Something big is happening: my take on the future of AI

I’ve been thinking about topics that are shaking the world right now, from Matt Shumer’s article “Something Big is Happening,” which was read by over 85 million people in its first week, to developers mourning their craft, to enthusiastic predictions about how AI will work for us, and we’ll finally have time to live.

I’m adding my own perspective, practical experience, and a path for how to make the most of it.

What do these predictions have in common? Where do they differ? And what does it mean for you?

I’m covering why the acceleration of AI is real and unprecedented, how two parallel worlds are forming: people who use AI and those who don’t, why agency is the most important skill of the future, where three enormous opportunities lie, and what to do about it today. Think about where you stand right now, and what you need to do to turn this opportunity to your advantage.

I’m going through the sources one by one, comparing them with my own experience, and drawing my own conclusions.

Links to the analyzed sources, with the option to download them into your favorite AI tool, are at the end of this article.

1. What everyone agrees on

AI acceleration is real and unprecedented

No serious author, not even the biggest skeptics, is talking about a bubble anymore. Three factors are driving the acceleration: models keep improving, the tools around models (planning, execution, self-correcting workflows) multiply their impact, and AI labs are now using AI to develop AI (Anthropic confirmed that 100% of the code for Claude Cowork was written by AI agents; OpenAI confirmed that model 5.3 Codex was co-created by previous models).

It’s being compared to COVID, not in its mechanism of spread (human labor is far more complex), but in the exponential growth curve that we as humans naturally struggle to comprehend. At first it looks like nothing is happening, and then suddenly everything is moving unbelievably fast.

The gap is widening

Data from AI labs show that power users use AI 6× more than average users and leverage 16× more advanced features. 75% of active users say they’ve gained entirely new competencies thanks to AI. By contrast, 84% of the world’s population has never used AI, only 0.3% pay for it, and a mere 0.04% use advanced features like coding.

Macroeconomic data is starting to confirm the impact: US productivity grew by 2.7% at the end of 2025, nearly double the 1.4% of the previous decade, even as the number of open job positions grew by approximately 420,000. Top startups that used to need 5–10 developers now operate with 2–3 people whose competencies overlap (a PM who can code, a developer who understands design).

Emotions are valid

I want to emphasize that the full spectrum of emotions around AI, from total panic to denial, from excitement to grief, is completely legitimate. I experience it myself as a rollercoaster. The key is to accept those emotions and focus on what you can actually influence.

2. Where views diverge

I identify several dimensions on which the analyzed articles and reports disagree:

Time horizon: AI-friendly voices say months to 1–5 years; traditionalists say decades, uneven adoption
What to do: ome say, become AI-native now; others say, find “messy” work that’s hard to automate
Whether it’s good or bad: Some mourn the loss of craft; others see liberation from drudgery
Panic vs. calm: Some push for immediate action; others advise patience
Author motivations: Everyone has their biases, including me, since my business is helping companies with AI

Your goal should be to form your own judgment. But to do that, you need to gain your own experience. Once you discover what the latest AI tools and models can do, you’ll be far better able to imagine where the world of work with AI is heading.

This applies to leaders, managers, and executives too. You can’t lead a transformation you don’t understand. You can’t take people into a world you’ve never been in yourself.

3. Parallel worlds and the K-shaped economy

Bifurcation is real

Jack Clark (co-founder of Anthropic) predicts that by summer 2026, people working with frontier AI systems will feel like they’re living in a parallel world. I already see it today, in my team, in hiring, with clients. When I analyzed how people use AI at the AI Predictions event, the AI itself identified this pattern and called it “bifurcation.”

Miles Deutscher describes a “K-shaped AI economy” emerging by 2030: a “superclass” that uses AI to create enormous value (including new millionaires from tiny teams), and a “permanent underclass” of people whose current roles become obsolete.

Agency is the most important skill

My key insight: the fact that someone still hasn’t started paying attention to AI after 3 years of availability isn’t a skills problem. It’s a signal of low proactivity and “low agency.” Agency (proactivity, ownership thinking, solving problems without being pushed) is the single most important competency for the future. Intelligence will become a commodity. Coding will become a commodity. High agency will not.

Naval Ravikant’s definition resonates with me:

“Hire people who just solve problems without even being asked to solve the problem — they identify the problem, they go solve it, they don’t even necessarily have to update you every step of the way, they’re not asking silly questions, and they’re just coming up with solutions.”

This is my number one criterion when hiring today.

Juniors vs. seniors

Nolan Lawson describes juniors with AI tools as “people with bazooka jetpacks,” while seniors without AI are “people on a child’s bicycle.” But I add a nuance: juniors still need experience. They’ll just get it faster. My own observation of younger people: they’re entrepreneurial, they come with both digital and non-digital business ideas. They may not want the kind of jobs previous generations had. Personally, I think this applies not just to juniors.

4. What work will survive?

The complexity spectrum of work

Luis Garicano’s framework: knowledge work exists on a spectrum of “messiness” (messy jobs). Routine, easily documented tasks (enter this invoice, file this data) will disappear. Complex work involving psychology, reading situations, persuasion, understanding client needs beyond the brief, that remains valuable.

The key question: how many people will be willing to pay a premium for human service? As AI improves, that premium increasingly goes to professionals who combine domain expertise with AI mastery. Data already shows that workers with AI skills earn 56% more, and according to a PwC study, this gap doubled over the past year.

Jevons’ paradox: higher efficiency creates more work

My fundamental counter-argument to the narrative of “AI will save you 20 hours a week”: in practice, when people learn to use AI well, they don’t work less. They often do dramatically more. Lower barriers to execution mean more projects, more ideas, more ambition. At the AI Predictions event, I personally completed 40 projects in 6 weeks, all in collaboration with AI agents.

Don’t just look at what AI will replace, look at what people can do instead. In customer service, the real value isn’t in answering tickets. It’s in using the freed-up capacity to fix the processes that generate tickets in the first place. In education, it’s not about automating grading. It’s about teachers finally having time for individual discussions and personalized teaching. In our predictions, we sometimes focus too much on replacement through the lens of today’s world, and too little on the potential of what people could do with AI augmentation.

5. How AI changes leverage

AI reverses the historical advantage of large organizations

Elena Verna’s observation: it’s a bad time to be a manager at a large company who can’t work with AI, and an amazing time to be an individual professional. Small teams and bold companies that let people use the best tools will have an incredible advantage.

Examples: Anthropic built Claude Cowork in approximately 10 days with AI. Someone from Google commented that they have 10 meetings a day debating whether to build something similar. And I see more and more examples like this.

David Shapiro reframes it: stop fixating on ROI (return on investment) and focus on the “I,” because investment costs are becoming marginal. Every individual and company today has access to talent and intelligence that only Fortune 500 companies had 10 years ago.

Three areas of opportunity

Personal productivity, becoming “superpowered,” delivering dramatically better work
New products and services, from extreme cases like Base44 (sold for $80M, 90% of the code written by Claude) to internal innovations (HR creating a new onboarding product, marketing building campaign tools)
Marketing and distribution, as AI commoditizes products and services, the ability to reach customers and build relationships becomes the key differentiator

6. What we’re losing

Mourning the craft

Nolan Lawson’s piece is the most emotionally powerful: “I didn’t ask for this.” He’s a senior developer who acknowledges that AI coding tools work, and that’s what makes it so painful. But like others, he acknowledges there’s no going back.

Ann Handley offers a counterweight: she spent two years writing her book. Was it inefficient? Yes. Was it wrong? No. The very process of thinking, struggling, deciding what to say, was the point.

The noise problem

AI may enable less qualified people to produce outputs that look competent, potentially drowning out genuine expertise. I acknowledge this is real and will be frustrating, but I also point out that this has always been true: people who are good at selling sometimes beat people who are good at their craft (but can’t market themselves).

7. Predictions and time horizons

Short-term (2026): Parallel worlds will become visible and very real. Friction between the two groups will grow. Demand for AI-first/AI-native talent will explode, both for hiring and as service providers.

Medium-term (3–5 years): Gradual structural transformation of companies. J-curve of adoption. K-shaped economy of people and companies alike.

Long-term (10–20 years): The end of the economy as we know it in the sense of exchanging time for wages. Need for capital redistribution, taxing capital instead of labor, universal basic income.

My position: I lean toward urgency, but without panic. Two reasons: (1) it can go really fast and (2) even if it doesn’t, the worst-case scenario of investing in AI skills is that you’ll be excellent at working with AI. There’s no risk here.

8. What to do: practical recommendations

Have high agency – take ownership, be proactive, try things right now, don’t wait
Build cross-disciplinary knowledge – designers should learn vibe coding, developers should learn design and product thinking; this is the golden era of generalists
Learn to create and build – materialize ideas, do projects from A to Z, work on side projects (whether at work or personally). If your current job doesn’t allow you to grow, consider taking time to invest in skills elsewhere.
Experiment every day – invest in good tools (Claude, Gemini, ChatGPT, Cursor, Lovable, Replit); just pick some of the top tools from these four categories: chat tools, content creation, vibe coding, orchestration (automation). And learn to work with them well.
Don’t let your employer limit your learning – learn tools in your own time; no one can stop you from mastering Claude in the evening or trying to build a project in Lovable.
Find meaning outside of work – as you delegate to AI, think about what personally matters to you. And what gives you fulfillment beyond work itself.
Create your AI vision – every individual, team, and company should have a clear picture of how AI will affect their work, what they will do, what AI will do, and what they’ll do together. Once you realize what reality will look like, you can do something about it.
Rediscover your love of work – don’t just optimize emails and presentations; invest in becoming dramatically better at what you genuinely enjoy; find the thing you don’t mind working on Thursday evening or Saturday morning.

The ultimate goal is to use AI not just for efficiency, but for rediscovering a passion for work, reaching a state where we fulfill our talents and strengths, collaborate with AI, and work doesn’t feel like work at all.

Here are the sources I analyze:

Something big is happening – Matt Shumer: AI progress is accelerating so fast that half of white-collar jobs will disappear within 1–5 years.
Parallel worlds – Filip Dřímalka: AI is splitting the world into two parallel worlds and the decisive window is closing.
Something messy is happening – Ann Handley: don’t get swept up in the panic – when speed is cheap, judgment carries the value.
There’s a short window to get radically ahead – Elena Verna: the question isn’t whether AI will take your job – it will – but whether you’ll use the window of months you have to become AI-native.
The K-shaped AI economy – Miles Deutscher: by 2030 there will be only two classes – those who master AI, and those who are managed by it.
A New Year’s letter to a young person – economist Luis Garicano advises finding “messy” work where AI can’t replace local knowledge, execution, and relationships.
The end of the office – Andrew Yang writes with sadness about how AI is replacing white-collar workers in real time and what it means for society.
We mourn our craft – senior developer Nolan Lawson writes not a manifesto of resistance, but an elegy: we are the last generation that coded by hand.
The AI productivity take-off is finally visible – economist Erik Brynjolfsson (Stanford) presents the first hard macroeconomic data: US productivity jumped to 2.7% and AI’s real footprint appears in the data for the first time.
85% of people will be unemployable – David Shapiro builds a mathematical model of a post-work economy and concludes that mass unemployment is not a catastrophe, but liberation.
Jevons paradox for knowledge work – David Shapiro: when you lower the cost of work, demand for it explodes – AI doesn’t create less work, it creates exponentially more.
Silent sirens, flashing for us all – Anthropic co-founder Jack Clark: AI progress is nearly invisible to those who don’t work with it – and yet by summer 2026 it will create a parallel economic world.