Ghil Labs is the research-lab brand of WAYPRIME LLC (Branchburg, NJ), run by a single Korean-American real estate investor working alongside a fleet of AI agents. It documents the process of unifying decisions once scattered across separate advisors — real estate, tax, investment, retirement — into a single personal operating system (Asset OS), and releases the tools built along the way as products.
When assets spread across forms, a problem emerges. Real estate people see real estate; tax people see tax; investment people see investments. The chain effects one decision triggers in another domain — say, a sale gain's impact on health insurance subsidies — fall into a blind spot that is nobody's responsibility.
Hold assets in both Korea and the US and the problem doubles. Dual taxation, FBAR and Form 8938 reporting, two entirely different real estate regimes — advisors who see both sides at once are rare, and expensive.
So I started building. A local AI stack and personal knowledge base (RAG) gathered the scattered knowledge in one place; a deterministic calculation engine made every number verifiable; agent orchestration automated the repetitive analysis. That system is Asset OS — and DealDesk and RE Terms Pro are the products born from it.
Everything Ghil Labs releases is a tool I actually use to manage my own assets. Built because I needed it, verified in use, then released. Build for myself first — that is the quality bar.
Built the local AI stack on Ollama, ChromaDB, and Obsidian. Personal documents and decision memos became a searchable knowledge base — the foundation for everything.
Completed the dual track: BGE-M3 embeddings, local ChromaDB plus cloud pgvector. A FastMCP server made personal knowledge retrievable from any AI client.
The GitHub Actions BTC macro monitor began running unattended. The first end-to-end automation — data collection → AI analysis → alerts — phase one of Neutral-Alpha OS.
RE Terms Pro v2 submitted to the Play Store; DealDesk in development. The Asset OS web platform kicked off on Next.js/Supabase/Stripe — from personal tools to public products.
Published Foothold (DealDesk mobile v1.0) on the Apple App Store — rental underwriting and tax-deadline math, fully offline. Android (Google Play) in review.
Systematizing the regime-switching market-neutral framework (iron condors · funding-rate arbitrage · straddles). Monitoring → alert → verification automation advancing.
Tax optimization, LLC structures, insurance, cross-border reporting — plug the leaks before chasing gains.
Real estate underwriting, market-neutral and trend-following trading, portfolio allocation — growth only on verified numbers.
RAG, agent orchestration, monitoring pipelines — repetitive analysis goes to the system; humans focus on judgment.
Generational transfer design — step-up basis, gift and estate strategy, and eventually handing the system itself to the next generation.
Facts and estimates stay separated. Confirmed figures and assumption-based simulations are always labeled apart — [FACT] and [EST.] are different things.
Numbers, not adjectives. "It saves taxes" becomes: federal savings $X, state add-back $Y, net effect $Z. Abstractions can't drive decisions.
Second-order effects included. Not just a transaction's direct tax, but how the AGI increase ripples into other deductions and subsidies — a view only a system can hold.
AI is the verifier, not the decision-maker. The system analyzes, checks, and warns. The final call always belongs to a human.
※ Everything on this site is general information, not tax, legal, or investment advice. Consult licensed professionals (CPA/EA/attorney) before acting.
Questions, press, or partnership inquiries — ghilpocket@gmail.com