J.Doe is promoting a free, invite-only private beta for a personal AI that it says connects conversations, documents and screen context, suggests next steps, and hands approved work to agents including Codex and Claude Code. The pitch: less time briefing the AI on the meeting you just sat through. Source Source
The company says users can import recordings or record directly on their phones. Its illustrated workflow combines a conversation about an upcoming board meeting with an on-screen client profile specifying Germany and mid-market customers. From that context, it proposes research, a sourced market-entry recommendation and a draft briefing email. Source
In a November 17 post, Brian Roemmele described J.Doe presenting the next step as a “decision card” that he approves or dismisses, with approved work moving to agents inside his existing apps. He also said corrections carry forward without having to restate a preference. Source
In J.Doe’s example, a user asks for recommendations first and linked evidence, and the system asks whether to apply that preference to another brief. Remembering the house style would be welcome. Source
A decision card isn’t a blank check
The site’s market-entry example says, “Ready for your review. Nothing is sent until you say yes.” But its broader FAQ says, “Your permissions and approval settings determine what J.Doe can do.” That isn’t a blanket promise that every action requires a separate confirmation. Source
J.Doe’s FAQ says users can choose to delete their data, while beta exports require an email request to support. It directs users to a separate privacy policy for retention and deletion rules. The FAQ itself doesn’t specify retention periods, processing locations or whether data is used for training. Source
The site advertises iPhone and Mac apps and links to a beta application. Spots are advertised as limited, and applying doesn’t guarantee an invitation. Separately, J.Doe says its O1 device is expected to launch in early 2027. The site gives no specifications or exact release date. Source Source
The company’s examples and Roemmele’s account don’t establish how reliably J.Doe anticipates work, remembers corrections or completes tasks. Source Source Checking the AI’s assumptions could eat into any time saved on the repeat briefing.
Comments