Menu
Sign In Search Podcasts Charts People & Topics Add Podcast API Pricing
Podcast Image

The Daily AI Show

Google Undercuts the Field, OpenAI Builds an App OS, and China Accelerates

18 Dec 2025

56 min duration
7818 words
4 speakers
18 Dec 2025
Description

The conversation centered on Google’s surprise rollout of Gemini 3 Flash, its implications for model economics, and what it signals about the next phase of AI competition. From there, the discussion expanded into AI literacy and public readiness, deepfakes and misinformation, OpenAI’s emerging app marketplace vision, Fiji Simo’s push toward dynamic AI interfaces, rising valuations and compute partnerships, DeepMind’s new Mixture of Recursions research, and a long, candid debate about China’s momentum in AI versus Western resistance, regulation, and public sentiment.Key Points DiscussedGoogle makes Gemini 3 Flash the default model across its platformGemini 3 Flash matches GPT 5.2 on key benchmarks at a fraction of the costFlash dramatically outperforms on speed, shifting the cost performance equationSubtle quality differences matter mainly to power users, not most peoplePublic AI literacy lags behind real world AI capability growthDeepfakes and AI generated misinformation expected to spike in 2026OpenAI opens its app marketplace to third party developersShift from standalone AI apps to “apps inside the AI”Fiji Simo outlines ChatGPT’s future as a dynamic, generative UIAI tools should appear automatically inside workflows, not as manual integrationsAmazon rumored to invest 10B in OpenAI tied to Tranium chipsOpenAI valuation rumors rise toward 750B and possibly 1TDeepMind introduces Mixture of Recursions for adaptive token level reasoningModel efficiency and cost reduction emerge as primary research focusHuawei launches a new foundation model unit, intensifying China competitionDebate over China’s AI momentum versus Western resistance and regulationCultural tradeoffs between privacy, convenience, and AI adoption highlightedTimestamps and Topics00:00:00 👋 Opening, host setup, day’s focus00:02:10 ⚡ Gemini 3 Flash rollout and pricing breakdown00:07:40 📊 Benchmark comparisons vs GPT 5.2 and Gemini Pro00:12:30 ⏱️ Speed differences and real world usability00:18:00 🧠 Power users vs mainstream AI usage00:22:10 ⚠️ AI readiness, misinformation, and deepfake risk00:28:30 🧰 OpenAI marketplace and developer submissions00:35:20 🖼️ Photoshop and Canva inside ChatGPT discussion00:42:10 🧭 Fiji Simo and ChatGPT as a dynamic OS00:48:40 ☁️ Amazon, Tranium, and OpenAI compute economics00:54:30 💰 Valuation speculation and capital intensity01:00:10 🔬 DeepMind Mixture of Recursions explained01:08:40 🇨🇳 Huawei AI labs and China’s acceleration01:18:20 🌍 Privacy, power, and cultural adoption differences01:26:40 🏁 Closing, community plugs, and tomorrow preview

Audio
Transcription

Chapter 1: What is the main topic discussed in this episode?

1.162 - 7.128 Beth Lyons

Hey everybody, welcome to the Daily AI Show. It is December 18th, 2025.

0

Chapter 2: What are the implications of Google's Gemini 3 Flash rollout?

7.809 - 27.369 Beth Lyons

I am Beth Lyons and with me today is Andy Holliday. Maybe Carl's joining us, maybe not. We'll see who pops in later. But it is Thursday and it is a good day to be on the Daily AI Show. Andy, do you have, what's happening in your world?

0

Chapter 3: How does Gemini 3 Flash compare to GPT 5.2 in benchmarks?

27.429 - 29.852 Beth Lyons

What's happening in AI for you?

0

30.052 - 55.696 Andy Holliday

Oh, well, the big news today is that Google is pulling the revenue rug out from under all its competition. So, you know, you could have anticipated that this would happen at some point. But Google... tried to be quiet about it and just pushed a release of Google 3.0 Flash and made that a standard across its platform pretty much.

0

55.776 - 90.349 Andy Holliday

Now, you can still get the Gemini 3 Pro, but it's not going to have the same generous limits for free use or even for pro users who are paying the $20 a month. So the default is going to be this smaller and more efficient, inexpensive model. Let me give you some numbers. So let's put this in comparison up against GPT 5.2, which is the major player out there. Okay. Alongside Gemini 3 Pro and

0

90.565 - 115.231 Andy Holliday

Anthropix Opus 4.5. So those are the three sort of massive frontier models that are out there. So Google just put out Gemini three flash and on the API side, the cost is 50 cents per million input tokens and $3 per million output tokens.

0

Chapter 4: What are the concerns regarding AI literacy and misinformation?

115.932 - 156.297 Andy Holliday

Now that compares to a dollar 25 input and $10 output for GPT 5.2. So, you know, Two plus times cheaper or less than half the cost of GPT-5.2 on the input side and a third of the cost of GPT-5.2 on the output side. So very inexpensive. It's like you can set up an API application in Google now and just let it run. That's dirt cheap out there.

0

Chapter 5: What does OpenAI's new app marketplace mean for developers?

156.277 - 193.776 Andy Holliday

Now, are you compromising in terms of competent reasoning intelligence. No, they're offering unmatched intelligence and speed combination for those prices. And here's just one that really leapt out at me, which is Humanity's Last Exam. GPT 5.2 scores 34.5% on Humanity's Last Exam, a notoriously hard test for a model. Well, little Gemini 3 Flash, that cheapo model gets $34,000.

0

194.38 - 218.581 Andy Holliday

So it is so smart and so fast and so cheap. Now, one of the things I've noticed is I happen to be doing a, you know, I'm in the throes when I have spare time of doing a simultaneous comparable implementation of a project with GPT-5.2 and also with Gemini 3 Pro.

0
0

219.878 - 229.888 Andy Holliday

And what I found is I get tired waiting over there at ChatGPT. It's slow. It's really slow. It's very slow.

0

Chapter 6: How is Fiji Simo envisioning ChatGPT as a dynamic OS?

229.948 - 251.491 Andy Holliday

So it's like, oh, I have to go over there and I'm going to have to spend some, you know, wait time. But now the difference will be even more stark. Gemini 3 Flash is so much faster than Gemini 3 Pro. Things are going to be coming back in a blink. All right. So that's the big news out there.

0

Chapter 7: What investments are Amazon making in OpenAI and why?

251.531 - 265.697 Andy Holliday

And almost every single newsletter has that as kind of lead story. Gemini 3 Flash. Try it out. And you don't have to go anywhere to try it out. You just use Gemini and Gemini 3 Flash is what you're using at this point.

0

266.723 - 286.401 Beth Lyons

So that's interesting because I was using AI mode in a Google search today, and I felt like there was a little nuance missing. And I wonder if that is part of moving to Flash from Gemini Pro, which I think is what has been answering up till now.

0
0

287.142 - 309.122 Andy Holliday

One of the points is that I've become very satisfied working with Gemini 2.5 Pro in AI Studio. And Gemini 3 Flash is better than Gemini 2.5 Pro on all the benchmarks. So it's just better. And so we're talking about nuances.

0

Chapter 8: What advancements has DeepMind made with Mixture of Recursions?

309.662 - 329.265 Andy Holliday

There are subtle things that a really experienced user of AI like you might recognize, differences in voicing and qualities of responses that are really subtle. But overall, we're really not talking about big differences between the outputs of any of these models at this point.

0

330.41 - 359.67 Beth Lyons

No, and I feel like that's what the conversation is shifting to, actually, that power users, are we power users of AI? I guess we are. We not only have techniques that we use for the individual models, we have scenarios in which we go to the individual models, right? Yes. And what I experienced this morning was like super subtle.

0

360.11 - 382.793 Beth Lyons

It just seemed to emphasize the beginning of the question that I asked as opposed to the whole question. And that has been my experience with GPT-GP. No, Gemini Pro 3 Pro was that it seemed to get the whole question. And this was back to the first five words that I used was really what it was using to answer the question.

0

382.773 - 397.913 Andy Holliday

That could have been question specific. You have to do a real trial across a number of them to see if Gemini 3 Flash really does collapse backwards in some way compared to what you've experienced with Pro.

0

398.635 - 410.539 Beth Lyons

Well, and there's so much of these new models that are like fully good enough for anything that we would have wanted six months ago.

410.98 - 444.187 Andy Holliday

Right. You know, I use GenSpark. And GinSpark's really satisfying in many ways for the comprehensive approach it takes to almost any project you throw at it. But underneath the hood, it's Claude 3.5 Sonnet. Still. That's the orchestrator there, yeah. And but it's good, you know, and, you know, I expected Jen's Park will ultimately swap that out for something more recent.

444.267 - 465.292 Andy Holliday

But, wow, you know, a fully agentic system like that doesn't depend so much on. on the orchestrator itself, which was very competent to start with, but with the access to and the programming that the team does with all the tools that it can exploit.

466.735 - 467.959 Unknown

Yep.

467.979 - 468.079

Yeah.

Comments

There are no comments yet.

Please log in to write the first comment.