Christopher Diak designed and built Deep State & Local to leverage the power of deep learning to support real estate developers, investors, architects, and contractors in dealing with their state and local governments.
The problem is deceptively simple: there is no standardized way to access the regulatory requirements that govern what you can build on a given piece of land. Every municipality in the United States maintains its own zoning ordinance, its own building code amendments, its own permit applications, its own fee schedules, and its own approval processes. There are over 19,000 incorporated municipalities, 3,000 counties, and 50 state governments — each with overlapping and sometimes contradictory authority over land use, construction, and development. The information is technically public, but it is scattered across thousands of poorly indexed government websites, buried in PDF ordinances that haven't been reformatted since they were scanned in 2003, and interpreted differently by every planning department staffer you happen to reach on the phone.
For decades, the real estate industry has relied on a patchwork of consultants, land-use attorneys, and institutional knowledge to navigate this landscape. A developer considering a site in an unfamiliar jurisdiction might spend weeks and thousands of dollars in professional fees just to answer basic feasibility questions: What can I build here? What permits do I need? How long will it take? What will it cost?
Deep State & Local takes a fundamentally different approach. Rather than building a centralized database of regulations — an effort that would require an army of researchers and be outdated the moment it was published — the system uses browser-automated AI agents that research regulations the same way a diligent human would: by reading the actual source documents. When you enter an address and describe your project, a coordinator agent analyzes your specific situation and writes six tailored research briefs. Six specialist agents then fan out in parallel, each with its own ChatGPT session, searching government websites, reading zoning ordinances, cross-referencing permit requirements, and extracting the specific provisions that apply to your parcel. The coordinator then synthesizes all findings into a unified report with citations to primary sources.
This architecture — what we call a "coordinator-first agentic loop" — means the system gets smarter with each round of research. The coordinator can identify gaps and contradictions across modules, then write targeted follow-up prompts that push each specialist deeper. A first-round report might identify your zoning district and basic dimensional standards; a second round might cross-reference the flood map, flag a demolition delay ordinance, and calculate your actual FAR from the assessed lot area. The user controls the loop: review the report, decide if it's good enough, or click Refine for another pass.
The technical insight behind this approach is that browser automation provides a uniquely low-cost, low-code way to interface with the regulatory landscape. Rather than building and maintaining API integrations with thousands of government databases — most of which don't have APIs — the system leverages ChatGPT's existing ability to search the web, read PDFs, and reason about complex regulatory language. The Chrome extension simply orchestrates multiple ChatGPT sessions in parallel, each working from a carefully crafted prompt that specifies exactly what to look for and how to return the results. The portal renders everything into a structured, downloadable report with source links.
The result is something that didn't previously exist: a way to get a comprehensive, jurisdiction-specific regulatory feasibility assessment for any property in the United States, in minutes, for the cost of a ChatGPT subscription. Not a template. Not a checklist. An actual research product, with real citations to real ordinances, tailored to what you actually want to build.
Christopher is an AI researcher and holds an MDiv from Harvard and a BA from Middlebury.