For years, the browser was this perfectly safe sandbox. We kept our AI models neatly contained, you know, behind digital screens. Yeah, totally contained. But right now, the tech is aggressively stepping out. It's literally taking physical control of the real world. We're watching it
run the back end of major banks. We're even seeing it... um control human hand muscles directly which is just crazy i mean we are talking about raw electrical signals bypassing our brains entirely so welcome to today's deep dive i'm really glad you're here with us yeah thanks for having me today we're looking at the landscape of may 2026. we are unpacking a massive shift happening right now ai labs are just Well, they're done waiting for users to sign up. They're completely done.
Right. They're forcibly embedding themselves into the global economy. They're entirely rebuilding their infrastructure just to handle real time human interaction and, you know, causing real world hardware shortages in the process. It's an incredibly wild time to be watching this space. The whole landscape is shifting right under our feet. We're seeing this rapid transition from like digital toys to. physical reality. And it's happening much faster than anyone modeled. Exactly.
So let's start by just following the money. Always a good idea to understand where AI is going. We first have to look at the enterprise side. The biggest labs in the world are forcing their way in. They're actively drilling into the corporate bedrock of our economy because they have these massive IPO goals they need to hit. And to do that, they basically need to be indispensable. Right. And. Subscriptions alone just aren't enough anymore. OpenAI and Tropic realized a very hard
financial truth recently. Yeah. You simply can't wait for individual people to buy a plus plan. Consumer subscriptions are just historically fickle. Users churn all the time. They do. If you want to hit... multi -billion dollar IPO numbers, you need guaranteed revenue. You need deep institutional contracts that span years. You essentially have to physically go into their offices. You have to custom build the intelligence for them. OpenAI just made a structural move
here. They quietly created something called the deployment company. Oh, yeah. It's already valued at. $10 billion. Which is staggering. Right. OpenAI owns the vast majority of the equity, but they've got serious heavy hitters backing them. TPG, Brookfield, SoftBank, they're all getting involved. That is undeniably serious institutional capital entering the chat. And their deployment strategy is, frankly, highly aggressive. They're targeting over 2 ,000 major
portfolio companies. The ultimate goal is integrating GPT 5 .4 and Codex directly into their systems. They want to fundamentally rewire how these companies operate. It's not a tool anymore. It's a completely new operating system. A totally new operating system. Yeah. And Anthropic certainly isn't sitting on the sidelines here. They formed their own private equity coalition to compete. Right. Blackstone,
Goldman Sachs, Hellman and Friedman. Anthropic is aggressively going after the midsize market. But, you know, their actual integration method is fascinating to me. It really is. They aren't just sending a download link. They're literally embedding their own flesh and blood engineers into these companies. Right. And that is a crucial distinction. They aren't just selling an API key and hoping for the best. No. They're sending highly paid human engineers physically into the
building. These engineers are actively rewriting entrenched legacy workflows. They're using Claude to rebuild operations from scratch. So they're basically hiring a totally new kind of worker, not just developers who sit in code. They need engineers who can persuasively talk to a CEO. Yeah, translators, basically. Exactly. The biggest example of this is Anthropic's partnership with FIS. FIS is a staggeringly massive deal. For context, it's a software running a huge chunk
of global banks. We're talking about the core financial plumbing of the world. Anthropic is bringing Claude directly into that sensitive ecosystem. By late 2026, it's going to be widely available across the sector. Institutions like BMO and Amalgamated Bank will be running on it. It's kind of like stacking Lego blocks of data directly into the foundation of Wall Street. That's a great way to put it. But these blocks are constantly thinking. They're actively adapting.
If one block decides to change its shape, the whole tower shakes. Beat. It's honestly a little terrifying. Oh, absolutely. I mean, I still wrestle with prompt drift myself. We all do. Getting a model to just stay on track is genuinely hard. Yet they're trusting AI to run global banks seamlessly. Well, that underlying risk is exactly why they embed human engineers. They absolutely cannot afford any prompt drift at a major bank. Right. Joint ventures secure that guaranteed revenue
stream. When you're deeply integrated into a bank's ledger, you're incredibly sticky. You don't just get canceled like a Spotify subscription. No, you become a permanent utility. Exactly. Sierra is another perfect example of this enterprise demand. Yeah. Sierra's wild. They just raised $950 million. Yeah. That puts their valuation at $15 billion. They already serve 40 % of the entire Fortune 50. Their annual recurring revenue jumped from $100 million to $150 million. Almost
overnight. It's crazy fast. It clearly proves the corporate market is starving for this infrastructure. It effectively shows that enterprise AI is not a bubble. It's a structural rewiring of how corporate infrastructure operates. The raw demand for automated workflows is just staggering. Companies are panicking that they'd be left behind. Okay, let me pause and get this straight. If they are embedding AI into the core of global banking, how does the infrastructure keep up without crashing?
Well, they realized they couldn't use the old Internet backbone. They had to completely rip out the old web architecture. They built a totally new split brain system for real time scale. So they basically built a brand new Internet just to handle the load. Exactly. And that naturally brings us to the voice infrastructure miracle. You simply can't rewrite global bank operations without flawless systems. You need split -second
reliability at a massive scale. OpenAI recently released an engineering deep dive detailing this exact flex. And the raw numbers are absolutely staggering. As of May 2026, OpenAI officially handles 900 million weekly active users. Think about the physical volume of data moving there. To make a digital conversation feel human, it has to be fast. The AI constantly has to handle unpredictable human interruptions. It has to manage conversational turn -taking seamlessly.
And traditional web setups just couldn't handle it. that kind of dynamic load. Yeah. Standard HTTP requests naturally have way too much inherent lag. It breaks the illusion of a conversation instantly. Yeah, it does. So OpenAI just ditched the old setups entirely. They moved to a custom WebRTC on Kubernetes stack. Let's define that jargon quickly. That's just tech that keeps live audio streams stable without crashing. Spot on. For you listening, think of the old internet
like sending letters. There's always a noticeable delay waiting for a response. WebRTC is like keeping an unclosable phone line directly open. That's a perfect visual. And to actually make it work, they created a brilliant split -brain infrastructure. They didn't just shove everything onto one overheating server. They divided the computational labor to mass. maximize speed. Right. They built a very lightweight component they call a relay. The relay just handles the
fast moving data packets. Its only job is to move raw audio quickly. Then. They built a much heavier stateful transceiver. The transceiver does all the actual heavy cognitive lifting. It handles the complex AI thinking and all the deep encryption. And they aggressively pushed this entire architecture out to the edge. They systematically deployed global relays at the Internet's physical edge. Which is wildly expensive. Incredibly expensive. It effectively means your
voice hits a server almost instantly. The millisecond a sound leaves your lips, it... It aggressively cuts down the jitter and the conversational lag. It happens before the data even reaches the main model. And, you know, the core models themselves are entirely different now. We're talking about GPT 5 .5 and GPT real -time 1 .5. They are entirely audio native from the ground up. This is a profoundly deep shift in computer science. They aren't translating
your natural speech into text anymore. Instead of reading a transcript, they're processing the pure sound waves. They can actually hear the hesitation or excitement in your voice. They process pure sound and speak with raw emotion directly. The median latency is currently sitting under 500 milliseconds. That is literally faster than human reaction time. Whoa. Imagine scaling to a billion queries. Two sec silence. It's genuinely hard to even wrap your head around the physics
of that. Managing millions of concurrent stateful audio sessions is computationally brutal. Stateful means the server constantly remembers the entire conversation. It holds the context open. Exactly. It holds the context completely open while you talk. Doing that seamlessly for 900 million people without dropping context is brilliant engineering. But hold on. Think about the real world impact here. With 900 million people talking to audio native AI in real time, isn't that putting insane
stress on actual physical hardware? Oh, it absolutely is. It's heavily straining the global power grid. It's gotten to the point that developers are panic buying consumer hardware to keep up. Wow. The cloud alone simply can't handle everything locally anymore. Right. The digital boom is creating a massive physical supply chain bottleneck. Exactly. And that seamlessly leads us to the profound fiction we're seeing. We are actively moving from digital infrastructure into physical reality.
The scale of pure audio models. hits the real world hard. So if OpenAI is processing 900 million live audio streams, they're shifting a massive compute burden onto the grid. And that macroeconomic cloud strain is why we're seeing a microeconomic panic at the consumer level. Specifically, we are seeing this happen with Apple hardware. Yeah, Apple didn't foresee this. They didn't completely foresee this sudden crunch. Right now, Mac Mini and Mac Studio stock is running critically low.
Software developers are aggressively hoarding these specific desks. cop machines because they desperately need them to run their local AI agents. Prices on the secondary market are already rising significantly. People intensely want to run AI locally for strict data privacy. They want to avoid that frustrating cloud latency entirely. So they're literally buying up every single Mac they can find. Industry experts think this severe
shortage could easily hit iPhones next. If the iPhone supply chain actually gets hit, that is massive. It changes consumer tech availability overnight. It really does. And it's definitely not just physical supply chain friction we're dealing with. We're actively seeing major regulatory friction popping up, too. The White House is currently weighing some very serious new AI oversight rules. Yeah, they urgently want early access
to new foundational models. The stated rationale from the administration is actually pretty straightforward. They want to manage the societal risks if things go completely wrong. It isn't really about broadly blocking AI development. They just want a thorough look under the hood before public deployment. Wait, hold on. If I'm understanding this dynamic right, the White House isn't trying to shut these local agents down. They just want a backdoor
view into them because they're nervous. They're realizing an uncensored AI running freely on a local Mac studio is a wild card. Is that basically what's driving this sudden oversight push? Yeah, because the sheer lack of visibility is deeply concerning to regulators. If we connect the dots, the entire trend makes sense. The intense consumer desire for privacy and edge computing is driving this shift. Right. Everyday people want autonomous
agents running locally in their own homes. That desire directly drives the massive Apple hardware shortage. And that decentralized deployment immediately prompts governments to seek visibility. They desperately want to know what is actually running on those machines. But let's bring this down to the listener for a second. If governments and hardware can barely keep up. How is this rapid integration actually showing up for everyday
users right now? It's actively bleeding into highly personal physical applications at an unprecedented pace. We're seeing it fundamentally shift physical wearables and daily digital tools. So the tech is already fundamentally altering our daily physical and digital routines. Completely. We're rapidly moving from macroeconomic shortages down to the micro level. This is fundamentally about what you can actually use today. The sheer speed of practical applications application development
is staggering right now. Let's actually look at the wildest edge case first, because this one honestly blew my mind when I read about it. There's a brand new AI wearable device out there right now. Well, this is crazy. It literally controls your human hands using direct electrical signals. It actively lets you perform physical skills you never learned. It's fascinating. It sends calibrated impulses straight to your hand muscles. It's almost literally like downloading
physical abilities directly into your body. Yeah. It bypasses your brain's slow motor learning process entirely. It feels very sci -fi. It feels exactly like that scene in The Matrix. Totally. You just plug a cable in and suddenly you magically know Kung Fu. Or, you know, maybe you instantly know how to play the piano. You strap it on. Yeah. Your fingers are doing complex tasks. It's fascinating, but it's also profoundly weird to think about. It definitely bridges a crazy gap
between biology and machine. But we also have very practical everyday tools bridging a similar gap. Look at what XAI just launched into the market this week. They released a wildly powerful new API specifically for voice cloning. It's wild. It lets you create. hyper -realistic custom voices. You can use them for podcasts, autonomous agents, synthetic videos. You can freely pick from over 80 distinct voices. They cover 28 different
global languages with perfect accents. And the actual pricing model is what's truly disruptive. It starts at literally just $3 an hour. It completely commoditizes high -end voice production. Yeah. Anyone with a laptop can instantly spin up a multilingual ad campaign now. You don't need a massive recording studio or professional actors anymore. Right. Speaking of the advertising world, Meta just made a huge integration play. Meta now actively lets you connect AI straight into
your ad account. You can directly plug in ChatGP to your cloud to run campaigns. The AI natively talks to your potential customers directly in the chat. The rapid adoption rate there is absolutely staggering. Weekly automated conversations jumped from 1 million to 10 million almost overnight. Wow. Small businesses are entirely letting the AI handle their sales funnels. It's negotiating, answering questions and closing deals in real time. We're also seeing an absolute flood of
rapid fire. Daily tool updates, just highly practical tools you might use every single day. There's a new one called Avatar getting a lot of traction. It perfectly removes complex image backgrounds in one single click. It automatically balances colors flawlessly. It even intelligently restores missing parts of an old image. Then you have highly creative niche things like codex pets. I love them. Right? They're totally optional, small, animated companions specifically for Codex.
They just sit quietly on your screen and show your thread status. They physically reflect whether Codex is actively running or just waiting. It has an interesting bit of personality to the coding process. There's also Droppy, which I think is a massive shift in retail power. It isn't just a basic price tracker. It's an autonomous agent fighting massive retail pricing algorithms. It constantly tracks Amazon, eBay, and AliExpress silently in the background. You just get a notification
the exact millisecond a price drops. It shifts the power completely back to the consumer. And for independent website owners, there's a huge shift with sleek analytics. It's a completely... Privacy -first encrypted alternative to Google Analytics. It aggressively offers real -time data, entirely cookie -less tracking, and fast dashboards. All these diverse tools share one major common thread. They're taking highly complex AI architecture and making it utterly simple
to use. Let's pause on that idea for a second because the societal implications are massive. With AI literally moving our hands and cloning our voices, where does the human element fit in? I think the human element fundamentally shifts up the cognitive chain entirely. As AI flawlessly handles the raw execution of tasks, we elevate. Our primary role permanently becomes one of deep curation. We handle the broad strategy and the
complex intention behind the action. We decide what needs doing and the AI natively does it. Basically, we stop doing the heavy lifting and become the directors. Exactly. We confidently call the strategic shots. We architect the vision and the machine executes the labor. Let's pull all of this incredible information together. We've covered a truly massive amount of ground today. Mid -roll sponsor Red Placeholder. We started by looking at AI labs infiltrating major
banks. They launched a massive $10 billion integration strategy to do it. They're actively securing guaranteed revenue from the global corporate economy. Then we look deep into the underlying infrastructure, making it possible. They absolutely had to rebuild internet audio architecture from scratch. Yeah. They're serving 900 million users seamlessly with a split brain system. They completely achieved sub -500 millisecond latency for pure
emotional audio processing. And that massive digital explosion inevitably caused serious physical friction. We saw severe Apple hardware shortages hitting the secondary market hard. Software developers are aggressively hoarding Macs to run local agents. And the White House is pushing hard for new regulatory oversight. And it all eventually culminates in the most personal edge cases imaginable. Wearables that literally control your human muscles with
electrical signals. XAI flawlessly cloning your exact voice for just $3 an hour. The great integration is officially no longer just a pending software update. It is a tangible, rapidly accelerating. physical reality happening right now. It really truly is. Thank you so much for joining us on this deep dive today. I highly encourage you to actively check out some of the specific tools we mentioned today. See exactly how they might seamlessly fit into your own daily workflow.
And please be sure to subscribe for our next deep dive into this rapidly changing world. It's always an absolute pleasure to critically unpack this landscape with you. There's literally always something entirely new and profound to learn. I want to leave you with one final lingering
thought today. If AI can perfectly replicate our voices with XAI and literally control our physical movements with new wearables, how long until we can't tell the difference between a skill we genuinely learned and a skill we simply rented for the afternoon? Beat. Let that sink in.