Claude Duet is beautiful. I still forgot to use it

The episode kicks off with the host expressing a mixture of excitement and exhaustion over new personal agents, highlighting Anthropic's recent release of Claude Duet. While impressed by its design and capabilities, the host laments personal agent fatigue, having tried many agents that haven't quite become essential. The episode's purpose is to evaluate whether Claude Duet, a new, well-designed agent, can rise above the rest and become the host's primary personal assistant.

The host describes Claude Duet as a personal agent app for iPhone and Android that integrates with Claude's ecosystem and tools. It's positioned for small business owners and busy professionals, not casual users or those managing extensive agent networks. The product emphasizes a polished design, smooth onboarding, and integration with existing plugins like Gmail, Figma, and task management tools. It aims to serve a middle-market user who needs reliable, straightforward assistance for work tasks.

The host praises Claude Duet's design as exceptionally polished, with attractive visuals, clear onboarding, and thoughtful organization. Key features include task lists to manage goals, artifacts like generated HTML or slides, and a channel system inspired by Slack for organizing conversations around projects. The app also supports plugins for various tools and keeps existing artifacts for context. These primitives are seen as strong foundations for an effective personal agent, facilitating smooth workflows and collaboration.

Despite its polish, the host criticizes Claude Duet for being insufficiently proactive; it doesn't nudge or alert the user routinely. Notifications are understated and infrequent, leading to forgetfulness. Performance issues also stand out; model latency is slow, often taking 10+ minutes to complete tasks like shopping for tickets or preparing documents, which kills engagement. Additionally, the absence of a desktop app and multiplayer/team capabilities means the agent isn’t integrated into the host’s regular workflow across devices or shared with team members, further limiting its usefulness.

The host tests Claude Duet on various tasks: reviewing email and Slack to identify priorities, shopping for event tickets, creating sales decks, and generating launch videos. Results are mixed; the agent does some tasks adequately but struggles with prioritization, speed, and fulfilling specific custom requests, such as creating launch videos from local skills. Frustration mounts over slow interactions, inability to close loops effectively, and the lack of real-time responsiveness, which impairs its role as a dependable coworker.

The host emphasizes that key features are lacking for the agent to be fully integrated into daily work routines. These include a desktop app, which would allow continuous, unobtrusive engagement during work hours, and multiplayer features to share agents with team members. Without these, Claude Duet remains a single-user tool that isn't yet capable of supporting collaborative workflows or team-based projects, limiting its potential even for small teams. The host recognizes that these are early days and expects improvements in future iterations.

The host concludes that Claude Duet is among the most polished agent launches but notes significant gaps in proactivity, speed, and collaboration. While its design, task management, and artifact handling are excellent, these strengths are undermined by delays and a lack of engagement hooks. The host advises trying the product but remains skeptical about its immediate fit for daily work, emphasizing that truly competitive personal agents need to be more proactive, faster, and connected across devices and teams to fulfill their promise.