Earmark, co-founded by Mark Barbir and Sanden Gocka, aims to revolutionize meetings by transforming discussions into actionable work—like specs and tickets—before the call ends. Their approach goes beyond traditional AI notetakers that merely summarize. Instead, Earmark employs multiple agents simultaneously during meetings, handling tasks such as translating technical jargon and drafting product specifications on the fly. This real-time capability sets Earmark apart and addresses a common pain point for professionals who often struggle with post-meeting follow-up. The team initially focused on developing a coaching tool for Apple Vision Pro presentations but pivoted to create a web-based meeting assistant. Their decision to implement an ephemeral architecture, which doesn’t store data, turned into a key selling point for enterprise clients. They successfully reduced the cost of running AI during meetings from $70 to under $1 through efficient techniques like prompt caching. Earmark also tackles challenges in search and analysis, recognizing that traditional vector search methods often fall short. They’re building an "agentic search" system that combines different retrieval methods to extract insights over time. Earmark's design targets product managers, treating them as the extreme user. This focus drives features that could benefit a broader audience. The founders envision an AI chief of staff that extends beyond automating tasks, aiming for a tool that effectively manages deliverables and enhances productivity. Their discussion highlights both the technical hurdles of developing such a product and the visionary aspects that could reshape how teams approach meetings and work output.