Riley Brown and Michael Shimeles discuss hands-on agent workflows from development and marketing perspectives. Their examples emphasize work completed inside real interfaces rather than chat alone. The episode's product comparisons and performance impressions are their experience at the time of recording, not universal rankings or a current product guide.
For development, they describe a build, test and fix loop in which computer use helps an agent inspect the result. Marketing examples use reference material, clear evaluation criteria and repeated feedback. They distinguish a usable workflow from a single generated draft; an agent's own quality score does not independently establish commercial performance.
The conversation considers reusable tools, separate threads and scheduled review tasks as building blocks for agent workflows. Shimeles discusses Convex while identifying it as his employer. Both speakers retain a role for human communication, design judgment and choosing the right interface, while presenting agent-first applications as an emerging direction rather than a guaranteed outcome.
Watch on YouTube
