Skip to content
Articles
ArticlesAI translation· 9 November 2026

Are We Already Fxcked and Just Haven't Noticed Yet?

by 3 min read· extended by 1 Nachtrag

In my post “2026 – the Tipping Point of the Social Innovator's Dilemma”, I argued that we are inevitably heading towards the decoupling of human value creation from the distribution of wealth. Back then (so roughly 9 months ago), that may still have sounded to some like an abstract philosophical thought experiment for the next decade.

But reality is currently overtaking our debates on the hard shoulder. We have to ask ourselves, seriously: Are we already fucked and just haven't noticed yet?

The answer lies in a development that gets little attention outside the AI labs but is increasingly costing the researchers inside them their sleep: Recursive Self-Improvement (RSI). Put simply, AI systems are starting to build smarter AI systems. The optimisation machine is optimising its own optimisation machine.

How far along this already is shows up in internal data from the AI company Anthropic. They recently measured how much of their own research and development work on future AI models is already being done by their own AIs.

Take a look at the orange bars: the work processes in which the AI takes the lead (“AI leads”) and the human merely assists or supervises.

A year ago, that bar practically didn't exist. In March 2026 it stood at a meagre 1%. By August 2026, the AI was already leading 26% of the research on its own next generation. And the curve leaves little doubt about where this is headed.

We are no longer talking about a human using AI as a better tool to get their job done faster. We are talking about an exponential infinite loop in which cognitive value creation is completely decoupled from humans. Jack Clark, co-founder of Anthropic, recently revised his already breakneck forecast for when true RSI will be reached: from 2028 to 2027. That's next year.

The “hyperdeflation” of labour costs described by Wissner-Gross is no longer a future scenario. It is happening right this very moment in California's data centres. The value of human (knowledge) work is in free fall, plunging towards nothing more than the energy cost of the servers.

Yet we keep clinging to a socio-economic rulebook from the 20th century. We debate payroll taxes, retirement ages and 40-hour weeks while, in the background, the machines are making the entire foundation of our economic model, human labour, obsolete.

I'm not here to pull the definitive master plan out of my hat. I'm here to sound the wake-up call.

We're sitting in a car heading for a cliff. And instead of yanking the steering wheel around, we're arguing about who pays for the fuel, while the car, driven by self-improving AI, is flooring the accelerator. If we don't grasp faster that we have to completely rethink the fundamental mechanics of participation and value creation, reality will run us over without mercy.

The clock is ticking. Faster and faster.


P.S. And this just landed on my news dashboard: At Anthropic, internal Claude usage now costs more than employee salaries. Yet another bad omen ...


The Social Innovator's Dilemma Trilogy

Our Society’s “Innovator’s Dilemma”DocIsInDaHouseSchwarzenegger vs. Keynes - Consequences of the Social Innovator’s DilemmaDocIsInDaHouse2026: The Tipping Point of the Social Innovator's DilemmaDocIsInDaHouse

Live long and prosper 😉✋

▶Listen