Can AI Build an Entire Business Without Your Help?
The Daily AI Show
October 16, 2026
The episode opened with OpenAI's release of GPT-Live 1.1, featuring new voices, improved accents and more natural conversational responses.
Speakers Beth Lyons, Andy Halliday, Karl Yeh, Gareth Hood
TopicsTechnology
Beth Lyons (0:00)
Hey, everybody, welcome, welcome. It is, today is October 16th. We are recording this at 10 a.m. Eastern Time, 7 a.m. on Andy's Coast. And I'm saying that because if something drops today, and you're watching it, and it's after 11 a.m. our time, that's why we didn't talk about it.
It is episode 835, 835 And I mentioned, this is The Daily AI Show.
And I'm Beth Lyons. With me in the studio today is Andy Halliday. So this is another of the, Andy and Beth have talked with our tech nerd friends in the chat and whoever else drops in. So let's do that.
Andy Halliday (0:52)
Yeah, I enjoyed it. And every time this happens, we say, or I say, let's make it a 30 minute show. Let's just knock it out and just be done in 30 minutes, which never happens. It hasn't happened ever in the history.
Beth Lyons (1:09)
So which, yeah, because there's so much great stuff to talk about. Agus Cupid, very nice to see you with the chat. Thank you very much. And Jen, yes. Hey, everybody. Happy Friday.
All right. So Fable did not get dropped yesterday, but Meta is dropping some open weight stuff. So I feel like yesterday, things that happened yesterday were a little more like, hey, this is the universe we're playing in, who's adding functions to that universe or something. Anything catch your eye?
Andy Halliday (1:47)
Well, speaking of drops, GPT-Live, which is, in my experience anyway, the very best voice API and voice system that's being used out there. I haven't tested them all, but it's the one that I turn to repeatedly just to get a really natural conversation that's completely informative and easy to digest. I haven't screamed at it once.
But anyway, that was GPT-Live 1 that I was talking to. Well, yesterday, they dropped to GPT-Live 1.1, and that provides new voices. It's got a mini version called Quip. That's a smaller model that you can use for voice applications.
The release says that it provides more natural accents. I chose for the voice that I use with ChatGPT application on my phone, which is the primary place that I interface with ChatGPT voice. I chose a lovely and reserved British female voice, which is very nice. So that's my girl.
Beth Lyons (3:00)
I think that's Brian's girl too.
So now, we need to find Carl and Gareth to see if they're also speaking to this one British lady. I think currently mine is the Australian lady, and what I was most excited about is getting rid of the annoying filler words, which was also a description from this drop. Like I'm really over the, hmm, well, let's explore that.
Andy Halliday (3:38)
Mine says, let me check that. Checking. And I like that there's this interjection, while the thinking and or reasoning steps of inference and agentic pursuit of my objective are underway. I like that it keeps me abreast of that.
I didn't mind that part. That seems kind of natural. But they did say in this release that they're going to use a wider range of phrases in that because those phrases that became very predictable, and were in a limited range. So now you're going to have them interjecting and humming and hawing in lots of different ways apparently. I'll report back on how my girl, you know, my girl worked out with the new 1.1 version.
Beth Lyons (4:32)
So we were having this conversation offline, meaning off-live the other day. And what I was saying was that I have a frustration with the real live voice in part because I tend to do a brain dump, right? So WhisperFlow is much more my thing. And I will demo later in the show what I've set up for WhisperFlow.
But I'm interested in what are the, what is the parameter of the conversation? Do you have specific things that you're working out? Are you saying, like, yeah, what are you, how are you using the ChatGBT in conversation?
Andy Halliday (5:22)
So, first, let me just say I, for the work that I do with Claude Code and Codecs, I use WhisperFlow because, like you, I want to be able to try to get my way through each of the points that I want to have included in that. And then I want a moment to look at it and say, okay, well, that didn't transcribe exactly the way I like. Or by adding this little phrase in here, I can then type that in or speak it in.
In order to kind of correct that. So I do have an editorial approach to my messaging into that coding environment. But when I'm just out and about, and one of those moments occurs where you say, hey, we should check that, there's an open question, what is the answer to that? Then that's where I'd say 80% of the time I go to ChatGPT voice, just click on that little voice icon there, little blue globe comes up on your phone, and you start talking.
46 more minutes of transcript below
Thousands of transcripts fetched by people building searchable podcast archives
Fetch the whole transcript
The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000796733056