0:00 More than seventy percent of all new code being written is now attributed to AI agents.0:05 That’s not a forecast.0:06 That’s a real number, from a real engineering team, today.0:10 Yesterday, we talked about SpaceXAI's massive sixteen-billion-dollar bet on the future of intelligence.0:16 It’s a huge number, it dominates the headlines.0:19 But the real story, the one that actually affects how work gets done, isn't happening in a boardroom.0:26 It’s happening on the ground, inside engineering teams, and it’s moving way faster than you think.0:32 The data comes from Sasha Kaletsky on X.0:34 He’s not talking about a small startup.0:37 He’s talking about his own team, and the numbers are staggering.0:40 Ninety-nine percent of their engineers are using AI tools.0:44 That's not adoption.0:45 That is saturation.0:47 And that seventy percent figure for pull requests — the actual code contributions that build and update software — that means the majority of day-to-day coding is now being done by local or cloud-based AI agents.1:00 And it's not just one or two approved tools.1:02 The engineers themselves have built over two thousand five hundred of these agents.1:08 This isn't a top-down mandate.1:09 This is a bottom-up revolution.1:11 So what are we talking about when we say "agent"?1:14 Let’s be precise.1:16 Sean Janoe put it simply this week: an agent is an AI that takes ACTIONS on your behalf.1:21 It doesn't just answer a question and stop.1:24 It performs a task.1:25 It does the work.1:26 This is the fundamental shift.1:28 Felix Rieseberg noted this change has been happening since mid-2024, but now we are seeing the full effect.1:34 The primary tool for developers is no longer the IDE, the integrated development environment where humans write code.1:42 It's the coding agent.1:43 The AI apps of 2026, he says, look nothing like the apps from two years ago.1:48 We've moved from asking for answers to delegating tasks.1:51 And this isn't just a niche for coders.1:54 The boardroom is obsessed.1:55 Rakesh Gohel says it's the main topic of conversation.1:59 Databricks just dropped a "State of AI Agents" report for 2026, based on usage data from over twenty thousand global organizations.2:07 This is MAINSTREAM.2:08 The sentiment is clear.2:09 One poll from Sean Janoe showed sixty-three percent of companies are planning to deploy agentic AI within the next twelve months.2:17 So the takeover isn't coming.2:19 It's here.2:20 The debate is over.2:21 The agents are writing the code.2:23 Okay.2:23 So if the agents are already doing the work, what's the new problem?2:28 What's the conversation that's just starting to bubble up?2:31 It’s about control.2:32 And anxiety.2:33 This is the thread Amplifi Labs is pulling on.2:36 They posted a manifesto of sorts, titled "AI UX beyond chat." And their entire premise is that for enterprises to truly adopt this stuff, you have to solve the human problem.2:47 You have to reduce user anxiety.2:49 Think about it.2:50 If seventy percent of your code is being written by a non-human entity, you're going to be a little nervous.2:56 Especially if you're the one whose job is on the line if that code has a bug.3:01 Or a security flaw.3:02 Or, you know, just completely fails.3:05 The stakes in an enterprise are way too high for "let's see what happens." This is why, as one analyst pointed out, so many companies are struggling to get beyond basic chatbots.3:16 They can't just unleash these autonomous systems without guardrails.3:20 They need to understand the use cases, they need to build evaluations.3:24 They need to trust the output.3:26 And here's the key word Amplifi Labs used: AUDITABLE.3:29 They argue that the next wave of AI products has to be built around making the AI's reasoning auditable.3:36 You need to be able to look under the hood.3:39 You need to see the steps it took, the logic it used, the data it referenced.3:43 You can't just accept the answer from a black box when you're running a billion-dollar company.3:49 You need a receipt.3:50 You need to be able to trace the work.3:53 This isn't just about better user interfaces.3:56 This is a fundamental mindset shift.3:58 It’s moving away from the "magic" of AI and toward the engineering of it.4:02 How do we build systems that are not just powerful, but also transparent and trustworthy?4:08 Because the risk is real.4:10 The same poll from Sean Janoe that showed sixty-three percent of companies are rushing to adopt agents… it also showed that only eight percent of them were "very confident" they could even identify emerging AI risks.4:23 Let that sink in.4:24 The vast majority are jumping in, while admitting they can't see the dangers clearly.4:29 This is the tension of 2026.4:31 The capability is here.4:32 The adoption is happening at a breakneck pace.4:35 But the frameworks for safety, for control, for auditability… they are lagging way, way behind.4:41 And that gap is where the real drama is going to unfold.4:45 The conversation is no longer about "Can it work?" but "How do we stop it from breaking everything?" Now, let's zoom out.4:52 From the individual developer and the anxious enterprise boardroom...4:57 to the level of nations.4:58 Because this shift to agent-driven work is creating another, even bigger, conversation.5:03 It’s about power.5:04 Specifically, AI sovereignty.5:06 Hazar Yılmaz, who tracks these macro trends, posted a thread two days ago that nails it.5:12 He says, "AI sovereignty is now well underway." But then he adds a crucial qualifier: "it is nonetheless just nascent." So what does that mean?5:21 It means countries are waking up to the fact that if all your critical work — your software development, your data analysis, your marketing, your logistics — is being done by AI agents, it matters WHO owns the agents.5:34 It matters who runs the infrastructure.5:37 Yılmaz breaks it down.5:38 Sovereignty comes down to who is offering the AI services.5:42 Who controls the inference, the part of the process where the model actually does its thinking?5:47 Who owns the APIs that everyone plugs into?5:50 And, critically, who is charging for it all based on usage, based on tokens?5:55 Think about it like oil in the twentieth century.5:58 It didn't just matter who had cars.6:00 It mattered who controlled the oil fields, the refineries, the pipelines.6:05 The value, and the power, wasn't just in using the resource.6:08 It was in controlling its supply.6:10 We're seeing the exact same dynamic play out right now with intelligence.6:15 As more and more economic activity is performed by these agents, the companies and the countries that provide the foundational models and the computing power to run them are accumulating immense leverage.6:28 Microsoft, Google, Anthropic, OpenAI, Salesforce...6:31 they're all deploying agentic systems that operate across different apps and data sets.6:36 They are building the new digital infrastructure.6:39 This is a quiet but massive redistribution of power.6:42 We’re moving from a world where value is created by human labor inside a country's borders, to a world where value is created by an AI agent running on a server rack that could be anywhere on the planet, owned by a company with allegiance to no single nation.6:59 So when Yılmaz says AI sovereignty is "nascent," he means that countries are just beginning to grapple with this.7:06 They're starting to ask the hard questions.7:08 Do we need our own national large language models?7:11 Do we need to mandate that our data is processed on servers within our own borders?7:17 How do we avoid becoming completely dependent on a handful of foreign tech giants for our basic economic functions?7:24 This isn't a theoretical debate anymore.7:26 The numbers from Sasha Kaletsky prove it.7:29 The work is ALREADY being externalized to these agents.7:32 The shift has happened.7:34 And now, everyone is playing a desperate game of catch-up.7:37 So you have this three-level game happening all at once.7:41 On the ground floor, developers are building and deploying thousands of agents because it makes their work faster.7:48 One level up, the enterprise is freaking out about how to control and audit these things to manage risk.7:54 And at the very top level, nations are realizing they might be losing control over their own economic destiny.8:01 So what does it all add up to?8:03 Here's the turn.8:04 Here's the thread that ties it all together, from a post by Jin W Jeong that just stopped me in my tracks.8:11 He wrote: "As intelligence externalises, something subtle but profound begins to shift.8:16 Humans cease to be the primary site of economic optimisation." Read that again.8:21 Humans.8:21 Cease.8:22 To be.8:22 The primary site.8:23 Of economic optimisation.8:25 This isn't about job loss.8:26 This is not another "the robots are coming for our jobs" panic.8:30 This is deeper.8:31 More fundamental.8:32 It's the idea that for centuries, the entire system of economics has been organized around making human labor more efficient, more productive.8:41 We built tools, we built factories, we built software, all to optimize what a person could do.8:47 Now, for the first time, the system is beginning to optimize for the AI.8:51 The agents are becoming the new center of the economic universe.8:55 The work is being re-architected to make it easier for them to do, not for us.9:00 Zvi Mowshowitz noted how even the way we write is changing to accommodate AI; he says 2026-era AI writing prefers long sentences with less punctuation, and so the work subtly changes to match the tool.9:13 We are no longer the subject of the sentence.9:15 We are becoming the object.9:17 This is the profound, underlying shift that nobody is talking about on the earnings calls.9:23 The move to autonomous agents isn't just a new feature.9:26 It's a re-platforming of the global economy, away from a human-centric model and toward an agent-centric one.9:33 Axl Ibiza hinted at this, noting that the dream of agents running entire companies of sub-agents will be "slower, spikier, and more data-limited" than people think.9:43 But the direction of travel is clear.9:45 The conversations this week about auditability, about sovereignty, about seventy percent of pull requests...9:52 they are all just tremors.9:54 They are the surface-level effects of this one, massive, tectonic plate shifting underneath our feet.10:00 The question for the rest of 2026 isn't "Can an AI do this job?" We already know the answer to that.10:06 The real question is, what is the purpose of human work in a system that is no longer being optimized for humans?10:13 This is the conversation that's just beginning.10:16 And it changes everything.10:18 The real challenge isn't building a better agent.10:21 It's figuring out what role a person plays in a world that no longer revolves around them.
