Descriptions:
Former Uber engineer and Lindy cofounder Flo joins David Ondrej to discuss one of the most underexplored problems in the current AI moment: how multi-agent systems will evolve from solo developer tools into genuine team collaboration infrastructure. The central analogy is compelling — today’s AI agent landscape resembles the era before Google Docs, when collaboration meant emailing files back and forth rather than working on a shared artifact. Right now, developers each run their own Claude Code, Codex, or similar setup locally, and the tooling to coordinate agents across a team simply doesn’t exist yet.
Flo walks through live demonstrations of agentic workflows powered by Deep API, a tool that provides agents with a single API key granting access to web scraping, deep research, email, and browser automation. The examples include an agent that autonomously compiled 100 licensed general contractors in Austin — with names, websites, phones, and emails — after a 30-minute run, and another that analyzed news sentiment for eight stock tickers and emailed a summary in four minutes. These illustrate how agents can now complete multi-step, multi-tool tasks that previously required human orchestration throughout.
The conversation also covers Flo’s own daily agent stack — a morning briefing, email drafting and labeling agents, a meeting preparation agent, and a calorie-tracking agent via iMessage — and why he believes the first platforms to establish themselves as the default agentic layer will be baked into the training data of future models, creating durable competitive advantages. DoorDash’s recently released CLI is cited as an early example of software being rebuilt natively for agent consumption rather than human interaction.
📺 Source: David Ondrej · Published August 10, 2026
🏷️ Format: Interview
