An Agentic Stock Research Platform

CASE STUDY: AI Agents That Read Every Filing and Flag What Changed

We built a private research platform that tracks AI, robotics, and space companies against a written set of beliefs, with AI agents that read new filings and a person who approves every change. It is a research tool and gives no investment advice.

→ The Challenge

An investment thesis usually lives in someone’s head. Nobody writes down what has to stay true, so nobody notices when it stops being true.

Pain Points

More Filings Than Anyone Can Read: Dozens of companies file reports all year, and the one line that matters is buried in the rest.
Beliefs Never Written Down: Without a stated belief and a stated warning sign, there is nothing to measure new information against.
Warnings Arrive Too Late: By the time a broken belief shows up in the price, the early signs have been sitting in filings for months.

→ Our Solution

We built Investr, a research platform that scores each company against written beliefs and uses AI agents to keep those scores current, while every change waits for a person’s approval.

We Built

A Written Thesis the System Tracks: Each belief is recorded with what to watch and when to worry, and is marked holding, weakening, or broken. Companies are linked to the beliefs they depend on.
Research Agents With a Verifier: One agent reads new official filings and proposes score changes or alerts. A second agent checks that work before it goes anywhere. The agents cannot write to the live record.
Scoring and Plain Labels: Every company is scored out of 100 on the strength of the business and the fairness of the price, then sorted into labels such as Buy zone, Wait list, and Watch-only.
An Approval Inbox and Full Platform: Proposals wait in an Inbox until a person accepts them. Around that sit prices, risk, upcoming events, holdings, and a journal of decisions, in an app that installs on a phone and connects to Claude.