Two guys, $1.8 billion in revenue, and one brother hired mostly so the founder wouldn't be lonely. That was the story the New York Times ran this week about Medvi, a telehealth startup built almost entirely with AI tools. Behind the Glass We shipped five new tasks this week. Two of them are LOI generators: one for commercial leases, one for purchases. Upload your deal terms, and the AI drafts a letter of intent using our document standard skills. We also added a Closing Checklist generator (built on a stack of due diligence skills covering PSA review, environmental, financing documents, and three different closing timeline frameworks), a Multifamily Rents research task, and a Property Map generator for marketing packages. Behind those five tasks sit 15 new skills. The due diligence and closing skills alone represent what would normally live inside a senior associate's head after years of deal experience: how to coordinate a 30-day expedited close versus a 60-day standard timeline, what to flag in a PSA, how to evaluate environmental risk, how to structure lender selection. We wrote all of it down so the AI coworker can execute it consistently, every time. That brings us to 72 published tasks on the platform. A few weeks ago we had fewer than 50. Hundreds more are on the roadmap, and a meaningful share of what we're building comes directly from your conversations with us. When you tell us what's eating your time, we turn it into a task. The library is literally growing because of your feedback, so keep it coming! The Horizon We're working on some ambitious ones. An industrial site selection engine that includes drive-time employment cost analysis. A set of small bay industrial tasks. A project to estimate usable land from wetlands, topography, and location maps (this one is tough, and I'm genuinely unsure if it will work). The ability to assess deals in bulk: upload a CSV of addresses, get a general go/no-go estimation for each. Plus integrations with HubSpot, and comps for self storage, senior housing, and student housing. Here's the thread connecting all of these: most of what CRE professionals pay for in expensive software subscriptions is now highly feasible with AI. The cost of building these tools has collapsed. The question isn't whether AI can do it. It's who builds it first for CRE, and whether they build it well. Bigger Picture Back to the Medvi story. Matthew Gallagher spent $20,000 and two months to launch a company that did $401 million in its first year. He used ChatGPT, Claude, Grok, Midjourney, and Runway to build the product, the marketing, the customer service, and the internal analytics. His only employee is his brother. The article is worth reading in full (Link) . But here's what I keep thinking about: if one person can build a $1.8 billion revenue company selling weight-loss drugs online, the day is coming when one person manages a billion dollars in real estate assets. One person, with AI coworkers handling the underwriting, the due diligence, the reporting, the investor communications, and the asset management. We're grateful for the opportunity to help deliver to each one of you a Multiplier Effect from AI. Spencer