Latest / Deep Values: Value Investing Deep Dives / Google’s 180 Billion Dollar AI Operating System Exposed
Transcript
- 0:00You know, usually when a massive company adopts like a new piece
- 0:05of technology, there is this expectation of finality.
- 0:09Oh yeah, like it's a finished product.
- 0:10Right. It's like buying a new car.
- 0:12You get the keys, you drive it off the lot, and it just, you
- 0:15know, does exactly what the manual says it's supposed to do.
- 0:17You turn the wheel left, the car goes left.
- 0:19It's predictable. It's contained.
- 0:21Exactly. But if you look at how companies
- 0:24have actually been using artificial intelligence over the
- 0:26last couple of years, that pristine new car looks a lot
- 0:31more like, well, a messy high school science fair project.
- 0:35Oh, completely. Just duct tape everywhere, Yes.
- 0:38Duct tape everywhere, wires crossing.
- 0:40We've basically been looking at an enterprise landscape that is
- 0:43essentially total chaos. I mean, every department has a
- 0:47dozen little AI projects running independently, but nothing is
- 0:50actually talking to anything else.
- 0:52It is the absolute definition of the messy experimental pilot
- 0:56phase. And I mean, for a while that was
- 0:58fine, right? Yeah, you just test the waters.
- 1:00Exactly. Companies were just testing the
- 1:02waters to see what these large language models could actually
- 1:04do, but what we're looking at today really changes everything.
- 1:09Right, because we are getting into a deep dive today on an
- 1:11absolute treasure trove of information this just dropped
- 1:14from the Google Cloud Next 2026 conference.
- 1:17Oh, it's massive. It is and our core mission for
- 1:21this deep dive. What we really want to unpack
- 1:22for you is how AI is finally like moving out out of that
- 1:27chaotic experimental phase. We're looking at how it's
- 1:30becoming the actual functioning operating system of entire
- 1:34enterprises and. You know, to understand the
- 1:36sheer gravity of this shift, you really have to look at the
- 1:39money. The numbers are insane.
- 1:41Right. This is not just a software
- 1:43update, this is a complete rewiring of global
- 1:45infrastructure and the financial stakes are just staggering.
- 1:49Yeah, I was looking at the numbers in our sources and my
- 1:51jaw hit the floor. So look at Alphabets capital
- 1:54expenditure back in 2022. They spent about $31 billion.
- 1:58She's already a lot. Right, already a massive number.
- 2:01Fast forward to the projections for this year 2026, that number
- 2:05has skyrocketed to an estimated 175 to $185 billion.
- 2:11That is an almost 6 fold increase in just four years, and
- 2:15over half of that machine learning compute investment is
- 2:18flowing directly into Google Cloud.
- 2:20And they aren't just spending that money blindly either.
- 2:22I mean, the internal reality driving those numbers is what's
- 2:25truly fascinating here. Yeah, what's going on
- 2:27internally? Well, CEO Sundar Pichai revealed
- 2:30a statistic at the event that fundamentally changes how we
- 2:33think about software engineering inside Google.
- 2:36Right now 75% of all new code is generated by AI.
- 2:40Wait, wait, 75%? So human developers are barely
- 2:43writing the core architecture anymore.
- 2:45Right. Humans are essentially acting as
- 2:47editors now. They're they're merely auditing
- 2:50the code the AI generates and that number is up from 50% just
- 2:53six months prior. So the trajectory is straight
- 2:56up. That pace is terrifying, but
- 2:57also kind of incredible. And you know, what Google
- 3:00unveiled at this conference to support that pace wasn't just a
- 3:02handful of, like, disjointed new language models.
- 3:05No, it's much bigger than that. Way bigger.
- 3:07They introduced a massive cohesive blueprint.
- 3:10They are calling it a 5 layer Agentic Enterprise Operating
- 3:14System. Quite the name.
- 3:15Yeah, it's a mouthful O We are going to walk through this
- 3:18blueprint today to show you how a $180 billion bet actually
- 3:24functions in the real world. Because if you were going to
- 3:26build an enterprise run by thousands of autonomous AI
- 3:30agents. And remember, an agent is a
- 3:33piece of software that doesn't just chat with you, it actually
- 3:36executes tasks on its own. Right, it actually does things.
- 3:40Exactly. You first have to fundamentally
- 3:42change the physical hardware those agents live on.
- 3:44OK, so let's start at the absolute bottom of the stack,
- 3:47then layer one, the foundation. They call it the AI Hyper
- 3:50Computer because if you want to build this new world, you have
- 3:53to reimagine the physical brain. And the philosophy here seems to
- 3:57have, well, totally shifted. Completely shifted Google's
- 4:00infrastructure. Lead Aminvidat laid out the new
- 4:03paradigm perfectly. He essentially said compute is
- 4:06no longer defined by a single chip.
- 4:08Compute is the entire data center.
- 4:10I'm trying to wrap my head around this because, you know,
- 4:14for a decade, Google's TPU, their custom tensor processing
- 4:18unit chip, was kind of a Jack of all trades.
- 4:22Right, it did a bit of everything.
- 4:23Yeah, one single type of ship handled both the training of the
- 4:27AI where it like reads the whole Internet, and the inference,
- 4:31which is the AI actually answering your specific prompt.
- 4:33But for this new 8th generation TPU, they did something radical.
- 4:39They split it. They split it into two
- 4:41completely distinct chips. Why break something that was
- 4:44already working? Well, it's a massive
- 4:46architectural divergent, right? And it comes down to
- 4:48bottlenecks. You have the TUAT, which was
- 4:51designed in partnership with Broadcom and that is entirely
- 4:55dedicated to training. And then you have the TPU 8 I
- 4:58designed with media tech and that is built exclusively for
- 5:01inference. You know, when I was reading
- 5:02through this in the sources, I kept thinking about a hospital.
- 5:05Hear me not on this analogy. OK, I'm listening.
- 5:07You have the medical researchers in the lab, that's the TPU 8T,
- 5:10right? They take months, maybe years,
- 5:13processing massive amounts of historical data, running endless
- 5:17trials to find a cure. Right, heavy lifting.
- 5:19Exactly. But then you have the ER
- 5:21doctors, that's ATPU 8I. They don't have time to invent a
- 5:24cure in the middle of a trauma ward.
- 5:26They need to make split second life saving decisions on the
- 5:29spot based on the patient right in front of them.
- 5:32That is actually an incredibly accurate way to visualize it.
- 5:35The training chips those researchers are all about
- 5:38massive scale and brute force. How massive are we?
- 5:41Talking well, a single super pod of the TPUAT strings together
- 5:459600 chips to deliver 121 exaflop S of compute.
- 5:51OK, pause 121 exaflop PSI mean. I know that sounds impressive,
- 5:54but what does an exaflop actually mean in practical
- 5:57terms? Right, it's a crazy word.
- 5:59An exaflop is 1 quintillion floating point operations per
- 6:02second. One quintillion.
- 6:04Yeah, essentially it's a billion billion calculations happening
- 6:07every single second. But the processing speed isn't
- 6:10even the craziest part here. The memory scale is what's truly
- 6:13mind bending. There are two petabytes of high
- 6:16bandwidth memory per pod. 2 petabytes.
- 6:19Yeah, and to contextualize that, two petabytes is enough memory
- 6:23to hold the entire digital collection of the US Library of
- 6:26Congress over 100 times. That is completely absurd.
- 6:30So the researchers basically have the entire library of human
- 6:33knowledge in their back pocket at all times.
- 6:35Exactly. But what about the ER doctors?
- 6:37The TPU 8I inference chips? Because if I ask an AI agent to
- 6:42do something, I don't want to wait for it to, you know, sift
- 6:44through the Library of Congress. Precisely.
- 6:47The ER doctor's job is supporting millions of
- 6:49concurrent AI agents for a business simultaneously, and to
- 6:53do that, the AI chips feature a dedicated collector's
- 6:56acceleration engine. What does that do?
- 6:58This specific hardware basically drops latency by five times, and
- 7:03it completely breaks through what engineers call the memory
- 7:05wall of long context decoding. Oh, I saw that phrase memory
- 7:09wall in the sources. Can we unpack that?
- 7:10Because it sounds like a massive roadblock.
- 7:12Yeah, think of it like trying to read a 500 page legal contract
- 7:16and having to hold every single clause in your short term memory
- 7:19at the exact same time. So you can like cross reference
- 7:22page four with page 499. Exactly.
- 7:25For an AI, loading all that context takes so much memory
- 7:28bandwidth that the actual thinking processor just sits
- 7:31there starving, waiting for the data to arrive.
- 7:34That is the memory wall. Oh.
- 7:35I see. So the new chip architecture
- 7:37essentially widens the pipe so the data flows instantly and the
- 7:41AI never stalls out. Wow OK so the brain is insanely
- 7:44fast but there's another piece to layer one.
- 7:47The sources mentioned these Axion N 48 ARM based instances.
- 7:53Oh yeah, for the cold starts. Right, being used to fix the
- 7:56cold start problem. What is a cold start and why
- 7:58does an AI agent even care? So traditionally to save power,
- 8:01cloud servers will spin down or go to sleep when they aren't
- 8:04actively being used. Makes sense, but if you have an
- 8:07autonomous AI agent running your supply chain and it detects a
- 8:09massive shipping play halfway across the world, it can't
- 8:13afford to wait 30 seconds for a server to boot up to take
- 8:15action. Right, it needs to act
- 8:17immediately. Exactly.
- 8:18The Axion instances ensure the agents are always awake, drawing
- 8:21minimal power, but instantly ready to execute.
- 8:24OK I see the picture now. We have this hyperfast
- 8:28specialized always awake physical brain.
- 8:31But if I'm being critical here, a lightning fast brain is
- 8:34completely useless if it's just hallucinating facts.
- 8:37Oh, absolutely. You've hit the exact flaw that
- 8:40brings us to Layer 2. Because without accurate
- 8:43context, reasoning is just guessing.
- 8:46And in a Fortune 500 business, you simply cannot afford a
- 8:49guess. Right, Which naturally leads
- 8:51into the Agentic Data Cloud. This is all about giving the AI
- 8:55what Google calls trusted context.
- 8:57Yes, crucial term there. Because the reality of
- 9:00enterprise data is that it is not neatly organized in a simple
- 9:03spreadsheet. Not even close.
- 9:05It is incredibly messy. It's locked in random PDF
- 9:08reports from 2019. It's in scanned images of
- 9:10receipts. It's spread across three
- 9:11different cloud providers. It's a data engineer's worst
- 9:14nightmare, honestly. So Google introduced the
- 9:16Knowledge Catalog. This is driven natively by their
- 9:19Gemini model. And what does it actually do
- 9:20with all that mess? It essentially ingests all that
- 9:24unstructured, messy data and maps it to business logic.
- 9:28And it does this without a human having to spend months manually
- 9:32tagging and categorizing things. The part that really stood out
- 9:35to me was how it learns the specific semantic meaning of
- 9:39words for a specific. Company, yeah, that's huge.
- 9:43So if the AI reads the word risk or net income, it isn't just
- 9:47pulling like a generic dictionary definition, it
- 9:50understands exactly what risk means to an automotive supply
- 9:54chain versus say, a retail clothing brand.
- 9:57Exactly. It creates A bespoke
- 9:58understanding of your company's reality.
- 10:01But it does this across borders too, which brings up the Cross
- 10:04Cloud Lake house. Yo man, this was arguably one of
- 10:08the most disruptive technical announcements at the entire
- 10:10event. Let's dig into this because the
- 10:12mechanics of it are wild. It's a game changer.
- 10:14Normally, if you want AI to analyze your company's data, you
- 10:17have to physically move all that data into the AI specific
- 10:20database. Right, which takes forever.
- 10:22And doing that costs an absolute fortune in what they call egress
- 10:25fees. It's like the hotel minibar of
- 10:29the cloud. Putting data in is free, but
- 10:31taking it out costs you heavily and worse, you get locked into
- 10:34one vendor. But the Cross Cloud Lake house
- 10:37uses Open data standards, specifically something called
- 10:40Apache Iceberg. Right.
- 10:42And without getting too bogged down in the code, Iceberg
- 10:45essentially acts as a universal translator.
- 10:47It allows Google's AI to reach directly into an Amazon Web
- 10:51Services server or a Microsoft Azure server and analyze the
- 10:55data right where it sits. That is wild.
- 10:57You don't move the files, you don't pay egress fees, you don't
- 11:00get locked in. OK, to really drive home how
- 11:02powerful that is, we have to talk about the frozen yogurt
- 11:05global launch demo from the sources.
- 11:07Oh. Yes, 5 minute miracle.
- 11:09Yes, the five minute miracle. So the goal of the demo was to
- 11:12see if a massive company could safely launch a new frozen
- 11:16yogurt flavor globally. Walk us through what the AI
- 11:18actually did there. OK, so the user types in the
- 11:21prompt first. The AI uses the Knowledge
- 11:24Catalog to dig through thousands of messy, hidden PDFs and
- 11:28internal memos, and in doing so, it finds a deeply buried warning
- 11:33about a soy allergen in the new flavors ingredients.
- 11:35Which a human analyst would have almost certainly missed.
- 11:38Exactly, just buried in a random file.
- 11:40Then, without moving any data, it reaches across the cloud into
- 11:44an AWS server where the company's actual customer list
- 11:47lives. Right where the data sits.
- 11:49Yep. It cross references that soy
- 11:51allergen against the purchase histories of loyal customers to
- 11:54see who might be affected. Finally, it calculates the
- 11:57potential drop in sales and outputs a highly detailed $15
- 12:01million adjusted sales forecast. And again, what was the timeline
- 12:05on that whole process? 5 minutes.
- 12:07The presenter stated that to do that manually finding the PDF,
- 12:10coordinating with the Amazon database team, having the
- 12:12finance team run the numbers that used to take five weeks of
- 12:16corporate analysis across multiple departments. 5 weeks
- 12:19down to 5 minutes. It's a staggering leap
- 12:21inefficiency, but if we step back for a second, this race is
- 12:25a massive red flag. A huge one.
- 12:27I mean, when I read that my immediate thought was panic.
- 12:30We are giving an army of AI agents instant, always on access
- 12:35to AWS customer data, internal PDFs, and the ability to write
- 12:3975% of the company's code. Yeah, it sounds like a recipe
- 12:43for disaster. A single hack could take down a
- 12:45Fortune 500 company in seconds. How on earth do you secure a
- 12:49system that has its hands and literally everything you're?
- 12:52Completely right to be terrified, and the security
- 12:54statistics shared at the event validate that exact fear.
- 12:58We've essentially entered an era of machine speed attacks.
- 13:01What does that mean exactly? Well, the exploit time for a new
- 13:04vulnerability, the time it takes for hackers to weaponize a flaw
- 13:07has dropped to -7 days. Wait -7 days?
- 13:10Meaning the hackers are actively exploiting the software a full
- 13:13week before the software company even knows the flaw exists to
- 13:16patch it. Exactly.
- 13:18And once they are inside, the time it takes for an attacker to
- 13:20hand off a breach to a secondary ransomware group has plummeted
- 13:24from 8 hours down to just 22 seconds. 22 seconds.
- 13:27Yeah, human security teams are now officially, mathematically,
- 13:31too slow to fight back. So relying on a guy in a chair
- 13:35watching a monitor is over. Which leads us to layer 3 the
- 13:39immune system, or agentic defense.
- 13:41To fix this terrifying reality, Google just dropped $32 billion
- 13:46to acquire a cybersecurity company called Wiz, and together
- 13:50they created an AI security posture management system.
- 13:53And Wiz approaches this problem with a really clever dynamic duo
- 13:58of agents. They essentially use a
- 14:00color-coded system to fight fire with fire.
- 14:03Yes, the red agents and the green agents.
- 14:05Let's breakdown how these actually work, because it's
- 14:06fascinating. So think of the red agents as
- 14:09friendly, continuous hackers on your own payroll.
- 14:12OK, they're constantly 24 hours a day attacking your own
- 14:16enterprise. They are trying to find
- 14:17bypasses, testing permissions, poking at the AWS data to see if
- 14:21they can steal it, and the millisecond red agent finds a
- 14:23crack in the armor it immediately alerts his partner
- 14:26at the green agent. And the green agent is the
- 14:28fixer. It automates automatically
- 14:29categorizes the risk level. It scans millions of lines of
- 14:32internal code to find the exact vulnerable line that the red
- 14:36agent exploited. And then rather than just
- 14:39sending an alert, the green agent writes the code to fix the
- 14:43hole and sends it directly to the developer's queue to
- 14:45approve. Yep.
- 14:47No human meetings, no urgent emails, just problem found.
- 14:51Problem solved. I have to push back here though.
- 14:54Having AI quietly writing and deploying code in the background
- 14:58to fix things sounds great, but doesn't that create the dreaded
- 15:02shadow AI problem? Well, shadow AI is when you have
- 15:06rogue algorithms running in the background doing things no human
- 15:09actually understands or approved.
- 15:11If we have agents fixing code silently, what happens when an
- 15:14agent makes a mistake and nobody even knows the agent was running
- 15:17in the first place? Right, that is the single
- 15:18biggest threat to this whole blueprint and enterprise leaders
- 15:21know it, which is why the system maps a live real time security
- 15:25graph of every single agent running in the company.
- 15:28Oh, so they map it? Exactly.
- 15:29It tracks every identity, every permission, every action an
- 15:33agent takes. So if an agent tries to access a
- 15:36payroll database, it shouldn't. The security graph catches the
- 15:39anomaly instantly. There is no dark corner where an
- 15:41agent can operate unseen. OK, so let's summarize where we
- 15:44are so far. The system has a massive
- 15:47hyperfast brain on the hardware side, it's deeply informed and
- 15:51connected on the data side, and it has an autonomous immune
- 15:54system securing it, right? But how do you and I, the actual
- 15:59human employees, command this thing without being totally
- 16:02overwhelmed? That brings us to the final two
- 16:04layers, Mission Control and the employee experience.
- 16:08Layer 4 is the Gemini Enterprise Agent Platform.
- 16:10The sources refer to this literally as Mission Control,
- 16:14and the thing that really surprised me here is that Google
- 16:16isn't forcing you into a walled garden.
- 16:18Not at all. They aren't saying you can only
- 16:20use our one omnipotent model. They're providing an arsenal of
- 16:23highly specialized tools. It's a very pragmatic approach,
- 16:26honestly. You have Gemini 3.1 Pro doing
- 16:29the heavy orchestration, managing the grand workflows,
- 16:32but for specific tasks you use specific models.
- 16:35Well, you have Nano Banana 2 for generating high fidelity
- 16:38visuals. Yeah, VO 3.1 Lite for video
- 16:40production, Lyria 3 Pro for audio generation.
- 16:43And here is the kicker, they actually integrated Anthropic's
- 16:48Clawed Opus 4.7 right into the platform, and Tropic is arguably
- 16:54one of their biggest competitors.
- 16:55And Google is letting Clawed run natively inside their own
- 16:58Mission Control. Because they know that at the
- 17:01enterprise level, companies want the absolute best tool for the
- 17:05specific job. If Claude is better at a certain
- 17:08type of coding task, Google wants you to be able to use it
- 17:11without leaving the Google ecosystem.
- 17:13But hold on, if I have all these different models running
- 17:15simultaneously, a Claude agent doing coding, a Gemini agent
- 17:18doing HRA Nano Banana agent making marketing graphics, don't
- 17:22you risk what they call context pollution?
- 17:25Oh absolutely. Context pollution is a massive
- 17:27liability. If the HR agent and the
- 17:29financial forecasting agent share the same underlying memory
- 17:32space, the HR region might accidentally leak the CEO salary
- 17:35into a public marketing slide deck.
- 17:37Which is a career ending disaster 100%.
- 17:39So how do they stop that? They built something called
- 17:41Gemini Enterprise Projects. It gives these agents high
- 17:45fidelity, completely isolated workspaces.
- 17:48It's essentially putting up soundproof, bulletproof walls
- 17:51between the departments so thousands of agents can work in
- 17:54parallel without ever cross contaminating their memories or
- 17:58accessing data they shouldn't. That makes a lot of sense.
- 18:01Which brings us to the final layer, Layer 5.
- 18:04The actual day-to-day employee experience.
- 18:07They call it Workspace Intelligence and honestly
- 18:10reading about this felt like a huge sigh of relief because it
- 18:14is finally the end of the context text.
- 18:17Oh, the context text. That is something every modern
- 18:19worker feels deeply. Oh absolutely.
- 18:21It's that feeling when you sit down to answer one simple client
- 18:24e-mail and 10 minutes later you have 15 browser tabs open.
- 18:27Yeah, you're digging through old Slack messages.
- 18:29You're checking a random spreadsheet.
- 18:31You're you're logging into your CRM.
- 18:32You forgot your password. Navigating the context of your
- 18:35job is exhausting. Workspace Intelligence fixes
- 18:38that by linking your Gmail, your Google Docs, Sheets, and Chat
- 18:42into a single unified neural network of your daily work.
- 18:46The demo for this was incredibly relatable.
- 18:49So a worker just types a single prompt into Google Chat.
- 18:52They ask for an update on a specific client.
- 18:54And the AI just figures it out, yeah.
- 18:56The AI instantly reads through their recent emails to
- 18:59understand the context. Then it reaches outside of
- 19:02Google entirely into a third party HubSpot CRM, pulls the
- 19:06latest sales data, formats it, and generates a fully branded,
- 19:10expertly designed Google Slides deck.
- 19:13All of that happens instantly from one chat window.
- 19:15And they also announced new Microsoft 365 interoperability,
- 19:20meaning you can pull data from Word or Excel effortlessly.
- 19:23It makes migrating between the two ecosystems five times
- 19:26faster. The AI just dissolves the
- 19:28friction between the tools. OK, so we have just walked
- 19:31through the entire $180 billion blueprint from the split TPU
- 19:34chips in the data center all the way to the Google Slides deck
- 19:37popping up on your laptop. But if we pull way back, what is
- 19:41the ultimate? So what?
- 19:42Right. How is it being used?
- 19:43Exactly how is this actually being used by massive brands
- 19:46right now? Well, the sources provided a
- 19:48rapid fire list of case studies that prove this isn't just
- 19:51theory anymore. Apple, for instance, has
- 19:53officially made Google Cloud their preferred cloud provider.
- 19:57That's. Huge.
- 19:58Yeah, they are using this very infrastructure, specifically
- 20:01Gemini to Co develop the next generation of Siri for billions
- 20:05of iPhones globally. And then you have NASA, they're
- 20:08using the Gemini Enterprise platform right now to prep the
- 20:11data and logistics for the Artemis, the second flight,
- 20:14which will be the farthest human Space Flight in history.
- 20:17They are trusting this operating system with astronauts lives.
- 20:21Then there is Virgin Voyages. In the commercial sector, they
- 20:24built something called Project Ruby.
- 20:26It's a highly personalized concierge agent for the
- 20:29passengers on their cruise ships.
- 20:30Did it work? Oh yeah.
- 20:32But using these specialized agents, they cut their
- 20:34production and planning cycles by 60% and it drove a record
- 20:38break, making 28% month over month sales growth.
- 20:41Incredible. But the one that absolutely
- 20:43fascinated me was in sports. Shaun White, the three time
- 20:46Olympic gold medalist, is using a Deep Mind 3D pose tracking
- 20:50model. Right.
- 20:51This shows the power of the unstructured data processing we
- 20:53talked about earlier. Exactly.
- 20:55The AI takes standard flat 2D video of a snowboarder doing a
- 21:00trick, just regular phone footage and it extracts full 3D
- 21:04flight dynamics. The exact angle of the board,
- 21:07the velocity, the rotation. That is wild.
- 21:09It is fundamentally revolutionizing how athletes
- 21:12understand the physics of their own tricks without needing
- 21:15expensive motion capture suits. It's just everywhere.
- 21:18So bringing this all back together, Google Clown CEO
- 21:21Thomas Curian synthesize this entire event, this massive shift
- 21:25into what he calls the three freedoms of the new enterprise.
- 21:29Let's lay those out for everyone #1 freedom of choice.
- 21:32You aren't locked into one piece of hardware or one model.
- 21:35You can use Google's Tpus, or you can use NVIDIA GPU's.
- 21:38You can use Gemini, or you can plug in Anthropic's clod. #2
- 21:42Freedom of data location. With tools like the Cross Cloud
- 21:45Lake house, you do not have to move your massive troves of data
- 21:48out of Amazon or Microsoft just to analyze it.
- 21:50The AI goes to the data. And #3 freedom of governance.
- 21:56The enterprise completely controls its own security
- 21:59destiny. You have those red and green
- 22:01agents acting as your autonomous immune system mapped by a real
- 22:05time security graph to prevent shadow AI.
- 22:08It is a comprehensive multi layered vision of order finally
- 22:13replacing the chaos of the last few years Which.
- 22:15Leaves you with a pretty profound question to Mull over.
- 22:18We have just explored a near future reality, a 20, 27
- 22:22enterprise where autonomous software agents write 75% of the
- 22:26code, where they fight off complex cyber threats in 22
- 22:30seconds, and where they can complete five weeks of corporate
- 22:32data analysis in 5 minutes. The messy duct tape science fair
- 22:36project is over. The hyper engineered, fully
- 22:38autonomous city grid is officially here.
- 22:41Exactly. So the question you have to ask
- 22:43yourself as we enter this new era is this.
- 22:45In a world where the enterprise is completely run by an AI
- 22:48operating system, what becomes the uniquely human skill that
- 22:50keeps you indispensable? Thank you so much for joining us
- 22:53on this deep dive into this massive architectural shift.
- 22:56We will catch you on the next one.