30초 요약
Peter Yang’s early test of Claude Duet centers on its scheduled check-ins: the assistant proposes tasks using information from connected services. He demonstrates a morning brief, writing guide, channel analytics, a mock sponsorship dashboard and a learning activity for his child. A ride-booking attempt fails at a human verification screen, and Yang cannot find a way to take over the cloud browser. He calls Duet proactive but still prefers Grok Bot overall at the time of recording.
- 1A recurring check-in drives the proactive suggestionsYang sees a scheduled task running every three hours during the day. It surfaces sponsorship and recording preparation; he considers the frequency too high for him.
- 2Built-in skills produce a morning brief and writing guideHe runs a morning briefing and a style-learning task based on his previous messages and documents, then reviews draft replies.
- 3A short request produces channel and sponsorship dashboardsDuet analyzes public channel data and makes a separate sponsor dashboard. Yang specifies that the sponsorship figures shown are mock data.
- 4Human verification blocks a ride-booking attemptYang cannot find a way to view or interact with Duet’s cloud browser to complete the check. He also cannot easily add his local skills.
- 5Duet is most proactive for Yang, but Grok Bot remains his favoriteHe compares several agents on features and his own experience, and says his ranking could change as Duet develops.
핵심 정리
- Yang says Duet’s scheduled check-ins suggest tasks from connected information; he prefers fewer check-ins than the default he saw.
- He uses generated reports and drafts for planning, while reviewing important messages himself.
- A browser verification screen blocked his ride-booking test because he could not take over the browser.
- His comparison is based on an early research preview and his own workflows.
챕터별로 다듬은 전문 읽기
Tasks, channels and proactive check-ins
Yang tours the early interface and its scheduled task suggestions. He says the default check-in frequency is higher than he needs.
I had early access to Claude Duet, Anthropic’s personal agent, and found it unusually proactive. I use it for a morning brief, channel reports, a sponsorship dashboard and a learning activity for my child. I also show a failed attempt to book a ride and compare it with other assistants.
Duet has a dedicated mobile app and a web interface. In my account it already had access to services connected to Claude, so I did not have to reconnect them. It read connected information about my work and suggested preparation for sponsorship calls, podcast interviews and other tasks.
Each suggested task opens a thread associated with the main conversation. The interface groups active, upcoming, scheduled and resolved work. Channels can be organized into groups and given icons and colors, though I cannot customize the main avatar as I can in some competing products.
The recurring check-in is the main reason I find Duet proactive. In my preview it runs every three hours during the day and surfaces items such as a sponsor agreement or an upcoming recording. Five check-ins feel excessive for my schedule; I would prefer one in the morning and one in the afternoon.
I think the generated artifacts are another distinguishing feature. A slash menu exposes built-in skills, including a morning brief that renders a styled page with calendar and priority information.
Briefings, writing and dashboards
Built-in skills produce a morning brief and writing guide. Yang also demonstrates analytics and a sponsorship dashboard with mock financial data.
Another skill examines examples of my writing and produces a style guide. In my test it used emails, messages and documents to describe habits such as opening with the point. A daily to-do list and email triage page provide drafts; I still want to review important replies before sending them.
I asked for a dashboard of my YouTube channel’s performance and linked the channel. The resulting report used public video data and suggested that personal-agent videos and some solo videos performed well for me. It included charts by date, topic and video length.
A separate sponsorship dashboard shows revenue, sponsor conversations and upcoming work. The financial figures in the demonstration are mock data. I can also ask for scheduled updates to a dashboard or for it to be sent to me.
A learning activity and a blocked booking
Yang builds a children’s learning artifact, then shows a ride request blocked by human verification. He cannot find browser takeover or an easy route for local skills.
With my child I created an AI Explorer Club artifact with a small animated character and lessons. The first lesson explains chatbots and provides a prompt to try. I find it too text-heavy, but could ask the assistant to make it more visual.
In the research preview I tried a built-in ride-booking skill. The assistant asked for pickup and destination, but Uber presented a human verification screen that it could not complete. I could not find a way to open or take control of Duet’s cloud browser to resolve it myself.
I also could not easily add skills from my local computer. I expect these features could change, but I see browser access and an easier path for personal skills as important gaps in the version I tested.
How Yang compares personal agents
Yang compares features and trade-offs in several products. His conclusions reflect the research preview and his own preferences.
For my work, Grok Bot remains my favorite because its X integration helps with research, although I find its email functionality unreliable and it is paid. I like Muse’s design and ability to call US businesses, while I consider its underlying model a limitation in some tasks.
I see the advantage of ChatGPT dots as its connection to existing ChatGPT threads and memory, but in my experience it has been slower and less useful than my own dedicated assistant thread. Duet is the most proactive out of the box for me, yet lacks the browser and skill features I want.
I still use Grok Bot daily alongside ChatGPT and Claude for individual work. If Duet becomes more capable after the research preview, I may use it more because Claude already knows much of my working context. At recording, I describe Duet as a preview with a waitlist and predict it may eventually be integrated into the main Claude app.
같은 채널의 다른 글
46:40EN
Peter Yang· 2026년 11월 8일팟캐스트 · 2명Ashley Alexander: How AI Helps Me Care for My Parents With Cancer게스트 Ashley Alexander · 진행자 Peter YangA caregiver uses AI to prepare for appointments, organize records and ask better questions—while keeping doctors at the center of treatment.
17:50ENPeter Yang· 2026년 11월 4일팟캐스트Peter Yang’s Five Steps for Using AI to Help Care for His MotherYang uses a shared care brief and organized records to navigate appointments, while making time for family.
15:03ENPeter Yang· 2026년 11월 3일팟캐스트How Peter Yang Used an AI Phone Agent to Negotiate His Internet BillAn AI call negotiated a lower price and faster service, but Yang still had to approve the change himself.
33:09ENPeter Yang· 2026년 11월 1일팟캐스트 · 2명Jaclyn Konzelmann: When AI Can Build More, Choose What Deserves to Exist게스트 Jaclyn Konzelmann · 진행자 Peter YangGoogle Labs pairs short strategy documents with prototypes, user co-creation and lessons from experiments that end.
13:05ENPeter Yang· 2026년 10월 28일팟캐스트Peter Yang Tests Genie 4: Convincing Worlds, Unfinished GameplayFive live demonstrations show scenes that respond to prompts and sound, alongside inconsistent actions and limited game mechanics.
1:06ENPeter Yang· 2026년 10월 26일팟캐스트Peter Yang on Setting Up Payroll for a One-Person BusinessIn a paid partnership, Yang describes the paperwork he encountered and a payroll product integrated with a business account.