The Vergecast - Dots are OpenAI's cute lil guys

Episode Date: September 30, 2026

Jolly's got new competition: OpenAI just launched its competitor to Muse and Grok Bot. It's called Dots, and the headline feature is that it's also cute, with fuzzy lil triangles and clouds and hearts that you can give access to your personal accounts to do things for you. The Verge's senior AI reporter Hayden Field joins guest host David Imel from San Francisco right after OpenAI's DevDay presentation. Further reading: OpenAI DevDay 2026: The biggest news and announcements The Vergecast has been nominated for a Signal Award! You can vote to help us win here. Subscribe to The Verge for unlimited access to theverge.com, subscriber-exclusive newsletters, and our ad-free podcast feed. We love hearing from you! Email your questions and thoughts to [e-mail] or call us at 866-VERGE11. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Support for the show comes from Sonos. Most of us here at The Verge are big movie buffs. When watching at home, I don't want that experience to be any less immersive than it would be in a theater. And sound plays a huge role in that. The Verge has actually published a few pieces over the years about how much sound matters to a movie. From headphones to sound bars, a Sonos system pulls you in. So far, you might actually think the phone call is coming from inside the house. Visit sonos.com and build your system.

Sonos, love the sound, feel the system. Visit BetMGM Casino and check out the newest exclusive. The Price is Right Fortune Pick. BetMDM and GameSense remind you to play responsibly. 19 plus to wager. Ontario only. Please play responsibly. If you have questions or concerns about your gambling or someone close to you,

please contact connects Ontario at 1-866-531-2,600, to speak to an advisor. Free of charge. BetMGM operates pursuant to an operating agreement with I. Gaming, Ontario. I'm Ina Garten, and each week on Happy Hour, I invite old and new friends for a drink and a fun conversation at my kitchen table in New York City. On this week's episode, I'm joined by Andy Baragani. We talk all about how he started his career as a teenager at Chez Pinesse to becoming an

internet heartthrob from his cooking videos to opening his first restaurant in New York City this fall. New episodes of Happy Hour are available every Wednesday. You can watch on YouTube or listen wherever you get your podcasts. Cheers. Welcome to the Vergecast, the flagship podcast of cute little AI agents that are definitely not dots. They're like different shapes. I'm David Amel.

Yesterday, OpenAI hosted its annual Dev Day. It's its big annual conference where it shows what's on the frontier of what it's developing. This year, there's a lot of companies that are all rushing to make vaguely the same thing. And the biggest news this year was OpenAI's new. dots. Dots are little agents that will go do things for you, but with one big quirk, they're really cute. I brought on senior AI reporter Hayden Field to talk about it. But first, here's what's happening on the verge today. I'm Travis Larchuk, and this is 90 seconds on the verge for Wednesday, September 30th, 2026. We didn't see any smart home stuff at Apple's September event, but,

according to Bloomberg's Mark German, we're about to on October 13th. He says, That's when Apple will debut its long-rumored smart home hub, which will maybe be called the Homepad or something else. Who knows? We should also expect a new HomePod Mini and an upgraded Apple TV, according to German. Apple Home users have been waiting on these updates for literally years. So what's another Fortnite to wait? Next up, Instagram announced a new AI Creative Assistant that will give you feedback on your posts and suggestions on how to increase engagement. YouTube announced a similar feature recently, and I know that this feels icky to some people,

but personally, I'm interested. If you make content for social media, you know that sometimes it's really hard to figure out why one thing hit and another thing didn't. It feels arbitrary a lot of the time. And so a tool that can give you suggestions that are data-based rather than vibes-based actually feels useful to me. And finally, the most important news of the day. The Virges Nathan Edwards reports the dads in his test.

Texas neighborhood are sending each other single potatoes by drone. They're ordering them from Walmart using wing drone delivery. The dads reportedly discovered that they can have the potatoes deliver to anyone's yard, not just their own yards. And the best part is that each potato delivery only costs 82 cents. Total. At that price, you can't afford not to send your neighbor a potato. You can read more at the verge.com.

That's 90 seconds on the verge for Wednesday. September 30th, 2026. Visit BetMGM Casino and check out the newest exclusive. The Price is Right Fortune Pick. BetMDM and GameSense remind you to play responsibly. 19 plus to wager. Ontario only. Please play responsibly.

If you have questions or concerns about your gambling or someone close to you, please contact connects Ontario at 1-866-531-2600 to speak to an advisor. Free of charge. BetMGM operates pursuant to an operating agreement with Eye Gaming Ontario. Support for the show comes from Sonos. Think about the first time you heard your favorite song. Who were you with? What did it make you feel?

Was it the music or the lyrics that grabbed you first? And what were the specific tech specs of the sound system it was playing on? Yeah, nobody remembers that last part. Sonos gets that. Their system is designed to just be there. Clear, ready, out of the way. So nothing stands between you and the next song that hits you out of left field. for a bunch of years now, I've been incredibly jealous of my friends who have Sonos speakers.

I was just always impressed how all that sound and punch and energy came from that speaker in the corner. But Sonos recently set me two of its play speakers, and now I truly, truly get it. Everyone knows that the quality is immediately noticeable when you listen to a sono speaker. But the bigger thing that I've learned is how everything is working together and creating this little ecosystem. It's just really important to me, especially as audio has become more interesting. integrated into our daily media lives, whether that's the music, a movie, and yes, the Vergecast or version history or decoder, I want it to sound as good as possible. Visit sonos.com and build your system. Sonos, love the sound, feel the system. Hayden, thanks for joining us.

Thanks for having me all the way from SF. Oh yeah. So you just got out of the developer conference. This just happened. Right. Number one thing that the company announced was this AI dots feature. Can you explain exactly what OpenAI, chatGBTD dots? What is the official name for this exactly? So OpenAI's dots is basically the company's answer to Meta's Muse and OpenClaw. It's the long-anticipated continuously running AI agent from OpenAI, the consumer-facing one. You know, this is something that we wrote a couple of days ago that Open AI really needed to catch up on. It was one of the areas that everyone was looking at the company saying, okay, well, It's strange that they're behind on this. That was kind of the consensus that Meta had suddenly

pulled ahead and it was unexpected and that, you know, Open AI was working on something rumored to be almost the same. Now they've announced it. And also, they made it really cute, similar to Meta's muse. They're basically cutesy little blobs with eyes and you can customize them however you want. So I think that's part of how OpenAI is banking on people, you know, feeling maybe better about using these agents. Same with all the other companies. It's like make them cute, make them customizable. And, you know, they're hoping people will be okay with giving them a lot of data. So Open AI, yeah, I mean, like, you know, one of the ones that they demoed was named Alfred.

And it's, you know, you can name it whatever you want. And then, you know, it immediately transformed into like a butler-looking blob with like, you know, glasses and a mustache. I think they had a lot of different little blobs on display. But yeah, I think, you know, there's a lot of customizable options. You can change the color. You can change like the eyewear hat maybe. And the name. So I think they're really, you know, hoping that people kind of go ham on the personalization here.

The most interesting part of it, though, to me is that although this is directly, you know, aimed at, you know, kind of gaining users away from Meta's Muse, it's only rolling. out to paid users, pro and business users. So it's not free for everyone like Mews is. And so I think that's kind of an interesting hallmark here, the fact that, you know, Open A High desperately wants to compete with meta on this, but they can't afford to fully because it's such a compute-intensive product. Yeah. Besides having an embodiment of this agent, the primary question that I have is how exactly is this different from just a regular chat GPT chat that can do things for you?

You know, if you're using Claude, co-work can go control your computer. What is it about these agents that are uniquely different from just using the regular service? Essentially, there's a couple different ways. They can go off and do things for you in the background continuously even without you having the window open.

So that's a big selling point for people. You know, they don't have to, you know, keep their computer open and stay, stay awake or whatever. They can just go off and live their life and the agent is doing stuff in the background. Another part of it is that, of course, you know, you're supposed to be able to text with them, maybe phone call with them, talk to them in a way that, you know, is really accessible for people beyond just like a chat GPT chat window. However, Open AI was even behind in that because they said the texting and calling function

will be coming soon, not right now. So that's, a lot of this is, I think, front-loading something so that they have a horse in the race. But, you know, there are a couple features that they haven't yet rolled out. So Open AI noted that they have over 4,000 integrations for these dots. For people that don't know, what is an integration? Yeah, basically an integration is like the way that these agents are useful at all. Because, you know, you can use an agent in a vacuum, but, okay, that's not that helpful. that they become useful as using them for your life and with the other tools that you use.

Think about if you had a personal assistant and then they didn't have any of your logins. You know, I was a personal assistant back in the day and I needed my boss's Sky Miles number. I needed her calendar access. I needed her like email login, basically everything for her whole life. So same thing with these agents, you know, integrating with different platforms you use, different tools, different accounts. And yeah, with Open AI, it's interesting because they don't naturally have a bunch of this stuff in-house. So their integrations may work a little worse with certain products than, you know, say Google's AI agent,

it's going to have better integrations with Google products. Meta's AI agent is going to have better integrations with meta products. So, you know, Open AI is really gunning for the, oh, the quantity, you know, of integrations maybe over some of the quality. Right. Because they don't have anything in house, but they're like, well, you know, we can connect to everything. It's just, it's going to be, they basically are saying it's going to be good, but it might not be as good as some of these other companies in-house AI agents. Right. I also heard something about the specialist dots that they're announced.

Can you explain what those are? They're launching these specialist AI agents that are aimed at being really good at certain niche enterprise functions. So, you know, one for legal. one for marketing, one for accounting, things like that. And they launch with a bunch of partners to start, like Harvey, Lagora, a bunch of different companies that are in these different fields. So it'll be interesting to see how those play out. I don't know if they're live yet, but they will be soon.

And so, yeah, they're supposed to be, you know, agents that are just, you know, really good in a certain field. And, you know, they can kind of be a coworker in that sense where, you know, they're senior and they've been in the field for a long time is kind of the way they're positioning it. Right, because Muse is sort of around for listing things for you on Facebook Marketplace, which we've had some problems with so far already.

It's supposed to do things for you in your personal life. A lot of people use ChatGPT, and a lot of people probably have pro. People that have pro are, you know, it's $20 a month and they can access these dots. Do you think that they're going to see a lot of interesting use cases from this, or do you think that it's primarily targeted at work?

I think that we are going to see people interested and trying it out, but I don't think it's going to be a general purpose as Meta's Muse, because with Muse, it's such an easy access situation. You know, it's a one-tap, and you're good. With Dots, not only do you need to be a pro user of chat chvety, but also, you know, there's a little bit more friction in terms of integration. So I think with meta, a lot of this stuff just automatically seamlessly integrates into all meta products. And same with Google.

You know, they have a bunch of existing in-house stuff. So I think that with Open AI, there's a little bit more friction there. They did try to talk a lot about how they're overcoming that friction and how, you know, there's a lot of different, you know, sign-ins and, you know, apps that you can use with these things. But I still think there is more friction to adopting this. And people are already a little bit uncomfortable with AI agents to begin with, especially. you know, with the AI populism rise and also the news of how AI agents can be pretty non-secure with handling your data and your stuff, not only muse with Facebook marketplace and those missteps, but also back when OpenClaught was especially viral, you know,

all the security issues that happened there, deleting emails, things like that. So, you know, it'll be interesting to see how people, how many people choose to trust Open AI with a lot of this data. They did make a big point to talk about privacy and data privacy in particular on stage today. So, you know, they did anticipate, I think, some of that pushback. But, yeah, we'll see. I mean, right now they did announce today that 1.2 billion people use chat GPT. So it'll be interesting to see how much of a chunk of those people are going to try dots. Support for the show comes from Idira by Palo Alto Networks.

Here's this statistic for you. Nine out of ten organizations suffered an identity breach last year. That happens because adversaries are logging in with valid credentials, not breaking in, and non-human identities outnumber human identities 109 to 1. That creates a massive, unmanaged attack surface and cloud environments. The problem is that legacy tools were built to manage human employees, leaving modern machine and AI access largely unmanaged. That's where IDERA by Palo Alto Networks comes in.

It's one identity platform for human, machine, and AI. IDira replaces permanent permissions with dynamic access so you can lock down every identity without slowing down your business. Secure every identity at your organization with IDA by Palo Alto Networks. Learn more at palo Alto Networks.com slash IDIRA. Again, that's Palo Alto Networks.com slash IDIRA. Visit BetMGM Casino and check out the newest exclusive.

The Price is Right Fortune Pick. BetMGM and GameSense remind you to play responsibly. 19 plus to wager. Ontario only. Please play responsibly. If you have questions or concerns about your gambling or someone close to you, please contact connects Ontario at 1-866-531-2,600 to speak to an advisor. Free of charge.

BetMGM operates pursuant to an operating agreement with Eye Gaming Ontario. As you were alluding to, the Metamuse agent has been under a little bit of fire recently for some mishaps that it's done agentically, of course. You know, we've seen these agents do things to governments. Now we're seeing them do things to regular people. One of them was that someone was selling something on Facebook Marketplace. And without even telling the user, it told someone his address. It told him that he was there and ready to come down and give him and sell the item

when he just wasn't there. All of this was automated. There were some problems with this. Did Open AI address any of these issues or maybe some potential issues that these agentic objects are going to start doing? They talked about the fact that, you know, this is going to be really secure,

really private, really trustworthy. But I was chatting with Glenn Codes, the head of app platform at OpenAI in a one-on-one interview today. And he said that he thinks they're in a different position to meta in that they, since they have 1.2 billion users, they want to launch something that doesn't make those types of mistakes. And they're trying to really be deliberate with how they let agents use plugins and how they allow agents to use people's identity and sign in and sign up for things, how they let agents pay for things. And they're also working with third party standards

bodies about how agents will attest that they're an agent when they're doing things, like how they'll say, hey, I'm an agent on behalf of, you know, XYZ. So he said we might move a tiny bit slower, but basically, you know, they don't want to have agents that are yolo buying stuff on Facebook marketplace for, you know, 1.2 billion users. So, you know, essentially all that translates to we're moving a little slower than meta, but we hope it's safer. And, you know, the jury's still on if that is the case, we'll have to test it. Right. How important do you think it is for Anthropic or Google to develop something of a similar caliber, right? We've seen Grockbot. Notably looks really similar to these little

muse agents. We've seen Jack Dorsey's project, which has little guys that kind of look similar to this. There's a big push in Silicon Valley right now to make agents look really cutesy and have these little agents you can interact with every day. Do you think it's important for Anthropic or Google to make something similar like this? Yeah, I mean, I think it's incredibly important. This is something that I have been seeing AI companies talk about for years. You know, I think 2023 was when AI agents had their chat, GPT moment in terms of them getting really buzzy and, you know, executives talking about them nonstop on earnings calls, really promising investors that this is the next big thing, the thing that's finally going to, you know, break through

and make a ton of money. So Open AI's answer to that is, okay, we're only getting able to pay to users and we're going to go really hard on enterprise. Anthropic, you know, we'll see. I'm sure it'll be a similar, similar offering. Right now you can only have one opening I dot per account, but they said that more are coming in the future. Do you think that that's a limitation in any way, or do you think that opening eye is just worried

about people spending up a thousand of these agents and kind of running them? Because notably, it does not count against your chat GPT usage limits, which seems like a big deal. So do you think that that's the reason they're limiting it for now? Yeah, they said that it was a really compute-heavy product during the press Q&A. Sam Allman said, you know, someday we'll be able to offer these for free to everyone, you know, or we hope so, but right now, you know, it takes a lot of compute.

And so I think that's the same reason why they're only offering one right now. They want to see how it goes and then maybe offer more later. But yeah, I do think it's a limitation in some ways, you know, the power users that are going to be really obsessed with this, like, you know, X open-cllaw users potentially are used having a bunch of agents working on their behalf. So with a subset of people, I think it will be a limitation for others. You know, if they're trying to make this a general purpose product that everyone wants to try, I don't think it's going to be much of a limitation there. Right. Do you think that they'll probably introduce a free version for free users eventually? They said they

hope so, but, you know, it seemed that he used the word someday. So, you know, I don't think that's happening anytime soon, at least ahead of their IPO, I would imagine. Right. Well, this is the major news from the Dev Day event, but were there any other big events or happenings at the event? I think the other big thing I was really thinking about is the protests happening outside. A lot, like more than a dozen organizations came together to kind of, you know, unite and plan this protest outside Deb day that was going on from early in the morning until about the time that Sam Altman arrived and started his keynote. And so, you know, there were art pieces outside.

There were a ton of protesters marching in a circle. There were a lot of chance. a lot of different signs. Most of them were protesting, you know, AI lab contracts with military and the government, particularly ICE. So, yeah, it was really interesting to see how many different organizations came together. It was even sponsored in part by the San Francisco Labor Council. So, yeah, I mean, it definitely made waves. Everyone saw these people when they were going inside. And so I think that is something that, you know, was definitely a topic of discussion today. another thing that I thought was interesting

was the fact that the voice function for dots was not working the entire time at 10th. They tried to demo that a couple of times it did not work. So, you know, at least we know it was a live demo, but it also showed they may have rushed this, we don't know. So, you know, probably after getting a lot of questions about where's your

answer to Muse. I mean, I'm sure, they said they've been working on this for a long time, so I'm sure they always You know, they plan to, yeah, yeah, they plan to, you know, show it off at Dev Day, but, you know, it shows the kinks aren't all worked out necessarily. Yeah. Well, Haydenfield, thank you for coming on. Appreciate it. Thanks.

That's it for the Vergecast. Let's know what you think. Call the hotline 866 Verge 1-1 or email, Vergecast at theverge.com. And as for everything we do here and to get ad-free podcasts, subscribe to the Verge. We're part of the Vox Media podcast network. The Vergecast is produced by Josh Kahas, Eric. Gometh Gomez, Brandon Kiefert, and Travis Larchuk. See you next time. Rock and roll.