A single source of truth that captures the records, watches the portfolio, remembers the decisions, and runs the operations — so a portfolio built over a lifetime can be run, and handed down, without living in one person’s head.
Why it exists
Local real estate is often a family’s primary multi-generational asset — yet it’s operated from memory, a shoebox of documents, and a phone full of contractor numbers. Existing software was never built for this.
Which plumber to call, why a unit’s rent is set where it is, where the insurance lives, what the TMK is — it’s all tacit. When the operator steps back, it evaporates.
TMK parcels, GET on rents, HRS 235 rules, island-by-island differences. Generic apps have no concept of any of it, so it’s tracked by hand or not at all.
A spreadsheet for rents, QuickBooks for books, email for tenants, a drawer for leases, the CPA for taxes. Nothing reconciles and nothing is proactive.
The next generation — or an advisor stepping in — inherits assets but not the operating knowledge. Succession is an afterthought, not a system.
The model
Each layer earns its place: capture the truth, watch it continuously, remember every decision, then execute the work — each one feeding the next.
Layer 1 · Capture
The Hawaiʻi-native foundation every other layer reads from.
Layer 2 · Watch
Continuously scans the records and surfaces what needs attention.
Layer 3 · Remember
Captures the why so the portfolio outlives the operator.
Layer 4 · Execute
Turns decisions into tracked, finished work.
The flow
Like a dynamic agent workflow, work fans out, gets scrutinized, converges on a human decision, then executes — and every pass feeds the next. Signals enter at the top; an updated, better-documented portfolio falls out the bottom and loops back. Nothing goes stale because the loop never stops.
What’s built
Every module below is live in the app today, backed by a four-migration data model (core → advisory → knowledge → operations). The colored dot shows which layer each surface belongs to.
Market fit
Hale OS serves the three roles around a family portfolio — and the moment of hand-off between them. The data model bakes these roles in from day one.
Operator
The owner managing rents, repairs, and renewals hands-on. Gets one proactive cockpit instead of five disconnected tools — and finally writes down the “why” without extra effort.
Family member
The spouse or next generation who will one day take over. Gets visibility today and a documented, ready-to-run portfolio tomorrow — not a shoebox and a phone number.
Advisor
The CPA, attorney, or property manager who needs scoped, accurate access. Reads the same source of truth the operator lives in — no year-end document scramble.
Enterprise PM suites (AppFolio, Buildium, Yardi) are built and priced for management companies with hundreds of doors and per-unit minimums.
Landlord fintech (Stessa, Baselane) stops at bookkeeping and banking — no operations, no knowledge, no succession.
Spreadsheets + QuickBooks capture numbers but never relationships, decisions, or continuity.
None are Hawaiʻi-native. None are built to transfer the portfolio to the next generation.
Land: import the existing spreadsheet → an instant, Hawaiʻi-aware system of record. Zero-to-value in one sitting.
Stick: the advisory layer immediately starts watching — expiring leases, overdue rent, expense anomalies, tax deadlines.
Expand: capture playbooks, decisions, and a succession plan — the things that make switching cost real.
Why now: a generational wealth transfer is underway across the islands, and AI finally makes capturing the knowledge in someone’s head cheap.
Build notes
This was the first thing I treated as released rather than personal: deployed, one branch and one pull request per change, even though I was the only user, because I wanted the habits before anything real depended on them.
The admin approval gate on sign-in exists because the new-user questionnaire got confusing enough that I did not want anyone reaching it unattended.
Two things now run against this repo without me: a model policy check that periodically looks for newer models, which is how a stale Sonnet 4 fallback in the extraction and chat path got caught, and a security scout that sweeps the repo for vulnerabilities on a schedule.
The accounting suite. I ran gap analyses against Buildium, the software this replaces, and full accounting is the piece that still needs me at the wheel, with scenarios and QA runs I drive myself.