The dots demo, take two | OpenAI DevDay 2026

Summary
In this official OpenAI DevDay 2026 follow-up video, an OpenAI team member presents an unedited, "take two" demonstration of OpenAI’s "dot" personal agent after the prior day’s live launch demo did not go as planned. Over a real-time phone call and a synchronized web workspace ("Your dot's computer"), she instructs dot to handle personal travel logistics (booking flights and hotels in Los Angeles) as well as internal work tasks (analyzing user feedback signals, mapping Linear tickets, and drafting Slack replies).

What is shown

• [00:00 - 00:46] Introduction explaining that the previous day's live demo had issues, leading to this recorded, unedited walkthrough of everyday dot usage.
• [00:47 - 03:03] Voice phone call to dot requesting flight and hotel arrangements for a next-day meeting in Hollywood at 3:00 PM; dot recommends flight options, evaluates LAX exit and traffic travel times, and proposes hotel options (Silver Lake Pool & Inn vs. Cara).
• [03:04 - 03:36] Presenter displays her laptop screen showing the ChatGPT interface alongside a sandboxed browser ("Your dot's computer"), where dot sends hotel photos, maps, and flight booking progress.
• [03:48 - 04:06] Presenter selects Silver Lake Pool & Inn; dot prepares the booking and checks cancellation terms.
• [04:07 - 06:45] Presenter assigns dot a company workflow: analyzing user feedback channels, auditing gaps without solutions, and setting up an automated loop to reply to Slack threads with PR links once fixes land.
• [06:48 - 07:25] Dot summarizes top feedback categories, identifying reliability as the primary issue.
• [07:26 - 07:42] Presenter holds her phone to the camera, showing dot’s mobile call UI displaying live status ("Finding relevant repositories") and an animated avatar.
• [07:44 - 08:15] Presenter tasks dot with identifying unanswered go-to-market Slack messages and drafting responses.
• [08:42 - 09:07] Screen recording showing dot's written breakdown in ChatGPT (131 of 386 signals attributed to reliability) alongside the live booking checkout on dot's virtual browser ($351.48 total), where the presenter prompts dot to run the feedback flow again.

Claims & numbers

• Dot identifies a Southwest flight at 11:30 AM landing at 12:55 PM (refundable, ~$326) and an earlier option landing at 11:20 AM.
• Dot estimates 30–45 minutes to get out of LAX and 45–75 minutes for the drive to Hollywood.
• Dot reports that reliability is the largest feedback category, accounting for 131 out of 386 feedback signals over a nearly 24-hour review window (6:25 PM yesterday to 6:02 PM today, excluding the last 22 minutes).
• The checkout screen for Silver Lake Pool & Inn on dot's computer displays a total of $351.48.

Notable quotes

• [00:24] "So we're just going to show you it here, unfettered, unfiltered, how I use my dot."
• [02:36] "I'd roughly budget 30 to 45 minutes to get out of LAX, and then maybe 45 to 75 for the drive."
• [06:58] "The clearest issue looks like reliability, when a dot looks active but doesn't reply or can't access tools."

Assessment
This is an official OpenAI demonstration presented as an authentic, single-take makeup demo following a botched live presentation at DevDay 2026. The demo shows real-time speech duplexing, browser automation on a cloud instance ("Your dot's computer"), and simultaneous multi-tasking across travel booking and workspace triage without apparent cuts in the voice conversation.

Described by gemini-3.8-flash on 2026-10-01 from the video's audio and frames.