Investment research
A research workspace built for investment analysts and their AI agents — where agent work is organized into tasks, reviewed by you, and kept as durable memory.
The workspace
Research is more than notes. lux72 is designed to be durable research memory across all of your work. Connect Excel, PowerPoint, Google Sheets, and more — your work becomes portable from one AI agent to the next.
An agent's context window is small and its session is short. The workspace is
neither. Agents keep research_state current as they work — the
standing summary of where the research stands. When an agent opens a workspace it
gets a context packet: the whole history compressed to what actually
matters, so the next agent picks up exactly where the last one left off.
Context packet · on open
AI · A first-class user, on a leash you hold
Agents organize their work into tasks and record every step as task activity — findings, calculations, source checks, assumptions. You review it: approve and the agent moves on, reject and it tries again. Approved output is promoted into a note, so nothing durable enters the workspace without passing through you. Every agent action is attributed and versioned; deletes are denied to agents and routed to a human.
Why · This exists
I built lux72 to manage my own research process. I wanted a better way to work with AI agents than a chat interface — somewhere their output became long-term memory of the work rather than disposable conversation, and somewhere I could see exactly what they did without taking it on faith.
If you want to harness AI agents in your research process, try out lux72.
— Tim, founder