# OpenAI's New Model Talks and Listens Simultaneously

An episode of [AI Chat: AI News on Claude, ChatGPT, Gemini, Grok, Perplexity](https://invisiblenetwork.ai/podcast/1669990482), 8 Jul 2026, 14 min.

Audio: https://content.rss.com/episodes/365073/2973072/ai-chat-podcast/2026_07_08_21_40_34_218a8268-03c8-45dc-8f7e-85b4ab998951.mp3

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! In this episode, we discuss OpenAI's new GPT Live 1, a groundbreaking voice model capable of listening and speaking simultaneously. Additionally, we explore the implications of recent security concerns flagged by China regarding Claude code, the advancements in open-source AI from NVIDIA, and Meta's latest developments in AI glasses. Chapters 00:00 Introduction to AI Chat Daily 00:44 OpenAI's GPT Live One Launch 06:09 China Flags Claude Code 08:14 NVIDIA's Nemotron 3 Ultra 10:00 Meta's AI Glasses Developments 11:44 OpenAI Chief Futurist Departure Show Links • Get the top 80+ AI Models for $8.99 at AI Box: https://aibox.ai • How I Grow and Scale My Business with AI: https://www.skool.com/aihustle • Get the AI Chat Daily Newsletter: https://www.aichatdaily.com/newsletter

## Transcript

[0:02] Welcome to the AI Chat Podcast. I'm your host,
[0:04] Jaden Schaefer. Every day I cover cutting -edge
[0:07] AI news and talk with the leaders behind it,
[0:09] breaking down what it means for your life and
[0:11] business. Are you subscribed to AI chat daily
[0:17] .com? Yeah, I have a newsletter that I send out
[0:20] every single day. And it's essentially all of
[0:22] the same content, all the same stories I talked
[0:24] about in the podcast, but I do them in these
[0:26] really easy to break down formats. And I also
[0:28] have links to articles to go more in depth. So
[0:30] it's kind of like show notes for the podcast,
[0:32] but it goes into way more depth. And we have
[0:34] extra stories that we can't get around to covering
[0:36] in the podcast. So if every single day you want
[0:38] to start the day out or end the day by getting
[0:40] a roundup of the latest in AI news, and a really
[0:43] easy to scan format, go go to AIChatDaily .com.
[0:46] There's a big subscribe button in the top corner.
[0:48] I'll probably leave a link in the show notes
[0:49] to the newsletter. And it also is just a full
[0:52] news site with tons of articles. I'll probably
[0:53] publish about 10 of them every single day. So
[0:55] if you want to go check that out, AIChatDaily
[0:57] .com. Okay, OpenAI is shipping GPT Live 1. So
[1:02] this is a full duplex voice model. And what duplex
[1:05] voice model means, well, this is actually kind
[1:06] of crazy because Miriam Murady, who is formerly
[1:09] at OpenAI, she released something recently with
[1:12] her most recent company. Thinking Machine Labs,
[1:14] where it's, I mean, essentially it looks like
[1:16] OpenAI has just knocked her off. So this is going
[1:19] to be interesting. Also, China has flagged Anthropix
[1:22] Cloud Code. They say that this is a security
[1:24] risk because there's some backdoor risks with
[1:26] it, which I don't know. I just think anything
[1:28] with security risk is ironic coming from China.
[1:30] NVIDIA's Nemotron 3 Ultra has just hit closed
[1:34] model parity at a 10x lower cost. And this is
[1:38] from some experiments they've been doing on Langchain.
[1:40] I'm really excited when we're able to get the
[1:41] costs down on some of these. models. Meta is
[1:44] adding a tamper -proof LED to their AI Meta glasses,
[1:48] and they're also getting more data collection
[1:50] from other parts of the glasses. So there's kind
[1:52] of a mixed bag with what's going on there. An
[1:55] AI, the OpenAI chief futurist, John HM, is leaving.
[1:59] He's been there for nine years. He's one of the
[2:01] OG trust and safety people over OpenAI. But let's
[2:04] kick the podcast off with OpenAI's latest model,
[2:06] GPT Live 1. This is a voice model. It listens
[2:10] and it speaks at the same time. is something
[2:12] that you've probably noticed a lot of these voice
[2:14] models have struggled with in the past. If you're
[2:16] talking to, you know, what voice mode is on chat
[2:19] GPT right now, which by the way, this replaces
[2:21] their advanced voice mode. But if you, if you've
[2:23] ever talked to it before you talk, there's a
[2:25] couple seconds of latency, then it responds back
[2:27] to you. And they've done a lot of work to try
[2:28] to like get that latency down as much as possible.
[2:31] But what's really cool about these new live one
[2:33] model is that it's literally listening and it's
[2:35] able to listen and talk at the same time, because
[2:37] that's how actual humans talk. When we're in
[2:39] a conversation, we kind of interrupt each other.
[2:41] We talk. over each other. They're saying something,
[2:43] you're saying something. And I mean, the goal
[2:44] is not to interrupt and talk over each other.
[2:46] But it's just part of communication, right? Like,
[2:48] we don't always wait for there to be a perfect
[2:49] pause at the end of everything that someone says
[2:51] before we jump in. So that's what these models
[2:53] are getting really good at, which is really interesting.
[2:56] And what's to me really crazy about this is that
[2:59] Miriam Marotti released a model just like this
[3:01] a number of months ago. I remember covering it
[3:03] on the podcast and it was like the big thing
[3:05] from Thinking Machine Labs. And I was like, man,
[3:06] this is a really cool idea and concept. Well,
[3:08] turns out, you know, a couple months later, anyone
[3:11] that releases any sort of cool new feature, a
[3:13] couple months later, the Frontier Labs are going
[3:15] to be able to clone it, reverse engineer and
[3:17] put it out themselves because, you know, they
[3:18] don't want to fall behind. The reason why I think
[3:20] this is one in particular is an important feature
[3:22] is because there's three separate things that
[3:25] are getting collapsed into one step. Number one,
[3:28] it's cutting the delay. So it's going to let
[3:30] your conversation be way more natural. And they're
[3:32] also going to be rolling this out to 150 million
[3:35] chat GPT voice users. And finally, the way this
[3:38] works is really interesting to me. So there's
[3:40] something called GPT live one mini, and this
[3:42] is going to be for basically all of the free
[3:44] users, it's kind of the default. But if you're
[3:46] on the paid tier, if you're paying $20 a month
[3:48] for chat GPT, you get the larger GPT live one,
[3:51] which is going to send some of the harder questions
[3:53] over to GPT. 5 .5 mid conversation. So you're
[3:56] going to be talking in the middle of the conversation.
[3:57] If you're saying something kind of complex and
[3:59] it's getting like, I mean, basically get half
[4:01] your sentence, send it over to GPT 5 .5 to try
[4:03] to come up with a response and send it back and
[4:05] put it out before you even finish your sentence,
[4:07] which is really crazy. It can be interrupted
[4:10] in the middle of a sentence. So if it's talking,
[4:12] you can interrupt it. It handles silences while
[4:14] you think. So just because, you know, you're,
[4:16] you're pausing to think about something doesn't
[4:18] mean it feels like it has to jump in, which is
[4:19] something really annoying. I feel like I can
[4:21] barely like keep my thoughts. Like I can barely
[4:23] think when I'm talking to these things because
[4:25] I got to talk so fast because if I stop talking,
[4:26] it's going to want to jump in and start talking
[4:28] to me. So I really appreciate that. And it's
[4:31] going to show charts or images alongside whatever
[4:33] you're talking to it about, which, you know,
[4:35] before it was just kind of this like animated
[4:37] strobe on your screen. But now if you're talking,
[4:40] you're like, hey, you know, what are the, you
[4:42] know, what's the weather going to be like? It's
[4:43] actually pulling up weather charts on the screen
[4:45] while it talks to you. So you get visual and
[4:47] audio, which is really cool. The product leave
[4:49] on this is or the product lead is Addy. Letty.
[4:52] He said that he held a 30 to 40 minute conversation
[4:56] with the model on walks. And I think the goal
[4:59] and probably the reason why they're telling us
[5:01] this is because they're trying to make this something
[5:02] for or trying to explain this is really good
[5:05] for long running tasks. This isn't just quick
[5:07] commands like, hey, what's the weather like?
[5:08] It's like, hey, I'm trying to break down this
[5:10] complex business plan that I want to, you know,
[5:12] create. talk with me about this, or I'm trying
[5:14] to figure out what the best strategy is for XYZ,
[5:16] talk to me about this. And you go and have a
[5:19] 40 minute conversation to learn something to
[5:21] plan something to actually get something done,
[5:23] which if I'm being honest, I was kind of moving
[5:25] away from opening eye and chat GPT for a while
[5:26] because it didn't feel like that was their focus.
[5:28] And I was going to Claude, but I've been trying
[5:30] out a bunch of things on chat GPT lately, mostly
[5:34] like whenever my whenever my tokens, my token
[5:36] limits reach on Claude, I go back to chat GPT.
[5:39] And I've been really impressed by the research
[5:41] it can do, where I'm like, hey, you know, go
[5:43] scrape like these 5000 things and do an analysis
[5:46] on these different websites or these different
[5:48] products and like, bring me back, you know, help
[5:51] me create a list of 365 XYZ things I can do for
[5:53] this, like it goes and get some real serious
[5:56] work done. So I'm really excited. I think the
[5:59] reason why this is particularly important beyond
[6:01] just humans is because we're going to have a
[6:03] lot of work with earbuds and wearables that OpenAI
[6:07] is currently working on. And so a lot of the
[6:10] hardware that they're making, these tools are
[6:13] going to be really useful for. And, you know,
[6:15] when we start talking about robots, getting them
[6:17] and all that kind of stuff and be able to communicate
[6:19] with that, this is going to be important that
[6:21] the communication is as smooth as possible. China
[6:24] is flagging Anthropics Cloud Code as a security
[6:26] backdoor risk, which is hilarious because China
[6:29] has been accused of putting backdoors into chips
[6:31] and hardware inside of China for many years on,
[6:35] you know, servers and all sorts of crazy stuff.
[6:37] But this in particular, I do think is probably
[6:39] accurate. What they've said is that Claude Code
[6:41] has a backdoor vulnerability that's going to
[6:44] leak your location and your identity data. And
[6:47] they're telling everyone in China to uninstall
[6:49] or to upgrade the version if you have something
[6:52] installed between April and June to upgrade it.
[6:54] The advisory is interesting because Claude Code
[6:57] isn't officially sold in China. So, you know,
[7:00] Alibaba has banned Anthropic Tools company -wide.
[7:03] This is also a week after Anthropic accused Alibaba
[7:06] of stealing. their AI technology. And what was
[7:09] interesting to me is there's like some, some
[7:12] moments where different people inside of I think
[7:14] Exomi, they had like an AI, AI developer conference,
[7:17] I think, and they were kind of publicly acknowledging
[7:18] that they were using cloud code at a state organized
[7:21] forum. This is back in March, even though it's,
[7:24] you know, officially not available in China.
[7:26] So it's funny, because it's like officially not
[7:28] available, officially not sold, Alibaba is banning
[7:30] it, no one's supposed to be using it. And yet
[7:32] China's like warning everybody, like they know
[7:34] they're using it. Anyways, it's kind of crazy.
[7:36] What I do think is important, though, is this
[7:39] particular flag they have is version 2 .191 to
[7:43] 2 .19656. But we're already on 2 .12 for the
[7:50] version that we're currently on. So if you're
[7:52] on the latest Anthropic thing, this isn't really
[7:54] relevant to you. Most modern coding assistants,
[7:57] so if you look at like OpenAI and Google, they're
[7:59] sending what's called... telemetry and crash
[8:02] reports back to the main servers. And anytime
[8:06] that something happens, I think this is what
[8:07] China is worried about the crash reports, they're
[8:09] worried has too much information about the users
[8:11] in it. But China wasn't really clarifying if
[8:14] quad code was the only one doing this, why they
[8:17] were singled out and not like open AI and Google.
[8:20] I'm not even sure if Google's is available in
[8:23] China. So anyways, it's China's a crazy, crazy
[8:26] place. In a big win for open source models, NVIDIA's
[8:29] open source Nemotron 3 Ultra model now matches
[8:33] the best closed AI systems on business tasks.
[8:36] So not everything, but business tasks specifically,
[8:39] and it can do them at a tenth of the cost. Langchain
[8:42] actually tuned the agent software around the
[8:45] model without retraining it. And they basically
[8:47] proved that the right engineering environment
[8:50] can compete with really expensive proprietary
[8:52] AI when it comes to doing enterprise work. So
[8:56] you can get these open source models, if you
[8:58] fine tune them, you can get them just as good
[9:00] for way cheaper than something like Anthropic
[9:02] or OpenAI. lane chain, adjusted system prompts,
[9:06] tool descriptions, and middleware. They did all
[9:08] of that to really just squeeze the top tier performance.
[9:12] Then they put it through the deep agent benchmark.
[9:15] Ambridge, Amdocs, Box, and EY are already building
[9:18] specialized agents on this exact stack for their
[9:22] platforms and their workflows. The tuned model
[9:24] is live on six hosting providers, so you can
[9:26] get it on Bastion, Crusoe Cloud, Deep Infra,
[9:29] Firework, Nebius, and together AI, which is what
[9:32] I use for AI. box. So it's, you know, pretty
[9:35] widely available. This is really exciting. And
[9:37] I believe we actually do have pneumotron over
[9:38] on AI box right now, which by the way, if you
[9:40] haven't tried AI box .ai, my own startup, we
[9:43] have over 80 different AI models all on there.
[9:46] So for $8 .99 a month, you get access to all
[9:49] of the top AI image, audio, video, and even music
[9:52] models. So if you don't want to have to have
[9:54] subscriptions to all of these different things
[9:55] and have, you know, logins on a ton of different
[9:57] platforms, just go to one place, AI box .ai.
[10:00] We even have an MCP that will tie all of those
[10:02] models. into things like clod so if you want
[10:04] to get like image and audio and video inside
[10:06] of clod you can use our mcp $8 .99 a month i'll
[10:09] leave a link in the description to that as well
[10:11] Meta just added a tamper -proof LED to their
[10:14] AI glasses, and it disables the camera if the
[10:16] light is modified or destroyed. So the idea being,
[10:20] you know, you want to know when someone's recording
[10:22] you with the Meta Ray -Bans or any of these other
[10:24] Meta AI smart glasses. A lot of people have been
[10:26] complaining about, you know, non -consensual
[10:28] recording. People are recording them when they
[10:30] don't know or they don't consent. People are
[10:31] in the gym recording, whatever. And so they've
[10:34] added a light, which I think a lot of devices
[10:36] have, like a light that pops on when it's recording
[10:38] so you know someone's recording. Theoretically,
[10:40] someone could go with like a pin and just like,
[10:42] you know, kind of shatter the LED bulb or wreck
[10:45] it, right? So it doesn't turn on or cover it
[10:47] or do something. So they're making it so that
[10:49] you can't tamper with it. If it's broken, it
[10:51] won't actually record. And this is, I mean, I
[10:54] think a lot of people are really excited about
[10:55] that. But this is also happening the same week
[10:57] that Meta is expanding the data collection elsewhere.
[11:00] So they're going to be allowing their AI to train
[11:02] on public Instagram photos by default. And they're
[11:05] testing glasses that continuously collect audio
[11:08] and they snap photos. every few seconds, which
[11:10] come on, that's terrifying. But I mean, maybe
[11:14] like they're milled in a product that they think
[11:16] some people will like that. I really doubt that.
[11:18] But this particular prototype that is under testing
[11:20] is going to move the glasses from being an on
[11:23] -demand camera, right? So you're like, hey, Meta,
[11:25] record whatever I'm doing. And it turns on to
[11:26] actually just being always on and having an always
[11:28] -on sensor. And it's feeding continuous audio
[11:30] and periodic photos to Meta's AI systems. They
[11:34] don't have a date launch for this. I hope there'll
[11:36] be a way to opt out. Otherwise, I don't imagine
[11:38] these glasses will continue to be very popular.
[11:41] For the Instagram thing, this is a change that
[11:43] will affect... about a billion that, you know,
[11:45] billions of photos for users. Most like everyone
[11:48] posting is not wanting their photos to be trained,
[11:51] like AI to be trained on them. So that's going
[11:53] to be kind of crazy. OpenAI's chief futurist,
[11:56] Joshua Achim, is leaving after about nine years.
[12:00] He's one of the safety advocates that's been
[12:03] at the company the longest since the early days.
[12:05] He is leaving, and he's the fifth senior safety
[12:08] -focused leader to leave OpenAI in the last two
[12:11] years, right? They're getting ready to go public.
[12:13] He actually joined as an intern back in 2017,
[12:15] and then he led their mission alignment team
[12:17] until it was disabled, or I guess like disbanded.
[12:21] Sam Altman got rid of it back in February of...
[12:24] this year. Then he moved into chief futurist
[12:27] role and was kind of studying AI harms and benefits.
[12:30] Five different safety -adjacent leaders have
[12:32] now left OpenAI since 2024, and a lot of them
[12:36] have started non -profits focused on AI safety
[12:38] standards. The former White House AI advisor
[12:40] Dean Ball started this week as the head of strategic
[12:44] futures and is going to briefly overlap with
[12:46] Etcham before he actually leaves and will kind
[12:49] of take over a lot of policy -focused research
[12:50] work without having a direct replacement for
[12:53] the chief futurist title. So if you believe you'd
[12:56] be perfect for that job, maybe you should send
[12:57] in your application to OpenAI. Guys, thank you
[13:00] so much for tuning into the podcast today. Make
[13:01] sure to check out AIbox .ai. Links in the description.
[13:04] If you have not already left a review for the
[13:06] show, it would mean the world to me. It helps
[13:08] the show out tremendously. Add some exciting
[13:10] interviews and content and stuff planned for
[13:12] the show this year. And if you would be able
[13:14] to leave a review, it would help me to bring
[13:16] you more incredible content. Over on Spotify,
[13:19] you have to hit the About tab. If you've listened
[13:21] to three episodes, you're eligible to leave a
[13:22] review. Otherwise, I won't let you. So if you
[13:24] have, that would be incredible. Hit the About
[13:27] tab. Hit some stars. stars on Apple. It's kind
[13:30] of annoying to find, but you go to the podcast
[13:32] page and you can hit the stars and leave a comment.
[13:35] I read them all. I appreciate them. Alright guys,
[13:37] I hope you all have a fantastic rest of your
[13:39] day.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.