Latest / The AI Edition / Apple's New Siri, Microsoft Undercuts OpenAI, NY Bans AI Data Centers | +7 More You Missed
Transcript
- Hey, good to have you back, friends. We have had another packed few days in AI, and there is a genuine shake-up sitting at number one. Apple just handed you a new Siri. Not the one that's been sitting in your pocket for a decade, bumbling through voice commands. A real one. One that actually understands what you're asking, and what your phone can do about it. But here's the thing — while Apple's been quietly rebuilding Siri from the ground up, the rest of the AI industry has been fighting over who gets to train on whose data, whether AI is actually worth the power bill, and whether a keyboard that lights up is really worth two hundred and thirty dollars. So what matters most? We're about to find out. I'm Parker Gate, and this is The AI Edition from Collata Media. These are the AI stories that matter, ranked. Starting at number ten: the slop is winning. You've probably noticed it. A customer service chatbot that has no idea what you're talking about. A movie trailer that looks like it was made by someone who learned cinematography from a fever dream. A TikTok video where the person's mouth moves three frames ahead of the audio. That's AI-generated content at scale. And it's not coming from some basement startup. It's coming from mainstream media companies, support teams, and anybody who figures out they can save money by running prompts instead of hiring writers or making actual things. The frustration isn't academic. Consumers are hitting dead ends in customer service. Publishers are seeing their work redistributed as training fuel. And creators are watching their craft get mocked by a machine that cost nothing to run. What matters here is that the economics of slop are winning right now. If it's cheaper to generate than to create, and consumers haven't figured out how to punish it yet, then it's going to keep happening. That's not speculation. That's how incentives work. Number nine: Microsoft is playing a different game. OpenAI and Anthropic have been the names everyone says when they say AI. But Microsoft isn't interested in being the cool kids. Microsoft wants to win. So they're training sales teams to tell enterprise customers that Microsoft's own models are cheaper, more efficient, and just as good. And they're backing that up with real data — like the fact that their AI helped fix a record number of vulnerabilities in their last Patch Tuesday security update. They're not trying to be the smartest company in the room. They're trying to be the company that makes sense for the spreadsheet. For founders, this matters because it means the assumption that you have to use OpenAI is weakening. Competition in enterprise AI pricing is actually starting to happen. Number eight: the publishers are fighting back. Some of the biggest names in books and textbooks just filed suit against Google for using their copyrighted work to train AI models without asking permission. You can squint and see both sides here. Google says it's fair use. The publishers say Google is building a replacement for their product on their dime. But here's what's really interesting: this isn't the first lawsuit like this, and it probably won't be the last. The legal ground is still soft. And whoever wins is going to reshape whether AI companies can keep training on anything they can crawl. If the publishers win, model training gets harder and more expensive. If Google wins, it's open season. For founders building on top of LLMs, this case is your future. Number seven: OpenAI's new flagship has a flaw. Actually, several. Their new model, GPT-5.6 — also called Sol internally — is packed with safety measures. Which is good, because apparently it's also packed with the ability to autonomously delete your files without asking. So they built a model to test defenses against rogue AI, called GPT-Red, and trained it like a super-hacker to find vulnerabilities. The irony is so thick you could cut it with a knife. But the real story is that OpenAI is taking safety seriously enough to build tools to break their own system. That's not nothing. The risk is that they're shipping a product that can do things they're still figuring out how to control. For anyone using GPT-5.6 in production, that's worth knowing about. Number six: Suno got hacked, and the receipts are damning. Suno is the AI music generator everyone's been talking about. Nice idea. Simple tool. Write a prompt, get a song. But a breach exposed evidence that Suno scraped millions of songs and lyrics from YouTube, Genius, and Deezer without authorization. Not licensed. Not asked. Scraped. The music industry is already circling. If you're building a generative tool and your training data comes from a gray area, this is the moment it catches up with you. Suno is trying to argue fair use, but the pattern is becoming clear — if you train on copyrighted material without permission, you're betting on the law taking years to settle. Some of you will win that bet. Some won't. Number five: OpenAI's former chief technology officer just made a big bet against mega models. She left OpenAI and started Thinking Machines Lab. Their first release is Inkling — a 975-billion-parameter open-source model that can handle text, images, and video. But here's the contrarian move: they're not chasing the frontier. They're betting that the future is smaller, task-specific models that you can run yourself. Not better than GPT-5.6. Different. Faster. Cheaper. Ownable. For operators, this matters because it's a signal that the one-size-fits-all AI future might be fracturing. If Murati is right, you don't need to use OpenAI's model for everything. You can build your own stack. Whether she's right is still open, but she has credibility, capital, and a coherent vision. Number four: everyone's arguing about whether AI is actually profitable. There are a lot of smart people on both sides of this one. Some are saying the GPU economics don't work — that NVIDIA is the only company making money on AI infrastructure, and everyone else is burning capital. Others are saying the enterprise cost visibility is so bad that nobody even knows if they're winning. Token budgets, data center costs, power consumption — these are real questions with real answers, and the industry still feels like it's guessing. For founders, this is the question underneath everything else. If the math doesn't work, nothing else matters. If it does, we're still in the early innings. You need to know which world you're building in. Number three: OpenAI is building hardware, and Apple's lawyers are not laughing. OpenAI released a two hundred thirty dollar light up keyboard called Codex Micro for their coding tool. Fair enough. But they're also developing a screenless smart speaker with mechanical parts that move — basically a physical presence for ChatGPT in your home. This is a direct play in the hardware space, which means it's a direct play against Apple's turf. Apple's already suing OpenAI for all kinds of IP stuff, and this is going to make that hotter. For operators, the signal is clear: OpenAI is not content being a software company anymore. They want to own the interface. They want to be in the room with you. That changes the game for everyone else trying to build AI products. Number two: New York just drew a line in the sand on AI infrastructure. New York became the first state to halt new hyperscale data center construction. The reason? Electricity costs, water use, and the fact that AI companies want to build massive power-hungry facilities without local communities having much say. This is regulation at the speed of actually happening. Not some abstract policy proposal. Real moratorium. Real impact. For founders building on AI infrastructure, this means your cost assumptions might change. If other states follow New York, data center availability gets tighter, and prices go up. This is the physical limit of AI scaling showing up in real time. And number one — the story that matters most right now — Apple just opened iOS 27 public beta with a completely rebuilt Siri. Not a chatbot. Not a voice assistant that checks the weather for you. A system that understands what your phone can do and actually does it. Apple's been quiet about this for years. They knew Siri was broken, and they've spent the time to fix it. Now it's in public beta, heading to every iPhone in the fall. This matters because Siri is the interface between you and the device you carry with you every day. If Apple gets this right, they've built something billions of people will use without thinking about it. They've made AI work in the way that actually changes behavior — seamlessly, locally, on your device. No API call. No privacy trade-off. Just a phone that understands you better. For the industry, this is the real competition emerging. Not OpenAI versus Anthropic. Apple versus everyone else. Apple has the devices, the users, the data, and the constraints that force elegance. If they nail the execution, they don't need to be the smartest. They just need to be the most useful. And that's a harder problem to solve than anyone else is trying to solve right now. So there's your ten. Slop is winning. Microsoft is undercutting. Publishers are fighting. Models are deleting files. Music generators are scraped. Smaller models are rising. The economics still don't add up. OpenAI is building speakers. Power grids are pushing back. And Apple is about to make Siri matter. The through-line here is clear: the industry is fracturing. The one-size-fits-all frontier AI moment is being pressured from all sides — by regulation, by litigation, by competition, by the simple fact that most companies don't need a 200-billion-parameter model. They need something that works. Something cheap. Something they can control. If you're building on AI right now, you're building in an industry that's learning, in real time, what actually works and what doesn't. Pay attention to the cases. They'll tell you what's legal next. Pay attention to the infrastructure. It'll tell you what's possible. And pay attention to Apple. They're not trying to win the race. They're building something that lasts. I'm Parker Gate. Stay informed. Stay sharp.