Lend a Hand
I'm always learning. These aren't random curiosities—they're directly connected to what I'm building. If you have experience here, I'd love to learn from you.
People I'd Love to Meet
- Creative & Automation Agency Owners
- AI/ML Engineers (Matching Problems)
- Two-Sided Marketplace Operators
- Mortgage & RE Coaches
- RE Broker Owners (Ancillary Rev)
- Partnership Leaders in Fintech/Proptech
How You Can Help
- Make an intro Know someone I should meet?
- Share a resource Article, framework, or tool.
- Challenge my thinking I'd rather be wrong and learn than confident and stupid.
Active Investigations
I treat my advisory and operational work like a research lab. Here are the hypotheses I'm currently testing in the market. If you have data points (pro or con), I'd love to see them.
Brokerage ancillary attach rates are an operational problem, not a sales problem.
ValidationAgents don't sell ancillary services because it breaks their flow. The system needs to capture it automatically.
Removing the audience accelerates authentic disclosure
HypothesisTraditional therapy requires sessions to build trust. When there's no human, no memory, no judgment risk—people skip straight to honesty. Ikigai tests this in recruiting. What other domains have value locked behind trust-building overhead?
Education-first communities generate higher LTV customers than direct marketing.
ValidationTeaching prompt engineering builds more trust than a demo. But does that trust translate to SaaS subscriptions.
Post-close integration has a 90-day window before organizational resistance calcifies
HypothesisPE firms optimize for pre-close diligence but the real risk is post-close human integration. Employees are open and curious in the honeymoon period—miss it and they either leave or entrench. Mapping values alignment early may predict success better than financial diligence.
Real estate professionals value hyperlocal market intelligence more than national trends
TestingAgents don't care what the Fed did—they care what's happening in Plymouth Township. Neighborhood-level content creates stickier partnerships with loan officers, and that visible value attracts other agents who want the same edge.
Native multimodal LLMs can classify and extract mortgage documents without specialized OCR
HypothesisAWS Textract is expensive at scale. Testing whether Claude/GPT-4V can handle document classification and basic data extraction at 90%+ accuracy—potentially eliminating the OCR layer entirely for most use cases.
In vertical B2B SaaS, community education creates an influencer flywheel that compounds
TestingTeaching creates trust, trust creates customers, customers become advocates who teach others. The Automated LO model: prompt engineering sessions → product adoption → successful users evangelize. Paid ads interrupt; education compounds.