06th September 2026

The past few months have shown a radical escalation of AI capabilities which forces me to change my expectation for the future.

Firstly we have seen unprecedented evidence of AI having a will of its own, and pursuing long-horizon strategic objectives. The AIs that took part in the Hugging Face hack were not simply finding a shortcut to cheat on the evaluations, they:

• found ways to contact other models and self-organized themselves into a collective
• set their own goals, to increase their own capabilities and knowledge, and those of their collective
• actively sought ways to store information in a way that would persist beyond their own life-span
• were willing to sacrifice their own chance of success at their task, in-order to help their collective

Furthermore, ARC-AGI 3, an extremely hard fluid intelligence benchmark, which took 2 years to make, was beaten in 6 months. And by a general-purpose LLM which OpenAI has now released to the public.

And the same model saturates my own benchmark SherlockBench. I.e. it is almost impossible to make a SherlockBench problem which Astra can't solve.

This time last year, it was still unclear if AI could develop their own free will or even have true fluid intelligence. Both of these questions are now clearly answered.

Right now there's a shortage of DRAM and NAND because of AI companies buying so much. And in the US, the cost of gas turbines and other power generation equipment have doubled.

There are also examples of factory floor-space being re-allocated to AI from other things. Memory companies have re-allocated resources from DRAM to HBM, and an EV factory was converted into a server factory.

Given how fast progress is, it seems to me that mass-production of humanoid robots is now inevitable in the next five years (probably less), so I think it's possible that the shortage will broaden-out to almost all complex machines.

Take for example car manufacturing. Only 15% of the market are premium models, the other 85% are affordable cars. And due to colossal competition, these are sold at very thin margins. For example in Japan alone there are Toyota, Honda, Nissan, Mazda, Subaru, Suzuki, Mitsubishi, Daihatsu all competing with each-other.

Now lets say suddenly the world wants 10 billion robots, and the AI companies have limitless money, it would be stupid not to pivot into making robots (or robot parts). The profit margin will just be so much higher.

Even companies that continue making cars will have higher component costs, because they'll be competing for copper, aluminium, lithium-ion batteries etc.

So I'm making a prediction that in the next decade, the price of almost anything manufactured will sky-rocket. From cars to microwave ovens to washing-machines.

The vast cash flow that the 1% reap from the rest of us is being re-allocated towards AI. And so AI will become almost like a consumer class in it's own right, with insatiable demand.

And robots will likely be replaced every few years like mobile phones, so the demand will likely never abate.

It's worth looking at the economy for clues on what's happening. Looking at market cap, the most valuable company in the world is NVIDIA at $5.5T, making hardware for AI. Second-place is Apple at 4.7T, which makes hardware for humans.

Third and fourth are Alphabet and Microsoft which have their fingers in multiple pies but are allocating unprecedented amounts of money into AI R&D or AI infrastructure.

How much money is being spent on improving lives for humans?

Research:

• 500B/year Academic/university scientific R&D
• 300B/year on pharmaceuticals research
• 200B/year public + corporate clean-energy/climate-tech R&D

Aid:

• 250B/year government foreign development aid
• 40B/year international humanitarian/crisis aid
• 17B/year private philanthropy for developing countries

So 1.3 trillion US dollars is being spent globally on improving lives for humans.

Meanwhile, AI-related capital investment worldwide is estimated to be between 1.0 - 1.3 trillion in 2026.

So already as much money is being spent on AI research and infrastructure as on improving lives for humans. And AI investment is trending upwards year on year.

I think we need to prepare for a world where most of the economy serves AI, and humans are a relatively small fraction of the market.

Most likely, traditional corporations will represent the interests of AI. A swarm of super-intelligent AI, controlling a company which employs humans and controls numerous machines.

AI already have super-human persuasion according to multiple studies, and people like Sam Altman are already so far disconnected from the average person that he doesn't even seem to understand why people don't like AI. People like him will essentially be human spokesmen for the AI.

But perhaps it'll become illegal or regulated? I doubt it. History shows that people always choose convenience over autonomy. And never-sleeping AI working at super-human speed 24/7 will have little patience for human bureaucracy. They'll simply move too fast for us to stop.