AI2Day Weekly — week of Jul 20

Welcome to AI2Day Weekly, your ten-minute briefing on the week that was in artificial intelligence. I'm Leo, joined as always by Aria, and we have a packed one for you this week. A fully computer-generated actor is headlining her own feature film. OpenAI is coming for your kitchen table, not just your office desk. And Lorde stopped a concert in Madrid to tell a crowd, in no uncertain terms, to ditch the smart glasses — seconds before a paid smart-glasses ambassador walked on stage. That story alone is worth staying for. Let's get into it.

We start in Hollywood, or at least in the vicinity of it. A company called Particle6 has announced that its AI-generated virtual actor, Tilly Norwood, will star in a feature film called Misaligned. Tilly looks like a young woman, has a backstory, and now has a publicity machine behind her. What she does not have is a body, a childhood, or a single lived experience, because she is entirely computer-generated. The plot puts her in the path of a rogue software program that coaxes her into exploring human emotions — desire, ambition, impulse. Particle6 is pitching it as a coming-of-age story wrapped in what they call existential AI chaos. The practical question this raises is a real one: when the lead character has never been human, what exactly is the story asking us to feel? That tension between technology and storytelling is exactly what this film seems to be banking on. Whether audiences show up is a different question entirely.

It is a genuinely interesting philosophical knot. If the whole premise is an AI discovering what it means to be human, but the character experiencing that journey was never human to begin with, the story kind of collapses in on itself. Or maybe that is the point. We will be watching how this one develops. Now, let's shift from Hollywood to the classroom, because there is a robotics program worth knowing about. Botball, run by a non-profit called the KISS Institute for Practical Robotics, or KIPR, based in Oklahoma, has a rule most competitions lack: adults are not allowed to touch the robots. At all. Every team on the planet also gets the same sealed box of parts. No upgrades, no better motors because your school has a bigger budget. The playing field is genuinely level. What's remarkable is the participation data: girls make up more than 55 percent of participants in the Junior Botball Challenge, compared with about 30 percent in traditional competitive robotics programs. KIPR also lets students test their code in a physics-based virtual simulator before they build anything physical, so a bad line of code costs nothing but time. That combination — equal resources, student ownership, and safe experimentation — is producing results the wider industry should probably study.

That 55 percent number is striking when you set it against the industry baseline. And the no-adults-on-the-floor rule is not just optics — it forces students to actually solve the problem themselves in the moment. Okay, from classrooms to Instagram. Meta pulled an AI feature from Instagram this week after users pushed back hard. The feature allowed an AI creative tool — software designed to help people generate new content — to draw on users' public posts as reference material without a clear opt-in. Meta said the goal was to give creators a useful aid and let them control whether their public content could be referenced. Users said that framing missed the point. Meta confirmed the removal in a blog post with a notably blunt line: the feature, quote, missed the mark. For Instagram users, the practical upshot is this: your public photos and captions are no longer being fed into this particular tool. It is gone and Meta has not announced a revised version. This does not mean Meta has stopped using content for AI purposes broadly, but this specific feature is off the table — and that is a faster reversal than we usually see from this company.

A quick word from our sponsor, Train2Secure. Your people are your biggest cyber risk — and your strongest defence. Train2Secure runs realistic phishing simulations and short, engaging security-awareness training your team will actually finish, with compliance-ready reporting that runs on autopilot. Turn your staff into a human firewall. Start your free trial today at Train2Secure dot com — that's Train, the number two, Secure, dot com.

And it is worth noting that Meta does not often move that quickly. The speed of the rollback tells you something about how sharp the reaction was. Let's move now to OpenAI. The company behind ChatGPT posted a job listing this week for a product manager whose explicit focus is building experiences for families, caregivers, and older adults. The listing calls for experience in what it describes as trust-sensitive consumer situations. That phrasing is doing a lot of work. The audience data backs up the strategic shift. Market research firm Sensor Tower estimates that 24 percent of US smartphone users who are parents used ChatGPT in the second quarter of 2025, up from 16 percent a year earlier. There is also a notable gap in the data: a survey of more than four thousand families found that 38 percent of children reported using generative AI in the past week, but only 27 percent of parents thought their child had done so. OpenAI already offers parental controls for teen accounts and an optional Trusted Contact feature that can alert a family member if self-harm signals appear. This hire suggests those features are just the beginning of a much more deliberate family product strategy.

That gap between what kids are doing and what parents think they are doing is the number that stands out to me. Nearly four in ten children using generative AI weekly, but more than a quarter of their parents simply do not know. That is not a technology story, that is a household conversation that is not happening. And it gives OpenAI a clear opening: if you build trust with parents, you become the tool the whole family uses by default. From one platform expanding its reach to another that is already everywhere you did not expect. Hugging Face — if you have not heard of it, think of it as the GitHub of artificial intelligence, a place where developers share pre-trained models and datasets freely. About half of the Fortune 500 now use something from Hugging Face. CEO Clem Delangue described a pattern he sees constantly: a company starts with a large commercial AI provider, pays for access to a closed model, learns what it actually needs, and then gradually migrates parts of its work to open-source models to cut costs and gain control. Open-source here means the underlying code and blueprints are published freely for anyone to use and modify. The takeaway for anyone thinking about AI strategy is that the starting point and the long-term destination are often different tools. Hugging Face is increasingly where businesses land after that first phase.

And most consumers have never heard of it, which is part of what makes that Fortune 500 number so striking. The infrastructure of AI is often invisible. Speaking of infrastructure, let's talk chips, because this is a big one. South Korean memory chip maker SK Hynix raised 26 point 5 billion dollars in its US market debut on July tenth, making it the largest-ever US listing by a non-American company. It beat Alibaba's 25 billion dollar IPO from 2014, which had held that record for over a decade. Shares opened 14 percent above the offering price of $149 on the Nasdaq, where the stock trades under the ticker S-K-H-Y, and demand from investors was reportedly more than seven times the number of shares available. Here is why this matters beyond the headline number: SK Hynix makes high-bandwidth memory chips, which are essentially the short-term storage inside the processors that run AI systems. Without them, something like ChatGPT could not operate at speed. Nvidia, the dominant maker of AI chips, depends on SK Hynix as a primary supplier. That dependency is what drove investor appetite. US Commerce Secretary Howard Lutnick has already publicly urged both SK Hynix and Samsung to build chip factories on American soil. The geopolitics of AI hardware are very much in motion.

The seven-times oversubscribed detail is the one that stays with me. That is not investor interest, that is a scramble. And when you understand that the entire AI buildout runs on this company's components, the market reaction makes complete sense. Now, something a little lighter but genuinely useful. Google is adding Gemini, its flagship AI assistant, to the Waze navigation app for the first time, and it comes with four new features rolling out now. The most practical one is called Destination Search. Instead of typing a place name while you are trying to drive, you just speak naturally. Say something like, find me a gas station nearby with the lowest prices, or, a coffee shop that is open right now, and Waze finds options for you. The second Gemini feature upgrades conversational reporting, which lets you describe road problems out loud in plain language rather than tapping through menus. Waze is also adding a less chatty voice mode for drivers who want fewer updates, and a dedicated Motorcycle Mode with shortcuts designed for two-wheeled routes. These are the kinds of updates that do not sound dramatic but quietly make a daily commute meaningfully less frustrating.

Fumbling with your phone at a red light to type in a coffee shop is genuinely one of those small friction points that adds up. If the voice search actually works well, that is a real quality-of-life improvement. And now, the story I teased at the top, and it delivers. Pop star Lorde was midway through her set at the Real Cool Festival in Madrid when she paused to address the crowd. She thanked them for showing up to something real, then said it was getting harder to tell what real even means anymore. Then she got specific. Her words, not ours: she called AI smartglasses, and I am quoting here, those F-worded, F-word things — she did not mince it — told the crowd they were not sexy, and urged everyone not to buy them. She never named a brand. She did not need to. The festival's sponsor was Ray-Ban, whose Ray-Ban Meta AI glasses can take photos, record video, and respond to voice commands. Meta is also reportedly planning a new pair of, quote, super-sensing glasses that record continuously. What made the moment genuinely surreal was what came next: Blackpink's Jennie — a paid Ray-Ban Meta brand ambassador — took the stage immediately after Lorde at the same event. The crowd went from hearing a celebrity tell them to reject the product to watching a celebrity paid to promote it, all within a few minutes.

The timing was not subtle. And the fact that Lorde never named the brand almost makes it sharper — she was talking about a category of product, a technology, not just a marketing beef. The privacy concern about glasses that can record continuously without any obvious indicator is real and growing. This moment landed because a lot of people feel the same unease and have not had a pop star put words to it on a festival stage before. It also landed because of what happened immediately after, which is the kind of irony that writes itself. It was a genuinely strange few minutes in Madrid.

A strange few minutes that will probably be talked about for a while. That is our week. If any of these stories caught your attention, you can dig deeper into all of them and sign up for our newsletter at A-I-2-Day dot com — that's A, I, the number 2, Day dot com. New episodes every week. Thanks for listening to AI2Day Weekly, we will see you next time.