You're Reading This? Then You're Already Responsible for AI.

Opinion by Nicolas Martin, May 28, 2026

Last week, two of the biggest companies on the planet quietly admitted they had a problem. Microsoft canceled most of its internal Claude Code licenses after six months. Engineers were burning through token budgets at a pace the company simply could not sustain. Uber's CTO, Praveen Neppalli Naga, went further. He disclosed that his team of 5,000 engineers had already exhausted the company's entire 2026 AI budget, all $3.4 billion of it, by April. Four months in. Gone. Uber's COO then told Fortune he couldn't even draw a straight line between all that AI spending and a single useful consumer feature shipped faster. "That link is not there yet," he said.

Let that settle for a moment.

The most AI-eager engineers on the planet, at two of the most technically sophisticated companies in history, hit a wall that nobody saw coming. Not a wall of capability. A wall of cost, governance, and organizational common sense. The first enterprise AI correction is underway. And somehow, somewhere on LinkedIn and X, the discourse hasn't moved an inch. We're still debating whether AI is a revolution or a scam. We're still sorting people into three tired boxes.

The Trio of Fears

Here is how the world has decided to handle the AI transition: by categorizing everyone involved into three convenient villain archetypes.

There are the enriched, the investors, the founders, the VCs riding the wave and laughing all the way to the IPO. There are the illuminated, the techno-utopians who believe we are six months away from curing cancer and ending poverty through the miracle of AGI. And there are the enemies, the doomers, the Luddites, the academics who write four-thousand-word essays about what Foucault would have thought about large language models.

The result of this taxonomy is that public discourse on AI has become almost entirely useless. It produces heat, not light. Every serious point gets absorbed into one of these three categories and neutralized. If you say AI is disrupting employment, you are a doomer. If you argue we should regulate it, you are an enemy of progress. If you point out that Anthropic just raised at a $61 billion valuation, you are implying the whole thing is a grift.

The irony is that the skepticism is legitimate. The impact is real. The Microsoft and Uber situations are not glitches in the narrative, they are the narrative. This is what adoption actually looks like: messy, expensive, uneven, and full of surprises that no forecast anticipated.

But here is what I find striking. The people most loudly debating whether AI is good or bad are the same people feeding it data every single day. And most of them don't realize it.

The Smartphone Hypocrisy

You checked YouTube this morning. You sent a WhatsApp message. You scrolled Instagram for eleven minutes before your first coffee. You used Google Maps to find a restaurant. You watched a recommended video on a topic you didn't search for, because an algorithm decided you wanted it.

Every one of those actions trained a model. Fed a data pipeline. Contributed a behavioral signal to a system that is, in some form, AI or its direct infrastructure.

The idea that you can be philosophically opposed to artificial intelligence while owning a smartphone is not a position. It is a performance. A comfortable fiction that lets people feel morally coherent while doing nothing differently. I have met engineers in Bangalore who say they are "worried about AI" while spending eight hours a day prompting Claude to write production code. I have met HR managers in Paris who are deeply concerned about algorithmic bias, while using an AI-assisted ATS to screen CVs before they read a single one.

This is not a moral failing. It's a human one. We compartmentalize. We hold contradictory beliefs simultaneously because it is cognitively cheaper than resolving them. But the compartmentalization has a cost: it prevents the serious, grounded conversation we actually need to have.

There is no "against AI" position available to anyone who participates in the modern digital economy. The choice is not AI or no AI. The choice is: conscious engagement or passive drift. Informed adoption or accidental dependency. Those are the only real options on the table.

Stop the bad faith. The question was never "are you pro-AI?" The question is: what kind of AI user do you want to be?

Employment: A Fact, Not a Wish

Let me say something that should not be controversial but somehow still is: AI is already affecting white-collar employment, and denying it is not a neutral position.

Dario Amodei spent much of 2025 issuing warnings that most of his peers were too commercially cautious to voice. He told Axios plainly: "We, as the producers of this technology, have a duty and an obligation to be honest about what is coming. I don't think this is on people's radar." He predicted that AI could wipe out up to 50% of entry-level white-collar jobs within five years and push youth unemployment to 10–20%. He called it "unusually painful" disruption, bigger than any labor transition in modern history.

Then, this week, at a Commonwealth Bank of Australia conference in Sydney, Sam Altman said he was "pretty wrong" about the economic impact. "I'm delighted to be wrong about this," he told the audience. "I thought there would have been more impact on entry-level white-collar jobs by now than has actually happened."

Two quotes. Same week. Two opposite conclusions.

Who is right? Probably both, depending on your time horizon. Altman's observation is empirically defensible, the Yale Budget Lab found no significant changes in unemployment for high-AI-exposure jobs through March 2026. Amodei's warning is also defensible, tech layoffs through May 2026 have already surpassed 115,000, approaching the full-year total for 2025, with major companies explicitly citing AI as the driver. Big Tech's hiring of new graduates has dropped nearly 50% from pre-pandemic levels.

The actual truth is uncomfortable for both camps: the disruption is real but uneven, accelerating but not yet catastrophic, and structurally baked in for the medium term regardless of what any CEO says in any keynote.

The honest answer, and the one I try to give when I train teams, is this: the jobs that disappear first are the ones where human judgment adds the least marginal value. Tasks, not whole professions. But enough tasks, aggregated across enough roles, and whole professions start looking different. That's not alarmism. That is adaptation to observable reality.

Adapt or Pick Mushrooms in the Forest

There is a French expression that does not translate perfectly: se mettre au vert. Literally, it means retreating to the countryside. Metaphorically, it means stepping back from the world, opting out.

There are people who genuinely believe that the right response to AI disruption is withdrawal. Digital minimalism. Artisanal work. Growing vegetables. I respect the lifestyle choice. I do not respect it as a professional strategy for anyone who has rent to pay, a team to lead, or a career to build in 2026.

The world is going to keep developing AI regardless of what you personally decide. The models are going to get better. The cost per token is going to fall. The number of agentic workflows deployed in enterprise environments is going to increase every quarter. China just opened a factory in Guangdong capable of producing 10,000 humanoid robots per year. Shanghai has declared it wants 100,000 humanoid robots deployed in factories by 2030. The city of Hangzhou sent humanoid robots to direct traffic in May 2026. These are not experiments. These are deployments.

The strategy of waiting for this to stabilize before adapting is not a strategy. It is a countdown.

Adaptation means something more precise: prepare for the worst case, hope for the best, and build toward the version of your professional life that remains valuable regardless of how fast the models improve. Develop the skills that compound with AI rather than compete with it. Build your network like it's a survival asset, because in a disrupted labor market, it is. Invest in cognitive complexity, the kind of judgment, synthesis, and relational intelligence that no model has convincingly replicated at scale.

The cabane en forêt option is poetic. But for most of us, the only real path runs straight through the disruption.

Four Races, Not One

Everyone is talking about one race. AI versus humans. Models automating jobs. The usual apocalyptic framing.

But there are actually four concurrent races happening right now, and only people who understand all four are positioned to navigate the transition intelligently.

Race one: models against models. OpenAI, Anthropic, Google DeepMind, Meta, Mistral, and a dozen Chinese labs are locked in a capability arms race with no obvious ceiling. Each new frontier model redraws the boundary of what AI can and cannot do. Businesses that locked in workflows around GPT-4 in 2023 have already had to rebuild twice. The pace is not slowing. What looked stable six months ago is already legacy.

Race two: AI against humans. This is the race everyone obsesses over, and it is real, but it is slower and more uneven than the headlines suggest. The most vulnerable positions are entry-level, high-volume, low-discretion knowledge work, legal research, financial analysis, first-draft writing, customer support, code review. These are already being restructured. The NVIDIA VP of Applied Deep Learning, Bryan Catanzaro, said publicly this week: "For my team, the cost of compute is far beyond the costs of employees." When a tech company executive says compute is cheaper than labor, you pay attention.

Race three: Robots vs humans. This is the one most Western professionals are not watching closely enough, and it is probably the most consequential in the long run. China installed more than 295,000 new industrial robots in 2024, more than every other country combined. It now controls over 90% of global humanoid robot sales. It committed $138 billion in state venture capital to AI and robotics. AgiBot produced its 10,000th humanoid unit in March 2026. Bank of America projects 90,000 humanoid robot shipments globally in 2026, rising to 1.2 million by 2030.

The robotics race is not a future scenario. It is happening in factories in Guangdong right now. Morgan Stanley notes China is running the same playbook it used with electric vehicles: integrated supply chain, massive domestic testing grounds, brutal cost compression. The economic pressure this creates on low-wage manufacturing will not respect national borders.

Race four: humans against each other. The adopters versus the reluctant. The strategically balanced versus the recklessly enthusiastic. The people building career capital in an AI-augmented world versus the people waiting for the dust to settle. This race is quiet, unglamorous, and will not be covered by any major publication. But it is the race that will define professional outcomes over the next decade more than any model benchmark.

The advantage in race four does not go to the most enthusiastic AI adopters. The Uber engineers who consumed $2,000 of Claude Code per month and shipped nothing demonstrably better are not winning. The advantage goes to the people who adopt deliberately. Who understand what they are using. Who maintain enough critical distance to evaluate outputs rather than simply relay them. Who build skills that compound with AI, not skills that merely defer to it.

Not too much. Not too little. The calibration is the work.

You Are Already in This

Here is the conclusion I keep coming back to, and the one I want to leave you with.

You are reading this article. Depending on where you found it, LinkedIn, a newsletter, a shared link, there is a meaningful probability that the distribution was assisted by an algorithm trained on behavioral data. There is a possibility that a headline variant was A/B tested by an AI system. There is a near-certainty that some of the research supporting this piece was conducted using AI tools.

And you are fine with that. Because you are already in this. There is no reading position outside of the system. There is no clean, uncontaminated vantage point from which to observe the AI transition without participating in it. The moment you opened this article on a device connected to the internet, you rejoined the system.

This is not a condemnation. It is a description. The ethical response to being inside a system is not to pretend you are outside it. It is to engage with it responsibly. To understand what you are using, what you are feeding, and what you are building, whether you intend to or not.

The people who will navigate this transition well are not the ones with the most optimistic or the most pessimistic priors. They are the people who are honest about the complexity of the moment. Who can hold two contradictory data points, Altman saying the jobs apocalypse hasn't arrived, and Amodei warning the disruption will be unusually painful, and treat both as provisionally true, depending on time horizon and sector.

There is no formula for what comes next. Anyone selling you one is selling you something else. But there are principles. Understand the tools before you depend on them. Build skills at a level of complexity that models cannot yet replicate at scale. Maintain your network like it is the most durable professional asset you have, because in a volatile labor market, it is. And engage with AI the way you would engage with any powerful, imperfect instrument, with clarity about what it does well, what it does badly, and what you are ultimately responsible for.

Microsoft and Uber did not fail because AI failed them. They failed because they adopted without governance, enthusiasm without structure, consumption without measurement. The lesson is not "AI doesn't work." The lesson is that the adoption of powerful tools requires the same rigor as the tools themselves.

The new rules are not complicated. They are just unfamiliar. Responsible AI. Personal agency. Team-based intelligence. Complex skills that compound over time. These are not buzzwords. They are the actual playing field.

You are already on it. The question is whether you are playing consciously or just running.

Sources