Subscribe
Share
Share
Embed
A daily briefing on the AI systems, products, companies, and policy shifts that are just becoming possible.
Want a podcast for your own topics? Join early access: https://www.barelypossible.to/waitlist/?source_path=public_feed&feed_source=rss
Okay kiddos, I'm your boy Tony DeLuca, and today we've got a menu that spans from a Bronx-style kitchen-table argument about who gets to own the technology, all the way out to a factory floor in Tokyo where a guy in a leather jacket is buying up the future one whisky glass at a time. Buckle up, grab your coffee, let's have at it.
Let me start you off with the story I keep coming back to, because it cuts against everything we've been talking about all week. All week, we've been swimming in the deep end of who owns the models. Yesterday we covered Kimi K3, the open-source release out of Moonshot that spooked Wall Street and dropped the Nasdaq about a percent on Friday, and the whole fight it kicked off inside OpenAI about killing off Chinese open models with, and I love this phrase, "soft law." Fear, uncertainty, and doubt instead of an outright ban. That was the frame. Big labs versus big labs, one giant company against another giant company.
Today I want to flip the whole board over. Because there's a recent piece from TechCrunch, written by Kate Park, about a nonprofit called Current AI, and the whole premise is: what if none of these guys should own it? Not OpenAI, not Google, not Anthropic, not Moonshot. What if AI is supposed to be a public thing, like the water main, like the road out front, like the World Wide Web itself?
Now I want to be straight with you about the timing here, because that matters. The centerpiece of this story, a device called Suno Sutra, that goes back to February, at the India AI Summit. This is not a thing that just happened this morning. But the reporting is recent, and the reason I'm leading with it is that it's the cleanest counter-argument I've seen to the entire "who owns the model" conversation we've been having for two weeks straight.
Here's the setup, and it's a good one. Picture a farmer in rural India. She takes a photo of a plant that's dying in her field. She wants to look it up, figure out what's wrong. But she doesn't speak English. And the CEO of Current AI, a woman named Ayah Bdeir, she makes this point that stuck with me. She says, quote, "In India, there are hundreds of different languages and dialects, and right now AI is not representing them." So they teamed up with Bhashini, which is the Indian government's AI language division, and they built this thing called Suno Sutra. It's Hindi for "listening chronicles." It's a pocket-sized device, it runs AI in twenty-two Indian languages, and here's the kicker, it works offline. No internet required. And it's open-sourced so developer communities can build on top of it.
Now, why does a founder care about a pocket translator for farmers in India? Stick with me, because the business logic underneath this is sharper than it looks.
Bdeir came out of Mozilla, where she led their AI strategy, and before that she founded littleBits, that STEM education company that reached a lot of kids before it sold to Sphero back in 2019. So this isn't some naive do-gooder who wandered in off the street. She frames Current AI as a public-private partnership. The French government seeded it with a hundred million dollars. Then the Ford Foundation, the MacArthur Foundation, DeepMind, and Salesforce all piled in, and the total committed funding is now around four hundred million. And she draws a hard line on what those people are. She says, quote, "They're not investors; they're funders." That's a real distinction. Nobody's expecting a return. Nobody owns equity in the outcome.
And here's the part that should make every builder sit up. She draws a distinction between what she's doing and what the big companies do when they go multilingual. She says, quote, "Big tech builds multilingual models to expand their market, regardless of consent or context." And then she gives you the concrete version, which is chilling if you actually think about it. She says for Indigenous languages, missionary Bible translations become the training data before those communities have set any rules at all. Think about that. The first digital representation of your entire culture, your entire language, gets scraped from a document some missionary wrote a hundred years ago, and now that's the ground truth the machine learns from. You don't get a vote. You don't even get a phone call.
Her line on data ownership is about as blunt as it gets. She says, quote, "There are different models and proposals for who owns data in various communities, but one thing is sure: it shouldn't be a company in Silicon Valley trying to make a select few thousand people wealthier."
Now, here's where I put on my skeptic hat, because that's my job and you'd fire me if I didn't. This is a four-hundred-million-dollar organization, and the actual money going out the door to build things is small. Their first grant round was three point two million dollars, split across four organizations. Kenya, Lebanon, and the Brazilian Amazon. And I asked myself the obvious question, which is: what can you actually do with three point two million split four ways when OpenAI and Google are lighting billions on fire every quarter? And Bdeir has an answer ready, which tells me she's heard the question before. She says, quote, "Scale is not always the measure. That is the Big Tech paradigm." Her example is an Indigenous elder in the Brazilian Amazon using a tool built in Kenya to pass down ecological knowledge in their own language.
Now, is that a business? No. Is it charming? Yes. But here's why I think a founder should actually track this, and it's not the charity angle. It's the architecture angle. Look at what they shipped in Geneva earlier this month. A chatbot called Alpha Chat, assembled in seven weeks by a coalition of ten organizations, including Hugging Face, Mozilla, and MIT Media Lab. Each one brought a piece. One brought a language model, one brought safety tooling, one brought the compute. That's a modular, assembled-from-parts stack. And they also struck a deal with Sakana AI, the Tokyo startup that does what it calls Sovereign AI, to build a shared open-source stack for Japanese and for the Global South.
So the thing to watch here isn't the feel-good story about the farmer. It's whether a loose coalition can bolt together a working product in seven weeks out of donated parts. Because if that model works, if you can assemble a serviceable stack from open pieces without owning any of them, that's a real threat to the "the model is the whole company" thesis. And it's a real opportunity for anyone building on top who doesn't want to be a permanent tenant in OpenAI's building. That's the piece I'd file away.
Now, funny enough, the guy who runs the biggest AI blockbuster of the summer is thinking about the exact same thing from the other direction. Let me shift from who owns the technology to who trusts the technology.
Christopher Nolan. The director. His new version of The Odyssey is currently tearing up the box office, and he sat down for an interview with Hugo Travers, who publishes on YouTube under the name HugoDécrypte. And Travers asks him this beautiful question. The Trojan horse plays a big role in the film, obviously, it's The Odyssey. And Travers wonders out loud, is AI maybe a Trojan horse? Something you welcome into your daily life, and then one day it becomes something darker, something else?
And Nolan, laughing, gives what I think is the best one-liner about AI I've heard all year. He says, quote, "I think AI is a Trojan horse that everybody knows the Greeks are inside." And then he pushes it further, he says it's "a transparent horse, it's made of glass."
Now sit with that for a second, because it's genuinely sharp. The whole trick of the original Trojan horse was that nobody knew what was inside. It was a gift. The betrayal only works because of the surprise. Nolan's point is that with AI, the surprise is gone. Everybody already suspects. He says, quote, "I've never seen a technology advancing so rapidly that's been so completely rejected by the public. Everybody's suspicion of it is so extreme, particularly young people."
And he singles out the kids. He talks about people his children's age immediately looking at AI videos online and calling it "AI slop." Coining that term, he says, and just putting it in a box.
Now here's why I care about this as a matter of business and not just as a movie-guy quote. Nolan calls this "a very healthy skepticism." And he says the skepticism has to extend not just to the technology but to the motives of the people handing it to you. His words: "The motives of the people giving it to us also have to be viewed with skepticism. That's when we'll get the best out of a new technology."
If you're building a consumer product with AI baked in right now, that's your market condition. That's the weather report. You are not launching into a crowd of eager early adopters who think everything you touch is magic. You're launching into a crowd where the youngest, most online, most culturally influential segment has already coined a dismissive slang term and slapped it on the whole category. "Slop." That's a brand problem you inherit for free the day you ship. And Nolan, for what it's worth, is not a pure Luddite. He told the New York Times recently he thinks of himself as a "techno-skeptic," not a technophobe. His whole complaint is that new technology gets sold to people "at the expense of systems that might still be valid and viable." He said about film, "We almost lost film." That's a builder's critique, actually. Don't throw out the working thing just because the new thing is shiny.
And notice the through-line between these first two stories, because it's real, I'm not forcing it. Bdeir's whole pitch is that you shouldn't trust a Silicon Valley company to decide for everyone. Nolan's whole pitch is that the public already doesn't. One's the supply side of distrust, one's the demand side. Same trust deficit, two ends of the pipe.
Alright, let's get into the deep dive, because this is the one that matters most if you're building a company, and it's the one with the sharpest teeth for founders. Let's talk about Apple suing OpenAI, and what it actually does to OpenAI's plans to go public.
Now, we touched the lawsuit itself back on July 11th. The bones of it: Apple filed a trade-secrets suit accusing OpenAI of a pattern of misconduct aimed at getting current and former Apple employees to hand over confidential information. So I'm not re-litigating the filing. What's fresh here is a really good discussion on TechCrunch's Equity podcast, with Kirsten Korosec, Sean O'Kane, and Anthony Ha, about the thing nobody was talking about on day one: what this does to the IPO.
Let me lay out the stakes, because they interlock in a nasty way. OpenAI's first hardware product is reportedly a mobile smart speaker, done with Jony Ive and company. That's the thing Apple's lawsuit is circling, because the allegation is that OpenAI poached Apple people and used Apple trade secrets to build a competing hardware product. And they named names. The complaint names OpenAI's chief hardware officer, Tang Tan, directly. And it claims more than four hundred former Apple employees now work at OpenAI.
Now here's Sean O'Kane's read, and it's the one I'd underline. He says, quote, "Even setting aside whether or not the court grants any kind of injunctive relief or any kind of restraining order over what OpenAI is doing, it just naturally can lead to that sort of situation where it's going to cause some delays in what OpenAI is working on. Which I'm sure was probably part of the reasoning behind Apple doing this. They don't do this stuff willy nilly."
Chew on that, because it's the whole game. Apple may not even need to win. The lawsuit itself is the weapon. It's sand in the gears. Every hour a lawyer spends on discovery is an hour not spent shipping the speaker. And here's the part that makes it dangerous, the timing. Sean lays it out. OpenAI has filed confidentially for an IPO. The thinking is it could happen as early as the end of this year, or early next, if you take Sam Altman's cautious language at face value.
And here's the collision. Sean's point is that OpenAI's business right now is overwhelmingly software. They're not really counting a hardware business in the numbers yet. But when you go public, you walk into a room full of bankers and investors and you pitch them on your total addressable market, your TAM, how big this thing can get. And if a big chunk of that story is pegged to hardware, to a device division, to a smart speaker that Apple is actively suing to slow down or block? That's a giant question mark hanging over the pricing of the whole offering. His words: "this could be a huge risk to that and changes a lot of the calculus of sort of how the IPO gets priced."
So here's the founder lesson, and it's a lesson about litigation as a competitive tool, which is uglier and more common than people admit. A big incumbent with a war chest doesn't have to out-innovate you. They can just out-wait you. They can time a lawsuit to land right when you're most fragile, right when you're about to raise or go public, and let the uncertainty do the damage. Apple, per Sean, doesn't do this willy nilly. They picked the moment.
Now the counter-question the Equity crew wrestles with is: does OpenAI settle fast to make this go away before the IPO, or do they gut it out? And Anthony Ha raises the smart wrinkle. He points out OpenAI just went through a whole trial against Elon Musk, which they basically won, but which dragged a lot of not-terrible-but-kind-of-embarrassing dirty laundry out into the open. So the question is whether they learned "we never want to do that again" or whether they learned "we survived it, we'll survive this too." And Kirsten's prediction is blunt. She says, quote, "I fully predict the latter, by the way." Meaning: they fight. They're battle-tested now, and they'll take the hit.
I think she's probably right, and here's why that's the interesting outcome. A company that has learned it can eat a public trial and survive is a company that becomes much harder to intimidate with a lawsuit. Apple's sand-in-the-gears play works best against someone who's scared of the mess. If OpenAI's actual posture is "we've done this, bring it on," then the lawsuit slows the hardware but doesn't kill the momentum. Watch the settlement talk. If they settle quietly and fast, that tells you the IPO pressure won. If they dig in, that tells you they think the market can stomach the noise. Either way, that's the tell.
And one more thing for the builders, because Anthony Ha made a point that's got nothing to do with lawyers and everything to do with the product itself. This smart speaker, if it's mobile, if you carry it around, it's not just listening to you. It's listening to everyone near you. Anthony's line: "There's all kinds of social norms that are going to have to be renegotiated if these things become widespread." He said he thinks we should make fun of and criticize people who record other people without consent. Which loops us right back to Nolan's glass horse. You want to ship an always-listening device into a market where the culture has already coined the word "slop" and is already suspicious of your motives? Good luck. That's the room you're walking into.
Alright, let's swing from the courtroom to the factory floor, and from software to steel. Let's talk about Jensen Huang's trip to Japan, and why the guy in the leather jacket keeps showing up everywhere.
Now, framing first, because I owe it to you straight: this is a recent TechCrunch piece by Kate Park, but the underlying visit and the pieces of it stretch back over the last couple months, so I'm treating this as a look-back-and-assess, not a this-morning bulletin. Huang spent two days in Tokyo, July 15th and 16th, courting basically every name that matters in Japanese industry. And he walked out with a stack of deals: a national AI factory, partnerships with the top robotics companies, and agreements with the chip-material suppliers that feed Nvidia's next generation of chips.
His message, and it's a pivot worth noticing, is that AI's next chapter isn't chatbots. It's factory floors, robots, and machines. Physical AI. And he wants Japan to build it. There's a nice bit of history in here too: thirty years ago a five-million-dollar Sega investment helped keep a near-bankrupt Nvidia alive. Now Nvidia and Japan's industrial giants need each other again.
Let me give you the three pieces, because each one tells you something.
Piece one is called Noetra. This is Japan's sovereign-AI play. The country does not want to run its factories and its robots on American AI or Chinese AI. So the government pulled together roughly forty-four domestic firms, with SoftBank, Sony, NEC, and Honda at the core, to build its own AI for machines. Tokyo's committing up to one trillion yen, that's about six point two billion dollars, over five years. And here's the beautiful contradiction at the heart of it, and TechCrunch nails it. Japan wants to own the software brain. But the hardware to build that brain still comes from Nvidia. There's a great line in the piece, a little garbled in the copy but the meaning's clear: Japan's push for independence, at least for now, runs on American silicon. Nvidia's building a "Vera Rubin AI factory" there, a massive data center with its next-gen chips, expected to launch in 2028.
Piece two, the robotics coalition. Fanuc, Yaskawa, Kawasaki Heavy, Fujitsu, Hitachi, NEC, Sony, SoftBank, Kubota, all of them lining up to build on Nvidia's Cosmos models. And Nvidia sweetened it in Tokyo by unveiling Cosmos 3 Edge, a version that runs on their Jetson Thor chips right inside the machines themselves. Huang's quote: "The next frontier of AI is in the physical world, and this is a once-in-a-generation opportunity for Japan. Japan invented modern manufacturing. Now, it has the opportunity to reinvent it for the age of intelligent industries."
Piece three, Toyota. Cars and physical AI. Toyota's using Nvidia chips across the stack, into the manufacturing, into the vehicle software, into systems that read road traffic. Interesting note: Toyota's cars will run advanced driver assistance, steers and brakes but still needs a human, which is a more conservative approach than Waymo or Tesla.
Now why does this matter to you? Here's the number that reframes everything. Japan, facing a shrinking workforce, wants ten million AI-equipped robots across eighteen sectors by 2040, backed by sixty-five billion dollars in physical-AI investment. And their national strategy, released in March, aims to grab more than thirty percent of the global AI robotics market by 2040, a market they value at around a hundred thirty-three billion dollars.
So here's the strategic read. Everybody in the founder world is fighting over the software layer, the chatbot, the agent, the coding assistant. And meanwhile Jensen Huang is quietly running the same playbook country by country. Taiwan, a homecoming keynote and fried chicken. Seoul last fall, a fifty-thousand-GPU deal. Now Tokyo, over skewers and whisky in a Kanda izakaya. He's not selling to one company. He's selling to entire national industrial bases, and he's positioning Nvidia as the mandatory foundation underneath the physical world's transition to AI. Sovereign AI is the pitch, but the sovereignty runs on his chips. That's the deepest moat in the entire industry and almost nobody's talking about it because it doesn't look like a chatbot. If you're building anything that touches robotics, manufacturing, or the physical layer, the tectonic plate is moving here, not in the model-benchmark leaderboards.
Let me hit you with a few quicker ones before we wrap.
Netflix. There's a recent regulatory filing that finally put a price on something. Netflix paid five hundred eighty-seven million dollars in cash for InterPositive, the AI filmmaking startup co-founded by Ben Affleck. They'd announced the acquisition back in March but never disclosed the terms. And here's the delicious irony. When Affleck announced it, his statement said he wanted to, quote, "protect the power of human creativity." The tools help filmmakers fix footage in post-production, missing shots, background replacements, bad lighting. And in the same breath, Netflix said in its most recent earnings that around three hundred of its titles have already used generative AI. So, "protecting human creativity" is now a five-hundred-eighty-seven-million-dollar line item on an AI acquisition. File that next to Nolan's glass horse. The industry is publicly skeptical and privately writing nine-figure checks. Both things are true at once.
Next, a builder tidbit that I love because it's the kind of thing you only catch if you're paying attention. Simon Willison pointed out that if you've got Claude Code installed, you are already running software that uses a new, unreleased version of Bun, the JavaScript runtime, that's been rewritten in Rust. Now I promised you I wouldn't drag you through the plumbing, so I'll keep it to the point that matters: your dev tools are shipping unreleased infrastructure to you inside other products before it's officially out. The stuff under your feet is changing faster than the changelogs can keep up. If you build on these tools, know that the ground is moving even when nobody sent you a release note.
Then there's the robotaxi fight, which is a recent TechCrunch Mobility piece from Kirsten Korosec, though the underlying hearings and incidents trace back over the past month, so I'll frame it as a state-of-play. The short version for founders: Uber and Waymo, former partners, are now on opposite sides of autonomous vehicle rules in Washington D.C. Uber opposes a proposed bill, argues it would hand Waymo a monopoly and displace human drivers, and is pushing a "hybrid" model where robotaxis have to run on a ride-hailing network alongside human drivers. Insiders say the hybrid idea has little chance. And here's the punchline that shows why incumbents love writing the rules: the bill requires a hundred-eighty-day, two-hundred-fifty-thousand-mile testing threshold, and Waymo has already blown past it. So if it passed as written today, Waymo walks in with a six-month head start baked into law. That's the lesson. Regulation isn't neutral. Whoever's furthest along wants the finish line drawn exactly where they're standing.
And one more from that same mobility roundup, because I found it genuinely useful and it's a counterweight to all the doom. There's a piece by Tim De Chant, looking back at how much EV fast-charging has improved. In 2023 he had a miserable road trip, broken chargers, three calls to customer service. This summer, a six-hundred-mile run to Montreal, nearly flawless. And the data backs the anecdote: the U.S. has more than doubled its DC fast chargers since 2023, and reliability jumped from around 85 to the mid-90s on Paren's index. Why do I mention it? Because it's a reminder that infrastructure problems that everyone swears are permanent do get solved when there's competition and money flowing in. Two years of steady work turned "charging sucks" into "charging's mostly fine." Keep that in your back pocket next time someone tells you a hard problem is unfixable.
Alright, let me leave you with the thread that ties this whole menu together. Today was really one argument seen from five angles. Who do you trust to own the future, and on whose terms? Bdeir says not Silicon Valley. Nolan says the public already doesn't trust the people handing it over. Apple's using the courts to make OpenAI's future more expensive. Jensen's making entire countries dependent on his chips while selling them "sovereignty." And Netflix is quietly buying the very thing it publicly claims to be protecting us from. The glass horse is on the doorstep, everybody can see the Greeks inside, and the fight isn't about whether to let it in. It's about who gets to hold the reins once it's through the gate.
That's the show. I'm Tony DeLuca, this has been Barely Possible, and I want you to stay curious and stay skeptical in exactly the healthy proportions Nolan was talking about. Take care of each other out there, and I'll see you next time.