Fundamental-Ava

Fundamental Research Labs is not selling a chat box with a spreadsheet plugin. The company's published research — Project Sid, on many-agent simulations toward AI civilization — describes architecture, specialization, law, culture, and religion inside multi-agent systems. That research lineage is what Shortcut puts to work on spreadsheets, and it is why the product is framed around agent collaboration and traceability rather than conversational convenience. For an analyst who has to defend a model to a reviewer, controller, or auditor, the reasoning log is the whole point: you can follow the chain from question to model rather than accepting a black-box answer. The strengths line up with that thesis. Spreadsheet-native reasoning means the tool works on the artifact analysts actually use, not a text summary of it. Parallel simulations let you compare what-if scenarios side by side without rebuilding the file by hand. Multi-agent collaboration on a single problem is unusual in this category and is the direct output of the company's research focus. And the team's pedigree — MIT EECS, Stanford NLP Group, Google X, Citadel — plus backing from Andreessen Horowitz and Prosus, is the kind of signal that matters when you are betting on an early-stage vendor. The honest weaknesses are about the stage of the product, not the idea. Shortcut is explicitly the company's first step toward its digital humans mission, so you are an early customer, not a late one. The public surface is thin: the site describes what Shortcut does but gives few specifics about plan structure, feature limits, or how the audit trail is exported and reviewed in practice. There is no mobile app and no documented integrations. Before you put Shortcut in a workflow that feeds real decisions, validate three things against your own data: whether the reasoning log is granular enough for your reviewers, whether parallel simulation matches how your team frames scenarios, and how model outputs are exported or reconciled back into your existing workbook conventions. Where it fits: finance and operations teams running repeated scenario work who need a defensible trail, enterprise data teams evaluating domain-specific agentic analytics, and researchers who want to test multi-agent reasoning on real business data. Where it does not: casual spreadsheet users who will never open the audit trail, and anyone who needs a fully documented public feature spec, published integration list, or mature support surface before they will evaluate.