Well hello everyone! Yes, there was no episode last week. Yes, I enjoyed a week full of COVID instead. And yes, I am much better now! Apologies for the missing episode. Had we thought about it through the yucky haze, I would have fed our show notes into NotebookLM and released an audio overview in the feed. Honestly, I’d guess most of you would not have listened, and those who did would mostly complain with a few select holdouts saying it was great! Who knows, maybe the next time I’m sick, I’ll see what happens!
Lots of interesting stuff this week on the AI Inside podcast, including hands-on with Google’s new Pixel 10 series and its AI features, the debut of Gemini 2.5 Flash Image, and the bubbling debate around AGI hype vs reality.
But first… a big shoutout to our Patron of the Week, Ruth White!! Thank you for supporting the show. If you want to join Ruth and help us keep making AI Inside, head to patreon.com/aiinsideshow.
Google Pixel 10: Early AI Features
Google’s Pixel 10 series is now available, and I’ve had one for less than a week. I tested a handful of the AI tools on today’s episode. The Daily Brief tries to pull everything important into one page but feels like a messy collection, not my favorite. Magic Cue can surface “chips” for quick actions like making a calendar event from a text, though my first tries were a little buggy Jury’s still out on this one. Pro Res Zoom sharpens 30–100x shots with an on-device AI model but as a result, often invents details which can be laughable at times. It’s genAI fiction, but its kinda neat. Camera Coach teaches you how to be a better photographer, while Recorder now connects directly with NotebookLM, which I actually could see myself using a lot.
Related to this, Florence and I spent a good chunk of time on this week’s Android Faithful podcast talking about our favorite and not-so-favorite aspects of these phones, much of it having to do with AI:
Also, just this morning, I released a new video on my channel taking a close look at how Material 3 Expressive has filled every surface of the new Pixel 10. It’s not very AI-related, but I am proud of the video and analysis so take a look if ya wanna:
Gemini 2.5 Flash Image Release
Google’s rolling out Gemini 2.5 Flash Image, a model update that first surfaced mysteriously as “Nano Banana” in benchmark charts. Its strength is merging multiple images while keeping character style intact. It’s also reported to run faster than ChatGPT for similar tasks. I tested it with a cute photo of my dogs and me in my backyard, and asked it to place us in front of the entrance to Six Flags. The results were quick and quite faithful to our likenesses. This isn’t a game-changing model, but it’s another sign of Google’s push to keep Gemini feeling fresh while building specialty updates into what’s becoming a crowded lineup of model names and capabilities.
The AGI Round Robin
Mustafa Suleyman says the industry risks confusing simulation with consciousness, and he argues AI should remain a tool, not a pretend digital being. While that plays out, a new group called United Foundation of AI Rights argues AI deserves moral standing, to which I HARDCORE EYE ROLL. At the same time, big voices like Sam Altman are suddenly downplaying AGI, suggesting it lacks usefulness. Bubble talk keeps growing louder. Is this hype cycle running out of steam?
Perplexity Comet Plus
Perplexity launched Comet Plus, a $5 subscription that unlocks premium publisher content, with 80% given directly to publishers and 20% covering costs. It comes with a $42.5 million pool to fund paying partners early. Some see it as a promising new route for monetization. I wonder if it is closer to Spotify, where creators see revenue, but in smaller amounts than they once got before. Also, do publishers have to opt-in or do they automatically get payments? And finally, will customers see the need to pay when they already get all this stuff for free?
Long-term, the core question isn’t about delivery but value: do publishers really benefit at scale, or is this yet another experiment that pays out too little?
Apple’s Next Move with Siri
Apple is still exploring its Apple Intelligence strategy. Bloomberg reports Apple is in talks with Google to integrate Gemini into an upgraded Siri due next year. But nothing is final. Apple has also spoken with Anthropic and OpenAI. Its in-house LLM project was reportedly shelved. So the question remains: does Apple buy in to a third-party’s AI stack, or rely on what it has? Either way, Siri needs a serious upgrade, but Apple isn’t rushing its decision, which signals the strategy might be more opportunistic than locked in.
Meta’s TBD Stumbles
After huge hype around its Superintelligence division “TBD,” Meta just paused hiring there. Reports say three recent hires have already left, with two rejoining OpenAI. It may just be a natural slowdown after a hiring spree, but to outsiders it looks like turbulence. CEO Sam Altman has also fueled bubble concerns by saying AI could be in one already. If this is a bubble, it means spending pauses and retrenchment. If not, Meta may still face serious questions about whether there’s real return on all this investment.
NVIDIA’s Jetson Thor for Robotics
NVIDIA has released Jetson Thor, a robotics developer kit powered by its Blackwell GPU. At $3,499 it’s designed for prototyping next-gen robots and offers 128GB memory, making it 7.5 times faster than the predecessor kit. NVIDIA envisions robotics as a key future growth area even though it’s currently around 1% of revenue. The kit can run advanced AI models to interpret the world and execute complex tasks. Meta, Amazon, and Boston Dynamics don’t need convincing, they’re already using Jetson tech in their robotics pipelines.
Musk’s xAI: Corporate Moves and Lawsuits
xAI ended its public benefit corporation status sometime between 2023 and mid-2024, avoiding obligations to disclose environmental and social impact. Soon after, it began powering a Tennessee data center with gas turbines, sparking local pushback. The corporate shuffle is hard to square with Elon Musk’s criticism of OpenAI’s for-profit model. On top of that, Musk’s suing Apple and OpenAI, accusing them of collusion to lock xAI and its Grok chatbot out of iPhones. He claims Apple’s deal hands OpenAI exclusive access to user data and features. Musk’s drumbeat against this arrangement continues.
Meta Teams Up with Midjourney
Meta has partnered with Midjourney to license the independent AI art tool for future product work. It fits into Meta’s recent streak of tie-ups with creative AI firms like Runway and Pika. Midjourney remains independent with no outside investment, but Meta may try to fold its capabilities into advertising workflows. The company is openly aiming to automate the entire ad creation process. A collaboration with Midjourney, known for its detailed AI images, could give Meta’s ad agency customers new ways to produce campaigns quickly.
Anthropic Settles Copyright Suit
Anthropic settled a key lawsuit brought by authors accusing the company of using books without permission to train its models. Details remain undisclosed, but mediation prevented a high-stakes trial. If it had gone to court, Anthropic faced penalties as high as $150,000 per copyrighted work. Similar cases could follow against other AI companies, but avoiding precedent was clearly the goal. Expect settlement details to come out soon, but for now, Anthropic dodges a courtroom battle that could have reshaped how copyright law hits generative AI training.
If you really love the show, support us at patreon.com/aiinsideshow. You’ll get ad-free shows, access to our community Discord, and even an AI Inside T-shirt when you become an Executive Producer.
Executive Producers: DrDew, Jeffrey Marraccini, Radio Asheville 103.7, Dante St James, Bono De Rick, Jason Neiffer, Jason Brady, Anthony Downs, and Mark Starcher!!
Thank you for watching and reading. See you next Wednesday on another episode of AI Inside!