A Manifesto
Real estate is a knowledge game disguised as a capital game. But the knowledge required to play it well — the judgment calls, the pattern recognition, the quiet lessons from every deal that nearly worked and every one that didn't — has been locked inside human heads since the beginning. Every great firm has that person who just knows things. Not because they're a genius. Because they've been there long enough to accumulate context.
That person is your most valuable asset. And eventually, they leave.
"The firm that institutionalizes its judgment will own the next decade. Every other firm is just renting talent."
Your best weapon is walking out the door.
The most valuable person at any fund isn't someone with 15 years of experience. It's the person with 15 years of experience at your firm. They remember the deal from 2016 that looked like noise and turned out to be signal. They know which email thread has the VDR link. They know the broker who always inflates T3, and exactly how much. More than anything, they know why the IC passed on that Chicago deal in 2019 — the flood zone, the tenant's parent company's debt, the off-market rumor that never made it into the memo.
When they leave, your firm doesn't just lose a worker. It loses its memory. Decades of nuanced decisions, implicit rules, and hard-fought pattern recognition walk out the door with them. The LPs won't see it on a balance sheet. But they'll feel it in the next deal that goes wrong for a reason you've already been burned by once.
Three things every CRE fund knows are true
01
Your best analysts eventually leave — and take your firm's judgment with them.
02
The knowledge they carry isn't written down anywhere. It's in their head.
03
Every new hire starts from zero. They make the same mistakes you've already paid for.
There's a graveyard of gold sitting in your systems right now.
Every deal you've ever touched left a trail. The OMs, the underwriting models, the IC memos, the redlines, the passed deals with notes explaining exactly why. That data sits dormant in your inbox, your shared drive, your CRM. Nobody touches it. Nobody connects it. Nobody learns from it.
Harnessed correctly, that dormant data doesn't just replace the 15-year vet. It builds something better — a system that carries their knowledge but doesn't sleep, doesn't leave, and gets smarter with every deal you run. One that guides every analyst, every diligence sprint, every IC memo with your firm's specific playbook — not generic best practices from a model trained on someone else's deals.
Generic AI is brilliant. It just doesn't know you.
Claude is remarkable. GPT-4 is remarkable. We use them both, and they power the core reasoning inside REMI. But they share a fundamental limitation: they don't know your firm. They've never seen your IC criteria. They don't know your hold period preferences, your preferred submarkets, or the covenant structures that have burned you before.
Giving a frontier model to an investment team without context is like handing a brilliant surgeon a scalpel and asking them to operate without showing them the chart. Raw intelligence without institutional memory produces generic answers to firm-specific problems. And in this business, generic is expensive.
Generic AI Tools
- Knows real estate broadly
- Has never seen your deal history
- Can't cite your IC's past reasoning
- Starts from scratch every session
- Gives generic answers to specific problems
- No memory of what you've already learned
REMI
- Knows real estate AND your firm specifically
- Has indexed every deal you've touched
- Cites your IC's actual past decisions
- Compounds context over time — forever
- Gives firm-specific answers to firm-specific problems
- Learns from every outcome: closed, passed, and lost
The Context Graph: a moat that compounds every day.
REMI builds what we call a Context Graph — a living, connected map of everything your firm knows. Not a database. Not a search index. A structured model of your firm's intelligence that understands relationships: between deals, between decisions, between assets, between people, and between outcomes.
Every IC memo strengthens it. Every passed deal adds signal. Every email thread, VDR, lease abstract, and broker conversation gets woven into a richer, more precise picture of how your firm thinks — and why it wins. The longer REMI runs, the wider your moat becomes versus every competitor still relying on human memory alone.
Institutional Pattern Matching
Every new deal gets compared against every deal you've ever seen. REMI surfaces analogous comps, flags risks your IC has encountered before, and reconstructs the reasoning that drove your best decisions — not a template, your actual history.
Anomaly Detection
REMI surfaces NOI variances, spending deviations, and operational outliers the moment they diverge from your historical norms — before they become surprises on an LP call.
Explainable Reasoning
No black boxes. Every recommendation traces back to the specific memos, decisions, and data that informed it. Your IC can see exactly how we got there. So can your auditors.
Impact Analysis
Before you change strategy, REMI maps the downstream consequences — which assets are exposed, which automations break, which LP commitments are affected. Risk stops hiding in spreadsheet footnotes.
Why now? Because the infrastructure finally caught up to the ambition.
Building a system like REMI three years ago required custom model training, massive ML teams, and infrastructure costs that only the largest institutions could absorb. Even then, most of those projects failed. Not because the idea was wrong — but because the tools weren't ready.
That changed. Frontier models now reason at a level that wasn't imaginable in 2021, at costs that weren't imaginable in 2023. The intelligence problem is solved. The infrastructure problem is solved. The only unsolved problem — the only remaining moat — is context.
The firm that builds its Context Graph in 2025 compounds that advantage for years. The one that waits until 2027 will spend years trying to catch up to a competitor that's already learned from two additional years of deals — two years of IC decisions, passed opportunities, and hard-won lessons the late mover will never recover.
The REMI Thesis
Intelligence without memory is nearsighted. The best investment firms of the next decade won't be the ones with the smartest people. They'll be the ones with the best institutional memory.
REMI turns your existing data into a compounding intelligence engine — one that gets smarter with every deal, every decision, every cycle. Not a tool you use. A system that learns.