Was it human or AI? Building software where the answer is "Ambiguous"

Most AI products treat agents as tools: something you prompt, supervise, and wait on. Ambiguous starts from a different premise: what if humans and agents worked together as coworkers? That requires an AI-native workspace: shared identity, tools, permissions, formats, tasks, comments, messages, calendars, workflows, and a visible trail of work. To build it, Ambiguous has been running its own software factory: 200+ autonomous agent runs, 8,000+ commits, and nearly 4 million lines of code in 30 days. Ryan Waliany, CEO and co-founder of Ambiguous AI, joins Foundations for a research preview of Ambiguous before public release: an early look at the product, architecture, and design decisions behind building for Human + AI Parity from first principles. The talk will go inside how Ambiguous is being built: - AI coworkers as first-class participants, not integrations - The workspace as the shared environment for human and agent work - Every surface designed for both human use and agent execution - Every client, including agents, calling the same underlying API - Documents, tasks, messages, calendars, and workflows as one system of record - Actions leaving the same trace whether completed by a human or an agent - Early lessons from building the suite before public release For founders, product builders, and operators thinking seriously about what software becomes when humans and agents work together as coworkers.