CRE operators, AI builders & research teams—let’s connect.
I spent 30 years in industrial commercial real estate. Started with a desk, a phone & 200 dials a day.
Now I’m turning that experience into training datasets, Python evaluations & specialized AI workflows for site selection, underwriting, industrial facilities & cold chain logistics.
How we cook: take a real workflow, document the facts, check the math & build tests with known answers. Then see where the model gets it right, misses a detail or should say “I don’t know.”
We run those tests on our own compute and use the results to improve the system.
Why? AI needs to handle the details that change a deal: a lease clause, a power constraint, an expense assumption or a break in the cold chain.
I bring the operating experience & compute. Looking for collaborators who bring hard problems worth solving.
What would you test against 30 years in the trenches?
DM me. Let’s build. 🐝
🐂 DOES IT PENCIL? ✏️
30 years. $8B in CRE transactions.
My first question was never “What’s the cap rate?”
WHO owns it? WHY sell? Is it BANKABLE?
Environmental? Title? Debt? Hidden liabilities?
What’s the cost, time & who can cure?
Before the OM. Before the listing.
You can’t price what you can’t defend.
THE PENCIL ✏️
No sides. Just evidence.
#CRE #AI #DoesItPencil
“My mentor taught me to write it down on paper with a pencil. Every assignment. Every time. It became one of the most valuable processes in developing my intelligence and judgment. I did the work.”
A pencil with an eraser taught us something AI still needs to learn: the value isn't being right the first time. It's knowing how to find what's wrong, correct it, and defend what remains.
Why the Pencil?
Thirty years ago, when I was leading teams in commercial real estate, I had a simple requirement.
Before you brought me a deal, take out a piece of paper and a pencil.
Write it down.
Show me the income. Show me the expenses. Show me the debt. Show me why you believe the opportunity works.
People would ask why we didn't just use the software.
Because software could give you an answer without teaching you how to arrive at it.
The pencil made you think.
And the pencil had something else.
An eraser.
When an assumption was wrong, you corrected it. When the numbers didn't work, you recalculated. When you couldn't explain the result, you went back to the beginning.
That wasn't a limitation. That was how you learned.
That's where Does It Pencil? comes from.
Today, we have extraordinary computing power, specialized AI agents, and mathematics running in code. But the principle hasn't changed.
Show the assumptions. Work the math. Check the evidence. Correct what doesn't hold.
Let experienced people make the call.
The technology has changed. The discipline hasn't.
Does It Pencil? ✏️
GOT DATA? LET’S PENCIL IT. ✏️🐂
Before you take the asset to market, take its intelligence through preflight.
Books. Leases. Operating history. Equipment performance. The knowledge earned running the business.
Organize it. Validate it. Understand what you own.
We know the sale-leaseback play:
Sell the asset. Lease back its use.
Now pencil the sale-data-back:
Could you license a curated dataset while retaining the underlying records and your own operating intelligence?
In this analogy, frontier AI firms are the “1031 buyers”—the potential demand side looking for the next acquisition.
Who has the rights?
What does the buyer actually need?
Does the value justify the deal?
https://t.co/6IC7pV3OG9 | Your Private AI Office
Asset preflight. Informed decisions.
@CryptoALH The spaces are dog shit anymore - I am looking for GPUs I have a fleet of blackwells 6000 & 5090s I’m looking to buy bulk 3090s 3090ti - 500 and up 🆙
NVIDIA — DOES IT PENCIL? ✏️🐂 😂
Architecture is the location.
Workload is the tenant.
🏙️ CLASS A
1,000 RTX PRO 6000 Blackwells
96GB each × $16,995
GPU acquisition: $16.995 MILLION
Aggregate VRAM: 96TB
🐐 CLASS GOAT
10,000 RTX 3090s
24GB each × $1,000
GPU acquisition: $10 MILLION
Aggregate VRAM: 240TB
Add compatible NVLink bridges:
5,000 two-card pairs.
48GB physical VRAM per pair.
Software must support splitting the workload.
$6.995M less in GPU acquisition.
2.5× the aggregate VRAM.
Bridges and infrastructure cost extra.
But here comes the operating statement.
Reference GPU power:
CLASS A: 0.6 MW
CLASS GOAT: 3.5 MW
Before hosts and cooling.
Cheap acquisition.
Bigger electric bill.
More equipment to maintain.
Does your tenant need a 96GB floor plate—or can you lease 5,000 paired 48GB units to workloads that fit?
Show me paid utilization.
Show me completed jobs per dollar.
Show me cash flow after power, cooling and repairs.
30 years in CRE taught me:
The acquisition price starts the underwriting.
The rent roll finishes it.
CLASS A OR CLASS GOAT?
DOES IT PENCIL? ✏️🐐
Zero Cash Flow Deal 😂
HOTEL PORTFOLIO DISPOSITION ACTIVATED. 🐂
$85M sale.
$45M debt payoff.
$40M equity before costs.
Next move: 1031 into STNL industrial.
Credit tenant. 20-year lease.
ZERO CASH FLOW.
“Why buy something that pays you nothing?”
Because cash distributions and economic return are different calculations.
Tenant rent services the loan.
Principal amortization builds equity.
A properly structured exchange can defer the gain.
But here’s where the pencil gets sharp:
Replacing $45M of debt alone does NOT complete an $85M exchange.
Where does the $40M equity go?
What’s the replacement purchase price?
How much debt actually amortizes?
Who guarantees the lease?
What taxes arise WITHOUT cash distributions?
What’s the building worth when the lease expires?
Zero cash flow can still mean taxable income.
Does it pencil?
Potentially. Show me the basis, lease, debt schedule, and residual.
A 20-year lease is an input.
The structure is the deal. ✏️