Demis Hassabis & Josh Woodward Tell Us Why Gemini 3.0 Puts Google in Front of the A.I. Race

Summary

In this special emergency episode of Hard Fork, Kevin Roose and Casey Newton break down the launch of Google’s Gemini 3, which they describe as potentially a turning point in the AI race. The episode features an interview with Demis Hassabis, CEO of Google DeepMind, and Josh Woodward, VP of Google Labs and Google Gemini, who explain what makes the new model significant and why Google’s competitors are nervous.

Gemini 3 shows major improvements across benchmarks, including nearly doubling the score on “Humanity’s Last Exam” compared to its predecessor. The model introduces generative interfaces that build custom UIs in response to queries, a Gemini agent capable of managing email inboxes, and improved coding capabilities. Kevin and Casey note that Google’s massive distribution advantage through its existing products — search, Gmail, Android, YouTube — gives it an almost unfair edge as its models improve. Hassabis maintains his 5-to-10 year AGI timeline, saying the progress is “dead on track,” while acknowledging one or two more research breakthroughs are still needed.

Highlights

”Their Names Are Gemini 3 and Opus 4.5”

yt-dlp --download-sections "*1:00-1:50" "https://www.youtube.com/watch?v=rq-2i1blAlU" --force-keyframes-at-cuts --merge-output-format mp4 -o "rq-2i1blAlU-1m00s.mp4"

“Maybe more than other model releases, this one seems to have the attention of Google’s competitors. We’re hearing a lot of whispers from folks who work at other AI labs that it seems like Gemini 3 has managed to figure some things out in a way that may be bad for their businesses.” — Casey Newton, 1:00

”Step One: Build an Illegal Monopoly”

yt-dlp --download-sections "*5:46-6:10" "https://www.youtube.com/watch?v=rq-2i1blAlU" --force-keyframes-at-cuts --merge-output-format mp4 -o "rq-2i1blAlU-5m46s.mp4"

“Google’s big advantage here over their competitors is that they have products that are used by billions of people a day and they can kind of shove Gemini 3 into those products over time. Which is why we always tell students when they ask us for advice, step one, build an illegal monopoly.” — Kevin Roose, 5:46

”Dead on Track for AGI”

yt-dlp --download-sections "*11:00-12:00" "https://www.youtube.com/watch?v=rq-2i1blAlU" --force-keyframes-at-cuts --merge-output-format mp4 -o "rq-2i1blAlU-11m00s.mp4"

“I think it’s sort of dead on track. We’re really happy with this progress. I think it’s an absolutely amazing model and is right on track of what I was expecting… I still think there’ll be one or two more things that are required to really get the consistency across the board that you’d expect from a general intelligence.” — Demis Hassabis, 11:00

”Maybe It’s One of Those Cool Drag Racers”

yt-dlp --download-sections "*8:18-9:00" "https://www.youtube.com/watch?v=rq-2i1blAlU" --force-keyframes-at-cuts --merge-output-format mp4 -o "rq-2i1blAlU-8m18s.mp4"

“Two years ago, Sundar Pichai told us that Bard was a souped-up Civic in a race with more powerful cars. What kind of car is Gemini 3? Well, I hope it’s a bit faster than a Honda Civic. Maybe it’s one of those cool drag racers.” — Demis Hassabis, 8:18

”Not Working Out for the Rest of the Web”

yt-dlp --download-sections "*5:33-6:10" "https://www.youtube.com/watch?v=rq-2i1blAlU" --force-keyframes-at-cuts --merge-output-format mp4 -o "rq-2i1blAlU-5m33s.mp4"

“So far they say that the usage keeps going up for AI overviews and every quarter they continue to make more money. So seems to be working out for them. Not working out for the rest of the web, but it’s working out well for Google.” — Kevin Roose, 5:33

Key Points

• Gemini 3 Launch (0:00) - Google releases Gemini 3, prompting a special emergency episode of Hard Fork
• Competitor Nervousness (1:00) - Whispers from rival AI labs suggest Gemini 3 has “figured some things out” that may hurt their businesses
• Generative Interfaces (2:00) - Gemini 3 builds custom interactive UIs in response to queries rather than just returning text
• Humanity’s Last Exam (2:55) - Gemini 3 Pro scored 37.5% on this benchmark vs. 21.6% for Gemini 2.5 Pro
• Gemini Agent (4:05) - New agent can manage your inbox, propose replies, and organize emails together
• Free for College Students (6:11) - Google giving all US college students a year of free access to paid Gemini
• Distribution Advantage (5:46) - Google can push Gemini 3 into products used by billions: Search, Gmail, Docs, Android, YouTube
• AI Mode in Search (5:00) - Gemini 3 available in AI mode tab alongside main Google search, suggesting cost efficiency
• Multi-Step Reasoning (8:45) - Josh Woodward highlights the model’s improved ability to think many steps without losing track
• AGI Timeline Unchanged (11:00) - Hassabis maintains 5-10 year AGI timeline, says progress is “dead on track”
• World Models Still Needed (11:30) - Hassabis says improvements in reasoning, memory, and world model ideas from Genie and Simmer are still required
• Task Completion Metric (13:00) - Google tracking “how many tasks did we help you complete” as a new product metric
• Safety Testing (16:00) - Hassabis calls this their most thoroughly tested model, with extra caution on tool use and coding capabilities
• AI Bubble Question (18:00) - Hassabis says some parts of AI industry are in a bubble (multi-billion seed rounds), but Google is positioned to win in either scenario
• Scaling Laws Not Zero (17:00) - Hassabis clarifies diminishing returns doesn’t mean zero returns — still “extremely good return on investment”

Mentions

Companies

• Google / Alphabet (0:00) - Releasing Gemini 3, integrating it across all products
• Google DeepMind (0:37) - The “engine room” of Google’s AI efforts, led by Demis Hassabis
• OpenAI (3:55) - Referenced as competitor whose ChatGPT may be surpassed
• Anthropic (7:00) - Casey’s boyfriend works there; mentioned as competitor
• Isomorphic Labs (19:00) - Google’s drug discovery company using AI

Products & Technologies

• Gemini 3 (0:00) - Google’s new frontier AI model, the focus of this episode
• Gemini 2.5 Pro (2:55) - Previous model that scored 21.6% on Humanity’s Last Exam
• Google Anti-Gravity (9:00) - New Google coding product showcasing Gemini 3
• Notebook LM (13:00) - Google’s AI-first product mentioned alongside Gemini app
• Genie / Simmer (11:30) - Google’s world model projects that will extend Gemini’s capabilities
• Bard (8:18) - Google’s previous AI chatbot, now replaced by Gemini
• Nano Banano (24:00) - Google’s image generation/editing feature in Gemini
• AI Overviews (5:33) - Google’s AI-generated summaries in search results

People

• Demis Hassabis (0:37) - CEO of Google DeepMind, interviewed about Gemini 3
• Josh Woodward (0:48) - VP of Google Labs and Google Gemini, interviewed about Gemini 3
• Sundar Pichai (8:18) - Referenced for previously calling Bard a “souped-up Civic”
• Mark Humphries (21:00) - Historian who used an unreleased Google model to transcribe 1800s documents

Surprising Quotes

“If you look at seed investment rounds being multi-10 billion rounds with basically nothing — it seems like that might be the first signs of some kind of bubble.” — Demis Hassabis, 18:30

“We always tell students when they ask us for advice: step one, build an illegal monopoly.” — Kevin Roose, 6:08

“Not working out for the rest of the web, but it’s working out well for Google.” — Kevin Roose on AI Overviews, 5:43

“I’ve been getting back into my games programming. I’m going to set myself some projects over Christmas… I feel like it’s actually got to a point where it’s incredibly useful and capable on front end.” — Demis Hassabis, 10:30

Transcript

0:00 Well, Casey, we have a special emergency podcast episode today about the launch of Gemini 3. Yes, Kevin, hotly awaited, much discussed among AI nerds here in Silicon Valley. We are finally about to get our hands on the genuine article.

0:19 Yeah. So, normally we wouldn’t break our Friday publication schedule to publish a special episode just about a new model coming out from one of the big AI companies. They’re releasing models all the time. But there are a couple reasons that we thought it was worth doing this week to talk about this model Gemini 3 in particular. The first is that we got some time with Demis Hassabis and Josh Woodward, two of the leading AI executives at Google.

1:00 Yeah. I mean, one big thing, Kevin, is just that maybe more than other model releases, this one seems to have the attention of Google’s competitors. We’re hearing a lot of whispers from folks who work at other AI labs that it seems like Gemini 3 has managed to figure some things out in a way that may be bad for their businesses. And I think around the AI industry, there’s sort of this feeling that Google, which kind of struggled in AI for a couple years there — they had the launch of Bard and the first versions of Gemini, which had some issues — and now the question is, is this them returning to the top of the AI leaderboard?

2:00 In terms of what Gemini 3 can do, Google shared a few different things. In addition to being better at coding and vibe coding, it’s going to do some new things around generating interfaces for you when you ask it a question. They showed an example where somebody wanted to learn about Vincent Van Gogh, and Gemini 3 coded up an interactive tutorial with images and interactive elements. They showed another example involving building a mortgage calculator for buying a home over a million dollars, which is the lowest amount of money that anyone at Google can imagine spending on a home.

2:55 The theme of the briefing was this is just kind of better than their last model Gemini 2.5 Pro in basically all respects. One benchmark that caught my attention was Humanity’s Last Exam, a very hard interdisciplinary exam at a graduate or PhD level. Gemini 2.5 Pro got about 21.6% on that test, and Gemini 3 Pro gets 37.5%. That’s basically the story of all these benchmarks.

4:05 They also talked about testing what they’re calling the Gemini agent, which is going to be able to look through your inbox, understand its contents, propose replies, organize emails together, and really help you get your inbox under control in a way that I personally have never been able to.

4:38 They are not rolling this out to everyone right away. It’s going to be available this week for users in the Gemini app and in the AI mode, which is the tab alongside the main Google search engine. It will also be available for developers. But they’re not saying when this will come to the Gemini integrations in Google Docs or Gmail.

5:33 So far they say that the usage keeps going up for AI overviews and every quarter they continue to make more money. Not working out for the rest of the web, but it’s working out well for Google. Google’s big advantage here over their competitors is that they have products that are used by billions of people a day and they can kind of shove Gemini 3 into those products over time.

6:11 And speaking of students, the other notable announcement is that Google is giving all US college students a year of free access to a paid version of Gemini. Which is I think a smart move. I feel a little gross about it — essentially telling students, hey, why don’t you use this to maybe do some of your homework?

7:00 So that is what we know about Gemini 3. We should obviously make our AI disclosures. I work for the New York Times company which is suing OpenAI and Microsoft. And my boyfriend works at Anthropic.

8:18 Demis and Josh, welcome to Hard Fork. Two years ago, Sundar Pichai told us that Bard was a souped-up Civic in a race with more powerful cars. What kind of car is Gemini 3? Well, I hope it’s a bit faster than a Honda Civic. Maybe it’s one of those cool drag racers.

8:45 What can Gemini do on a concrete level that previous AI models couldn’t? One, we’re starting to see this model really excel on reasoning and being able to think many steps at the same time. The other thing is all kinds of new generative interfaces. And we’ve put a lot of investment in coding itself.

10:00 There’s been some discussion that for average users, the chat use case can feel solved. To what extent does Gemini 3 feel meaningfully different? I think this is a model that’s more concise, more expressive. It starts to present information in a way that’s much easier to understand. And it’s got a better style about it. I find it more pleasant to brainstorm with and use.

11:00 The last time we had you on the show, you said you think we’re 5 to 10 years away from AGI. Has Gemini 3 changed any of those timelines? No, I think it’s sort of dead on track. I still think there’ll be one or two more things that are required to really get the consistency across the board that you’d expect from a general intelligence. Improvements still on reasoning, on memory, and perhaps things like world model ideas from Genie and Simmer.

13:00 We’re really interested on the team to be able to track things like how many tasks did we help you complete in your day. That’s a new type of metric. The original Google search — you would come to it, get an answer, and sort of move on.

15:00 Some of your competitors have been very nervous in the days and weeks leading up to Gemini 3. Do you feel like Google is ahead in the AI race right now? It’s a ferocious competitive environment, probably the most competitive there’s ever been. We’ve always pioneered the research part of this. Getting into our groove in making sure that downstream reflected in all of our products. GDM being the engine room of Google.

16:00 Any new dangerous capabilities or safety concerns? We’ve taken quite a long time on this model because it’s frontier and has some new capabilities. We’ve worked really hard on things like tool call usage and function calling. But it also makes them more capable for riskier things like cyber. This is our most thoroughly tested model so far.

17:00 How are you guys thinking about scaling laws in relation to Gemini 3? People when they hear diminishing returns think of is it zero or exponential. But there’s also in between. It’s not going to exponentially double with every era, but it’s still well worth doing. Extremely good return on that investment.

18:00 Are we in an AI bubble? My view is that there are some parts of the AI industry that are probably in a bubble. If you look at seed investment rounds being multi-10 billion rounds with basically nothing — that might be the first signs of some kind of bubble. But there’s also robotics, gaming, drug discovery with Isomorphic, and all these new green field areas. I feel really good about where we are as Alphabet whether or not there’s a bubble.

24:00 What would you have our listeners show their friends at Thanksgiving to save the holiday? Our imagery models in Gemini are still best in the world. Grab your phone, take a selfie, put yourself in it, and edit it. People are still doing that at huge amounts. You heard it here — Nano Banano will save Thanksgiving dinner. Gentlemen, thank you.