Platform Scope
By the numbers.
What has actually shipped. Not features promised, not screens designed, not pitched — shipped, server-rendered, live on the public site, reflected in the public commit history.
Snapshot as of May 19, 2026. Live counts on the changelog.
Build cadence
Discipline of shipping, not just thinking. Every commit lands on main and is reflected on the live site within minutes.
AI architecture
Five specialist agents and a synthesis agent, each with strict JSON contracts and typed fallbacks so the orchestrator degrades gracefully.
Coverage
Market intelligence depth across the highest-conviction US investment metros, with submarket-level supply, demographics, and regulatory context.
Research depth
Long-form analytical writing, worked case studies, and live build notes. Read like an institutional research desk.
Product surface
The feature set an institutional investor actually uses — not a generic dashboard, but a tool kit built around the workflow of finding, analyzing, and tracking deals.
Architectural Decisions
Five decisions that shape the platform
The numbers above are scope. These decisions are why they came out the way they did.
Deterministic financial math, AI judgment
Cap rates, DSCRs, and cash-on-cash returns are pure functions — same input, same output. AI is reserved for OM extraction, market narrative, hidden-risk pattern matching, and IC-style synthesis. Numbers reproduce exactly; judgment is augmented.
Strict JSON contracts with typed fallbacks
Every agent has a Zod-validated output contract. If the model returns invalid JSON, the orchestrator falls back to a typed stub rather than crashing. The platform stays up even when the AI provider is unreachable.
Parallel specialist fan-out, sequential synthesis
The four specialist agents (underwriting, market, risk, operations) run concurrently against the same DealInput. The IC synthesis agent runs sequentially, consuming all four outputs. Latency stays low even as the report depth grows.
Memory layer for cross-deal context
Most tools forget the moment you close the tab. The platform persists every analysis and surfaces what changed when you revisit a deal — rent trajectory, market shift, thesis evolution. Institutional investors track deals over months; the tool was built to match.
Provider-agnostic AI client
Behind the agents is a single client that abstracts the model provider. Switching between providers is a config change, not a refactor. Reduces vendor lock-in and lets the platform route per-task to the model best suited for it.
Where to go next
See what the numbers translate to.
The scope is real. The features are live. Walk through the platform and inspect it yourself.
Related
Dive into the depth
Numbers are the scope. These pages are what the scope translates to.
About this project
Why the platform exists, the four design principles, and the application context.
Research Lab
The multi-agent AI pipeline, layer by layer, with file paths.
Long-form research
Institutional analysis on underwriting, market selection, and AI workflows.
Changelog
Every ship, dated and explained — the live source of truth for what is on the platform now.