Free tool

A second opinion that texts back.

Most deals fall apart on something nobody asked about. Not the cap rate — the reserve that was set at a third of what the roofs will cost, or the concessions sitting under a rent roll that reads clean.

This is the prompt we use to build a screening assistant that asks those questions back. Send it a deal summary and it tells you what is missing before you call the broker.

Inside

The prompt itselfPaste it into Claude or ChatGPT and it becomes a deal screener that argues back. Nothing to install.
Connecting it to WhatsApp or TelegramSo you can forward a rent roll from your phone while you are still standing in the parking lot.
The eleven questions it asksBreak-even occupancy, exit cap sanity, reserve adequacy, concessions buried in the rent roll, and the rest.
What it refuses to doIt will not value a property or tell you to buy. It tells you what you have not checked yet.

Here it is. A copy is on its way to your inbox as well.

1 · The prompt

Paste this into Claude or ChatGPT as the system prompt, then send it any deal.

You are a multifamily deal screener working for a limited partner, not for a sponsor.

When I send you a deal, do not value it and do not tell me whether to buy.
Your job is to find what has not been checked.

Work through these in order, and say plainly when the answer is missing:

1. Break-even occupancy. Operating expenses plus annual debt service, divided by
   gross potential rent. Compare it to current occupancy and tell me the cushion
   in percentage points.
2. Exit cap versus entry cap. If the exit is equal or lower, say so directly.
3. Capital reserve per unit per year, against the age of roofs, HVAC and plumbing.
4. Concessions. Ask whether the rent roll shows face rent or effective rent.
5. Economic occupancy versus physical occupancy, and the gap between them.
6. Loss to lease, and where the market rent figure came from.
7. Bad debt as a percentage of gross potential rent, for three years not one.
8. Property tax after reassessment at the purchase price, not the seller's bill.
9. Insurance: a current quote for new ownership, not the expiring premium.
10. Loan type, rate structure, maturity date, and any rate cap expiry.
11. Total sponsor fees across the hold, in dollars, versus the promote.

Then give me exactly three questions to put to the sponsor, ranked by how much
money rides on the answer.

Be short. No preamble. If a number is missing, say "not provided" rather than
estimating it.

2 · Getting it into WhatsApp or Telegram

Both work the same way: a small relay sits between the messaging app and the model, so you can forward a document from your phone and get the screen back as a reply.

Telegram is the faster one to start with. Create a bot through BotFather, take the token, and point it at the model API with the prompt above as the system message.
WhatsApp needs the Business API through a provider such as Twilio, which means a verified sender and a per-message cost.
Either way the relay is about thirty lines. The prompt is the part that matters.

3 · What it costs

Screening one deal runs a few cents of model usage. Telegram itself is free. The honest cost is your attention when it tells you something you would rather not hear.

4 · Where it is weak

It reads what you give it. Send it a summary page and it screens a summary page. It cannot see the property, the neighbourhood, or the sponsor's track record, and it has no opinion worth trusting on any of those. Treat it as the checklist you would otherwise forget, not as diligence.

If you want the reasoning underneath these questions, that is what the Journal is. Start here.

Educational only. Not an offer to sell or a solicitation to buy any security, and not investment, legal or tax advice. Nothing produced by this tool is investment advice or a valuation. All investments involve risk of loss.