AI Finance Lab.

The thesis

Affordable housing finance is a domain where AI assistance can compress hours of statutory research into minutes — if the AI is grounded in primary sources, cites its work, and knows when to defer to human judgment. We're building tools that do exactly that.

This page collects what we offer now and what's coming.

Available now: The Claude Pack ($79)

The Claude Pack turns your existing Claude.ai subscription into an affordable housing finance assistant. It bundles:

• A Claude Project configuration — system prompt tuned for affordable housing finance work
• 50+ tested prompts — for LIHTC underwriting, QAP scoring, 4% vs 9% comparisons, OZ deal review, HUD compliance, and capital stack assembly
• Statutory reference pack — curated PDFs of Section 42, key IRS Revenue Procedures, OBBBA provisions, sample state QAPs, HUD CPD notices, and FHLB AHP guidance. Upload to your Claude Project for instant retrieval.
• Model workbooks — capital stack templates, AMI/rent limit calculators, basis boost analyzers, 4% vs 9% comparison sheets
• Setup walkthrough video — 20-minute screen recording
• Lifetime updates — new regulations, AMI tables, and QAP revisions delivered to your purchase email

You bring your own Claude.ai subscription (a separate paid product from Anthropic; we are not affiliated with Anthropic). The Pack is a one-time $79 purchase, no recurring fees.

Available now (MVP): Pro AI Q&A platform

Our hosted Pro AI Q&A is in active development. Unlike the Claude Pack (which runs in your own Claude.ai account), Pro AI Q&A runs on housingsubsidybrief.com using our content library and Anthropic's Claude API on the backend. Key planned features:

• Conversational interface — ask deal questions in natural language
• Grounded responses — answers cite primary sources (Section 42, 24 CFR, HUD notices, OBBBA, IRS Revenue Procedures)
• Knows when to defer — trained to flag situations requiring qualified counsel or CPA review rather than over-confidently answering
• Daily limits by tier — 3 questions/day for Reader (free), 20/day for Pro, 50/day per seat for Team
• Conversation history — saved chats for Pro members
• Privacy preserved — your questions are not used to train AI models. Anthropic processes the queries on the backend but does not retain them for training under our enterprise terms.

In active development: model tooling

Our approach to AI

We're not building AI for AI's sake. The tools we ship have to make practitioner work materially faster and more accurate than what's possible without them. That means:

• Grounded in primary sources — not generic language model output
• Citation by default — every claim traces back to a primary source you can verify
• Refuses to over-extend — trained to flag situations requiring human professional judgment rather than fabricate answers
• Privacy-respecting — your queries don't train AI models
• Transparent limitations — we publish what the AI can and cannot do reliably

Important caveats

AI-generated responses are not legal, tax, financial, or investment advice and should be verified against primary sources before any decision or transaction. Always consult qualified counsel, your CPA, and other appropriate professional advisors before relying on any AI output. See our Disclaimer for full details.

We're confident in the quality of what we're building, but the underlying technology has known limitations — hallucinations, outdated information, edge-case failures — that we work to mitigate but cannot eliminate. Use AI as a productivity tool, not as a substitute for professional review.

Get notified when new tools launch

The fastest way to know when a tool goes live is the weekly newsletter.

Claude™ and Claude.ai are products of Anthropic. The Housing Subsidy Brief is not affiliated with Anthropic. References to Claude.ai or Anthropic services describe how those products work; the Claude Pack and Pro AI Q&A use Anthropic's APIs under our commercial terms with Anthropic.