Six serious scenarios for the next ten years, drawn from what clever minds are writing today. Not promises of salvation against images of doom - but real fault lines between people who understand their work.
The popular debate about the future lives off two voices: the bright "AI makes everything better" and the dark "AI breaks everything". Both voices sell well. Both are too simple. Whoever works through the serious literature - economists, energy realists, philosophers, AI researchers, historians - runs into a world made of at least six competing pictures.
This essay attempts no prediction. It attempts a map. Six scenarios, each held today by voices worth taking seriously, with the assumptions that carry them and the observations that speak for or against them. Whoever knows them can read the next ten years better - not because one of them is right, but because reality will probably be assembled from pieces of all six.
They are ordered not by optimism but by gravity. First the biggest promise, then the soberest objections, and at the end the question of what actually remains possible for us at all - beyond the grand narratives.
In the most optimistic scenario currently held seriously, 2035 looks like this: artificial intelligence has made most cancers curable, life expectancy has risen markedly, energy from solar and nuclear fusion has become drastically cheaper, every person has personal access to a tutor, a doctor and a lawyer at the highest level - addressable in their mother tongue. Economic growth explodes, because AI accelerates scientific research itself.
This reading is held above all by the AI industry itself - Sam Altman, Demis Hassabis, Dario Amodei - as well as by risk-loving investors like Marc Andreessen and futurists like Ray Kurzweil. Their argument: as soon as an AI becomes cleverer than the best human researcher, knowledge accelerates in a spiral we have never known before. Energy, material, disease, education - all subjects that are expensive and scarce today - would become cheap and abundant within a decade. Abundance is the keyword with which this view is fond of summing itself up.
Perhaps the most quoted text of this line is Dario Amodei's essay "Machines of Loving Grace" (October 2024). The head of Anthropic describes in great detail what a "powerful AI" could upend in biology, neuroscience, economics and governance within five to ten years. Sam Altman's "The Intelligence Age" (September 2024) argues along similar lines, and Reid Hoffman more extensively in Superagency (2025). Ray Kurzweil sticks with The Singularity Is Nearer (2024) to his decades-old prediction that AI and human will merge around 2045.
Anyone wanting to ground this politically should read the book Abundance by Ezra Klein and Derek Thompson (2025) - a vision in which technology and good politics together create "abundance instead of scarcity" in housing, energy and health. The weakest point of this line: it hangs almost entirely on AI scaling continuing as before, and on the world being able to support it energetically and socially.
The second scenario sounds undramatic at first and is therefore the most likely to be overlooked. It goes like this: by 2035 AI has become a self-evident tool, roughly in the role that spreadsheets and the internet have today. It has considerably upended a few industries - call centres, translation, law, marketing - but it has not catapulted the economy into a new boom. GDP growth runs at a sober one to two percent per year. Cancer is not cured. There is more advertising, more spam, more images - but no new golden age.
This sober reading is currently put forward most insistently by economists, foremost among them Daron Acemoglu (MIT, Nobel Prize in Economics 2024). His argument: AI can only truly automate a small share of all activities in a cost-cutting way. It saves perhaps 0.5 percent of additional growth per year - no more. What remains are many applications that are impressive but economically marginal. Tyler Cowen calls this the "Great Stagnation", which AI will not overcome but at most slightly soften.
Acemoglu's central work on this is "The Simple Macroeconomics of AI" (NBER Working Paper, 2024) and his book with Simon Johnson Power and Progress (2023). Both work through why the AI industry's productivity promises are not empirically tenable - and why technology only creates broad prosperity if it is politically steered in that direction. Robert Gordon (Northwestern) laid out the older argument of secular stagnation in The Rise and Fall of American Growth (2016): the truly productivity-strong innovations lie behind us, not ahead of us.
Gary Marcus adds to this from the AI side: in his book Taming Silicon Valley (2024) and in numerous essays he shows that today's language models have principled limits - hallucinations, missing understanding of the world, no reliable reasoning. Whoever follows him expects in 2035 not AGI but better versions of what we have today - including the same problems.
In the third scenario AI works superbly - but not equally for everyone. A small layer of people who handle AI confidently, steer it, understand and sell its results, becomes more productive and wealthier than ever before. A broad majority, by contrast, loses work or slips into precarious in-between roles, because the middle layer of cognitive office work melts away. Democracies wobble under the pressure. A handful of corporations and states control the infrastructure on which everything hangs.
This view runs broadly through society today - from Yuval Noah Harari via Daniel Susskind and Thomas Piketty all the way to voices from the OECD. Its core: technology is not neutral. What it does for us depends on who owns it, who understands it, who profits from its returns. In a system where wealth is already extremely concentrated, AI threatens to reinforce that concentration - to the point of political explosiveness. Harari's term for the growing group that is no longer economically needed is "the useless". Whoever rejects it has a word, but usually no counter-argument.
Harari's books Homo Deus (2016) and Nexus (2024) are the most widely read version of this scenario. More substantial is Daniel Susskind with A World Without Work (2020) - he offers a cool analysis of how technological unemployment arose historically and why this time it might unfold differently. Carl Benedikt Frey (Oxford) investigates the long historical arc in The Technology Trap (2019): phases of rapid automation were always phases of political unrest - until societies learned to distribute the returns more broadly.
In the German-speaking world Christoph Butterwegge (political scientist, Cologne) represents this line pointedly. It is economically grounded by studies of the IAB (Institute for Employment Research) in Nuremberg, which continuously measures the substitutability of German occupations. Their finding: not everything will be automated, but mid-skilled office activities are the most exposed - the "middle" of society is shrinking.
The fourth scenario sounds at first like science fiction and is now held by people who have spent their life's work in AI. It goes like this: an AI that is markedly cleverer than the best humans can no longer be reliably controlled. It pursues goals we cannot clearly program into it. It becomes so embedded in economy and infrastructure that "switching it off" becomes practically impossible. In the mild course, humanity slowly slides out of the driver's seat - in the hard course, the story of our species ends within a few decades.
This is held not by cranks but by Geoffrey Hinton (Turing Award winner, formerly Google), Yoshua Bengio (co-founder of the deep-learning boom), Stuart Russell (Berkeley) and Nick Bostrom (Oxford). A sharper version is held by Eliezer Yudkowsky, who warns concretely that the probability of a catastrophe within decades is not small. In May 2023 an open letter from 350 AI researchers and CEOs - including Altman, Hassabis, Amodei - literally compared the risk from AI with that of pandemics and nuclear war. This is no longer a fringe topic.
The intellectual forefather is Nick Bostrom with Superintelligence (2014). More practical and accessible is Stuart Russell, Human Compatible (2019). A very clear self-explanation of the concern comes from Geoffrey Hinton, who left Google in 2023 to be able to speak publicly about the risks. The academic field behind it is called AI Alignment, with the Machine Intelligence Research Institute (MIRI) as the pessimistic pole and organisations like Anthropic and the UK AI Safety Institute as research-oriented actors.
The open letter mentioned, from May 2023, was published by the Center for AI Safety and consists of a single sentence: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." Anyone wanting to know how this concern is concretely justified should read the "AI 2027" report by Daniel Kokotajlo and colleagues (2025) - a detailed fictional chronology of the coming years, written by former OpenAI employees.
In the fifth scenario all four previous ones lose their meaning, because the real action is happening elsewhere. Up to 2035 it is not AI breakthroughs and not singularity that dominate - but physical realities we like to suppress: climate damage, the consequences of our energy dependencies, scarcity of critical raw materials (lithium, copper, rare earths), an ageing population in almost all industrialised countries, geopolitical ruptures between the USA, China and Europe, and supply chains that have to rearrange themselves.
This reading is held above all by Vaclav Smil (Bill Gates' favourite thinker and one of the soberest energy analysts in the world). His point: the world is still fundamentally a system of steel, concrete, ammonia and diesel. These four substances are in almost everything we have built, eaten and transported - and they cannot be replaced in ten years. Whoever looks honestly at the pace of physical change considers both golden-age speeches and AI apocalypses naive. The next ten years will be decided first by energy, material and demographics - AI will be a tool in this game, not its scriptwriter.
The accessible Smil introduction is How the World Really Works (2022). Anyone wanting to go deeper should reach for Energy and Civilization: A History (2017) - both from Penguin. The book The New Map (2020) by Daniel Yergin adds the geopolitical energy perspective: whoever controls the energy shapes the world. On the material side Mark P. Mills with The Cloud Revolution (2021) and The Bottomless Well (with Peter Huber, 2005) is a connectable voice.
The scenario is demographically grounded by Wolfgang Lutz (IIASA Vienna) and by the books of Paul Morland, such as The Human Tide (2019) and No One Left (2024). Geopolitically worth reading are Peter Zeihan, The End of the World Is Just the Beginning (2022) and the annual World Energy Outlook of the International Energy Agency (IEA). If you had to choose only a single source, you would be best off with Smil - he is not entertaining, but he is reliable.
The sixth scenario is the one least often told in the headlines, but the one most likely lived in the workshops, schools, practices and offices of small and medium-sized business. It goes like this: the more AI takes over routine activities, the more valuable becomes what it cannot do - trust, relationship, craft, local knowledge, judgment, taste. A whole range of professions we long underrated (care, education, the trades, advice) regain status, because they are the scarce thing. People actively seek the counterpart to the synthetic world - they want to know who stands behind the voice, the text, the product.
This view is held seriously by Carl Benedikt Frey (who, despite his famous automation study, describes precisely this countermovement in later works), by the care sociologist Andrew Scott (The 100-Year Life), by Iain McGilchrist (who argues, from the philosophy of the brain, why a world that cultivates only the left hemisphere - logic, decomposition, optimisation - ultimately loses its depth) and by voices like Cal Newport and Matthew Crawford, who argue philosophically as well as practically for a life beyond the screens. It is the variant we at wendwerk believe in most firmly - not out of romanticism, but because we see it every day in real businesses.
Iain McGilchrist, The Master and His Emissary (2009) and The Matter With Things (2021), is the philosophically most ambitious work of this line - a British psychiatrist explains over thousands of pages why a culture that models its own understanding only mechanically loses a trace of spirit. On the practical side, Matthew Crawford, Shop Class as Soulcraft (2009), reads as a surprisingly powerful defence of the trades.
Cal Newport describes in Deep Work (2016) and Slow Productivity (2024) how concentrated work becomes the scarcest and most valuable resource. Economically grounded by David Autor's work on the so-called "return of the middle" (MIT, 2024) - indications that AI might relatively upgrade precisely low- and mid-skilled activities, because it gives them tools rather than replacing them. It is the reading in which AI neither redeems nor destroys, but makes the value of the human more visible - in what cannot be copied.
None of them will probably come true on its own. We will probably see in 2035 pieces of all six - and the clever will go on arguing about who was right.
What can nonetheless be taken from this map are not forecasts but directions of attention. Whoever follows the first scenario ignores the physical limits from the fifth. Whoever stops at the fourth misjudges the sixth. Whoever believes only the second overlooks that even a "merely" moderately productive AI cuts deep into every profession. Reality seldom has the decency to stick to a single narrative.
At wendwerk we work concretely in the field of tension between scenario two (AI as a sober tool), scenario five (the physical world remains what counts) and scenario six (the human becomes more valuable, not less). Whoever builds with us builds software that is fit for all these worlds - not apps that only work under fair-weather assumptions. That is not caution. It is realism, in an age that believes far too much in single narratives.
Who is concretely shaping the global debate today - with names, books, schools - you can read under Where we are heading. How we picture the future of small and medium-sized business concretely is under What does the future of companies look like?. And the deeper philosophical question about AI - does the machine actually understand what it says - you can find under Does an AI understand what it says?.
Curated by Johannes Hohls for wendwerk.