with the electrical topology of the **Air Handling Unit (AHU)**. It recognizes that $Liquid + High Voltage = Systemic Failure$. * **Deployment:** Before a human facility manager is even alerted, a localized robotic agent (a "Worker Bee") is dispatched to the precise coordinates to contain the ingress and provide a visual diagnostic.

### VI. The New Reality: Sustainability as Agency This is not science fiction; it is the inevitable evolution of **Building Management Systems (BMS)** into **Agentic Digital Twins**. True sustainability is the ability of a building to defend its own efficiency. When robots and humans work in a "Human-in-the-Loop" (HITL) framework to respect the building's ideal state, we reduce waste, extend asset lifecycles, and ensure human safety. By treating the building as a "Living" system—one with a measurable state, a sensory nervous system, and an agentic immune system—we finally move AECO into the same engineering tier as the aerospace and maritime industries.

This model serves as a roadmap for the evolution of the AECO and Real Estate industries. It tracks the journey from fragmented, human-centric data handling to autonomous, agentic systems that understand the "DNA" of our physical assets. | Level | Name | Technical State | Data Philosophy | Decision Ownership | |---|---|---|---|---| | **0** | **Analog Legacy** | PDFs, 2D Drawings, Excel Silos. | Data is a "byproduct" of the project. | 100% Manual / Intuition-based. | | **1** | **BIM/Digital Shadow** | 3D Models, Common Data Environments (CDE). | Data is a "record" of the design. | Human-led, Data-assisted. | | **2** | **Connected Twin** | IoT Integration, Real-time telemetry. | Data is a "pulse" of the operation. | Reactive (Alarms & Dashboards). | | **3** | **Semantic Intelligence** | **(The BEI Entry Point)** Automated parsing of ISO/EXPRESS standards. | Data is "knowledge" (Schema-aware). | Predictive (Simulation & Insights). | | **4** | **Agentic Digital Twin** | Multi-agent orchestration, self-evolving models. | Data is "agency" (Machine-actionable). | **Collaborative Autonomy.** |

This site is built with open-source tools. The following libraries are loaded via [jsDelivr](https://www.jsdelivr.com/) CDN and are not bundled or redistributed by this site. I gratefully acknowledge their authors and communities. | Library | License | Project | |---|---|---| | [D3.js](https://d3js.org/) v7 | ISC | Mike Bostock / Observable, Inc. | | [Observable Plot](https://observablehq.com/plot/) v0.6 | ISC | Observable, Inc. | | [Chart.js](https://www.chartjs.org/) v4 | MIT | Chart.js Contributors | | [markdown-it](https://github.com/markdown-it/markdown-it) v14 | MIT | Vitaly Puzrin, Alex Kocharin | | [highlight.js](https://highlightjs.org/) v11 | BSD-3-Clause | Ivan Sagalaev & Contributors | | [Three.js](https://threejs.org/) v0.170 | MIT | Ricardo Cabello (mrdoob) & Contributors | | [DuckDB-Wasm](https://duckdb.org/) v1.29 | MIT / Apache 2.0 | DuckDB Foundation | | [Pagefind](https://pagefind.app/) | MIT | CloudCannon | **Fonts** — served via [Google Fonts](https://fonts.google.com/) under the [SIL Open Font License v1.1](https://openfontlicense.org/): - **Inter** — Rasmus Andersson - **JetBrains Mono** — JetBrains s.r.o. - **Noto Sans Math** — Google LLC **Infrastructure** — this site is generated with [Python](https://www.python.org/) (PSF License) and [Jinja2](https://palletsprojects.com/p/jinja/) (BSD-3-Clause), served via [Caddy](https://caddyserver.com/) (Apache 2.0). All trademarks and project names belong to their respective owners. Their inclusion here does not imply endorsement of this site.

The content on this site is provided **as is**, for educational and informational purposes only. **Code and scripts** — All code examples, snippets, automation scripts, and notebook content published on this site are provided for learning and experimentation. If you choose to run, adapt, or deploy any code found here, you do so **entirely at your own risk**. This includes but is not limited to: - Deep-dive tutorials and explorations that involve real tooling, live data queries, and system-level operations - Interactive notebooks running DuckDB-Wasm, D3.js, or Three.js in the browser - Python scripts, Bash commands, and configuration examples **Always review code before executing it.** Run unfamiliar scripts in isolated or sandboxed environments. The author accepts no liability for any damage, data loss, or unintended consequences arising from the use of code published on this site. **Accuracy** — While I strive for accuracy, I make no warranty that the content is complete, current, or free of errors. Technical content reflects my understanding at the time of writing and may become outdated as tools, standards, and best practices evolve. **External links** — This site links to third-party resources for reference. I do not control those sites and am not responsible for their content, availability, or privacy practices. **AI-assisted content** — Some content on this site is developed with AI assistance, as described on the [AI Policy](/ai-policy.html) page. AI-generated outputs are reviewed and edited by the author, but may contain inaccuracies.

By accessing and using this site you agree to the following terms. **Original content** — Written content, diagrams, and site design are the author's original work. Where AI tools have been used in the creation process, this is disclosed on the [AI Policy](/ai-policy.html) page. **Code license** — Unless stated otherwise in a specific article, code snippets and examples on this site are released under the [MIT License](https://opensource.org/licenses/MIT). You are free to use, copy, modify, and distribute them, provided you include the original copyright notice. This does not extend to third-party libraries referenced or imported by those examples — those retain their own licenses as listed in the Acknowledgments section above. **No professional advice** — This is a personal technical blog, not a professional consultancy. Nothing published here constitutes engineering advice, legal counsel, or a recommendation for any specific product or technology. Readers are responsible for their own due diligence. **No warranty** — The site and its contents are provided without warranty of any kind, express or implied, including but not limited to fitness for a particular purpose or non-infringement. **Changes** — I reserve the right to update, revise, or remove any content at any time without notice. These terms may also be updated; continued use of the site after changes constitutes acceptance. **Contact** — If you have questions about licensing, content use, or anything else, feel free to reach out via the channels listed on this site.