What does an investing agent actually do?
An explicit way to distinguish research assistance, simulated decisions, and real execution before evaluating an AI investing tool.
Updated
The Thesis Lab field guide
Your principles.
One reviewable
first version.
Start with an investing principle. Review the Unknowns, resolve what matters, and approve one saved version before exploring a simulated decision.
Follow the first-use guideFree closed beta · New accounts require an invitation · Existing login host
Choose a question
Six paths through the work
From a first principle to an evidence review.
01
Agent fundamentals
What an investing agent can do, what its output means, and where human judgment remains necessary.
02Principles & approval
Turn an investing approach into explicit rules, unresolved questions, and versions a person can review.
03Evidence & verification
Check claims against original sources, publication times, definitions, and competing explanations.
04Simulation & testing
Read historical simulations with clear cutoffs, execution assumptions, and limits on inference.
05Operations & security
Understand saved-data encryption, AI processing, permissions, recovery, and device trust.
06Thesis Lab guides
Use the current free closed beta with an accurate account of invitations, workflow, and supported data.
The reading room
Start with something useful
8 original guides · Updated 2026-10-26
How do you turn investing principles into rules a person can approve?
A worked method for exposing ambiguity, recording Unknowns, and approving a specific strategy version before simulation.
Updated
How can you tell a sourced fact from an agent’s inference?
A claim-by-claim review exercise for separating reported evidence, calculations, interpretation, and unresolved questions.
Updated
What information belongs in a point-in-time investment replay?
How to distinguish a reporting period, publication time, retrieval time, and simulation time before reading a historical decision.
Updated
What should you check before trusting a backtest chart?
A practical audit of assumptions, friction, path risk, and sensitivity that keeps simulated results separate from an investment record.
Updated
Does browser-encrypted research stay private when you use AI?
A plain-language distinction between encrypted saved research, unlocked browser data, selected AI requests, and recovery.
Updated
How do you review your first strategy in Thesis Lab?
A hypothetical first-session walkthrough with one reviewable output: an approved principles version. A supported dated simulation is an optional next step.
Updated
How do Thesis Lab’s three closed-beta invitation slots work?
Answers about new-account invitations, used slots, unused-link expiry, verified issuers, and free-beta duration.
Updated