The dots demo, take two | OpenAI DevDay 2026

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: The dots demo, take two | OpenAI DevDay 2026 · Channel: OpenAI · 08:49 · Watch on YouTube

Topics: AI agents, Task and workflow automation, AI evaluation and benchmarks, AI developer tools

Summary

The presenter replays a demo after the previous day’s launch demo did not go as planned, showing how they use Dot to handle real work. Dot helps arrange a last-minute LA trip, including a refundable, in-policy flight and hotel options with photos and map screenshots. The conversation then turns to manually capturing user feedback, identifying major issues, and linking fixes to follow-up replies and retests. Dot identifies reliability as a leading concern and is also asked to draft responses to unanswered Slack messages about go-to-market questions.

Key points

The demo is presented as a practical example of using Dot for real work.
The presenter asks Dot to find a company-policy-compliant, refundable flight to LA and hotel options in East LA.
Dot checks whether the flight arrival time allows enough time to reach East Hollywood by 3 p.m.
The presenter selects Silver Lake Pool and Inn, and Dot prepares to verify the booking price and cancellation terms.
The current feedback workflow is manual: messages are added to a master list, classified, grouped, and linked to verified Linear tickets.
The feedback dashboard summarizes the workflow but does not automatically capture new messages.
The team wants to prioritize unresolved feedback themes and close the loop by replying with PR links and asking users to retest.
Reliability is identified as a prominent recent issue, including missing replies, tool-access failures, and unclear progress.
The presenter also asks Dot to find unanswered Slack messages and draft responses about the relationship between Dots, ChatGPT, Slack, and Teams.

Chapters

Start	Chapter	Summary
00:00	Replaying the demo and planning an LA trip	The presenter introduces the replay and asks Dot to arrange a last-minute LA trip with a suitable flight and hotel.
03:05	Choosing a hotel and reviewing the feedback workflow	The presenter selects Silver Lake Pool and Inn, then asks Dot to analyze user feedback and explain how it is currently captured.
05:42	Closing the loop on user feedback	They discuss prioritizing unresolved issues, linking fixes to PRs, and asking users to retest, with reliability emerging as a key concern.
07:17	Slack responses and demo wrap-up	The presenter asks Dot to draft responses to unanswered Slack messages and summarizes the feedback themes Dot is tracking.

Quotes

the whole point of the demo was to show you how Dots can be used for real work, real launches, real stuff. — 00:00

Can you get me a flight that’s like on policy and everything for the company? — 01:02

It doesn’t auto capture new messages. — 05:11

The clearest issue looks like reliability. — 06:46

Run the feedback flow again. — 08:49

Concepts and entities

Dots, Slack, Linear, PR, ChatGPT, Teams, Google Maps, Southwest

Full transcript (original language)

00:00 Yesterday’s launch demo didn’t go as planned. Um, a bit unfortunate because I was really really looking forward to showing you all what Dots are capable of. So, today we’re going to play it back. And uh the whole point of the demo was to show you how Dots can be used for real work, real launches, real stuff. Dots actually did launch, and I actually do have a lot of stuff to do. So, we’re just going to show you it here, unfettered, unfiltered, how I use my Dot. Starting with uh

00:31 the phone call, the the the infamous phone call to Dot. Um, I need to go to LA tomorrow. It’s kind of a last-minute thing. I’m going to see if my Dot can just like unblock a bunch of stuff for me cuz I’m a bit stressed out about it. >> Hey, good to hear from you. What’s up? >> Hi. Okay, so tomorrow I have an important meeting in LA. I need to be in Hollywood around like 3:00 p.m. Can you figure out when I think the latest I can

01:02 leave the office is probably 10:00 a.m. Can you get me a flight that’s like on policy and everything for the company? And then um also some It’s got to be refundable. And also some hotel options in like East LA. And I want to see screenshots of them so I can choose. And also screenshots of like where they are on like Google Maps, you know? So, I can see if I can hang out with anybody while I’m there. >> Okay. Yeah, let me line that up for you.

01:33 >> Awesome. Thank you. >> Yeah, and I’ll make sure the flight is both in policy and refundable. >> Okay, cool. Um what are the top options looking like right now? >> Pulling that together now. So, the leading flight I saw earlier was Southwest at 11:30 landing at 12:55 refundable around 326 I still need to

02:03 verify the checkout total and policy. >> Okay. >> Silver Lake Pool and Inn looks like the best Aerowan set and Cara is the other front runner closer to Hollywood. >> Okay, if I land at 1:00, do you think I can get to East Hollywood by 3:00? I don’t know what LA traffic is like. >> Let me check that. >> Okay, and send me screenshots of those hotels. I want to look at pictures. >> Probably if the flight’s on time and you

02:34 head straight over. I’d roughly budget 30 to 45 minutes to get out of LAX and then maybe 45 to 75 for the drive and regular ride shares mean getting to LAX it. I’d feel better about it if you landed closer to noon. >> Okay, yeah. >> I’m grabbing those screenshots. >> Yeah, yeah, yeah. Let’s get me in closer to noon then. >> Okay, I’ll recheck with that in mind. >> All right, one of my favorite things to

03:05 do just while I’m muted here is [snorts] to take a look at uh my dots browser. >> Also that earlier Southwest option I’d seen landed at 11:20 which would give you a nicer buffer. I’m checking if it’s still available with an in policy refundable fare. And I’ll send the hotel photos and maps in chat so you can look at both. >> Woah, okay. Cool.

03:37 Um Silver Lake Pool and Inn looks fun. I like Cara, too. Am I muted? Sorry. Okay, let’s just do Silver Lake pulling in. >> Okay, booking on to Silver Lake. I’ll prepare a one night booking and verify the full price and cancellation terms. >> Okay, great. Thank you. Um, hey, while you’re doing that

04:09 some other stuff. Um, our user feedback channel just has so much stuff going on in it and I know uh I think I would feel better if we just had a more kind of concrete workflow to capture all of that. Um, I think the biggest things I just need to know are like what are the biggest themes of feedback and do we have like solutions to the most pressing issues and if not, what are the outstanding things that maybe none of us are paying enough

04:40 attention to? Can you maybe like pull together that analysis? >> Yeah, I’m on it. The last dashboard we had was based on monthly. So, I’ll pull together the main themes, fixes, and gaps and make the capture workflow concrete. >> Okay, what is our capture workflow right now? >> Right, let’s make that explicit. >> What is the capture workflow? >> Checking.

05:11 Right now, it’s a manual new channel batch. We read the channel reports and replies, put each item in the master list, classify and group it, and then link actual issues to verified linear tickets. We also match tickets to PRs and track fixes in flight, shipped, or retested. The dashboard summarizes that work. It doesn’t auto capture new messages. >> That’s fine. >> is still incomplete. >> Cool, yeah. I I think it’s fine for us

05:42 to run the feedback manually. Um, I guess the bigger thing I want to make sure of is that we understand what kind of large issues that we don’t have solutions for right now. And then that be when a PR lands and we expect it to have addressed some big theme of user feedback that we actually go into Slack and have you respond and link out to the PR and like you know, tell the user to try again. >> Got it. That gap and reply loop is clear. I’ll search for biggest issues

06:14 with a credible fix and then close the loop with PR link replies asking people to retry. >> Awesome. Okay, and then let’s just um >> Yeah. >> What’s the >> And the retest confirms the retest Go ahead. >> No, you go. >> The retest is what confirms we actually close the loop. That’s the part that’s been inconsistent. I’ll make the no fix themes a priority and follow up in threads with the PR link asking folks to try again once it’s rolled out to them.

06:46 >> Okay, yeah, that makes sense. Um What’s maybe the biggest theme right now? Just in the last 24 hours. >> The clearest issue looks like reliability. When a dot looks active but doesn’t reply or can’t access tools. I’m checking the full ranking before I call that the largest by volume. My strongest read so far is people can’t tell whether their dot is actually working. That includes missing replies,

07:17 tool access failures, and unclear progress. Some real, some visibility. >> Okay, actually Drew, can you focus in on this for a second? I just want folks to see the really cute dot animation up here as it’s searching for repositories. Very cute. Okay. Awesome. Yeah, that makes sense. And then, hey, lastly,

07:47 I know there’s just been a ton of go-to-market movement. Folks are wondering about the relationship between dots and some of the other things we launched like ChatGPT and Slack and Teams. Um can you maybe pull together the top unanswered Slack messages I have and just maybe draft a response for me? >> Let me get that together for you. >> Okay. Thank you. Um

08:18 that was actually really awesome. It’s still booking my flight. It found the hotel. I like these pictures. I picked one of them. And it, you know, knows about the feedback flow that we are implementing and it’s going to go you know, prioritize my latest um things that I need to go work on. It’s messaging me now about what the top categories of feedback are. Looks like it is reliability, which we’re working on.

08:49 Um and some like UX polish that we need to do around mobile and notifications. And then, you know what? I’m actually just going to say like, “Run the feedback flow again.” And we’ll see.