InvestMoat.
An AI-era research framework for moat investing — scoring stocks on moat durability, growth trajectory, and live valuation to surface high-conviction opportunities.
Moat investing — the practice of identifying companies with durable competitive advantages — is one of the most proven long-term strategies in public markets. But actually doing it well is research-intensive: you need to assess industry dynamics, competitive positioning, financial trajectory, and valuation, all at once.
Most retail investors either skip this depth entirely, or spend hours per stock piecing together data from earnings transcripts, news, screeners, and financial databases — only to make a decision based on gut feel anyway.
The opportunity: use LLMs and structured data pipelines to systematize the moat assessment process — producing consistent, repeatable scores that surface genuinely strong businesses at reasonable prices.
Moat Durability
AI-powered assessment of competitive advantages: network effects, switching costs, cost advantages, intangible assets. Pulls from SEC filings, earnings calls, and industry analysis.
Growth Trajectory
Revenue growth trends, margin expansion, reinvestment rates, and TAM runway — structured into a consistent forward-looking score across every stock in the universe.
Live Valuation
Real-time price and fundamentals data mapped against intrinsic value estimates — so high-quality businesses are surfaced only when they trade at sensible prices.
Portfolio Builder
Combine the three scores into a composite ranking, filter by sector or risk tolerance, and construct a concentrated, high-conviction portfolio — with reasoning attached to every position.
LangGraph for structured multi-step reasoning
Moat assessment isn't a single-shot LLM prompt — it requires sequential steps: retrieve filings, summarize competitive dynamics, cross-reference with news, produce a structured score with citations. LangGraph lets each step be a distinct, testable node with clear state handoffs.
n8n for data ingestion and refresh triggers
Earnings releases, price updates, and SEC filings trigger n8n workflows that kick off re-scoring pipelines. This keeps the platform fresh without polling on a fixed schedule — scores update when the underlying data actually changes.
Next.js frontend with server components for performance
Stock pages are heavy on data — fundamentals, charts, score breakdowns, reasoning summaries. Server components handle the data-fetch-and-render path at the edge, keeping time-to-interactive low even for complex pages.
Scores as structured outputs, not prose
Every assessment produces a JSON score object — numeric ratings per dimension, confidence levels, and evidence citations — rather than free-form text. This makes scores comparable across stocks and auditable, not just readable.
This project sits at the intersection of two areas I care deeply about: AI agent systems and long-term investing. After building AI pipelines at Storia Technologies and seeing firsthand how LLM-based agents can compress research-heavy workflows, I wanted to apply those techniques to a domain where the quality of reasoning genuinely matters.
Moat investing appealed because it has a well-defined evaluation framework — Warren Buffett's original criteria, expanded by Pat Dorsey — which maps cleanly onto structured LLM outputs. It's a domain where AI can genuinely augment human judgment rather than replace it.