The Path to Intelligence...

Recursive Self-Improvement | AI Supremacy | Own all Futures

AI Dilemma:
The Two Timelines of Intelligence

What am I even doing guys? At this point it’s hard not to cry when I open the portfolio. But after the tears run dry and a quick coping session with my dad about not having paper-hands… I decided to just reassess the situation.

The narrative surrounding AI keeps shifting day by day with so many promises being made with few actually materializing. I look like the fool but am I the fool?

---

The New World of Intelligence

Concerns of an AI bubble have plagued the media for years. But even then hype has infiltrated society to the point where people stopped trying to solve real problems, instead banking everything on AI to fix them.

Most of us still scoff at the idea, although 3 years later things are a bit different. Now talks of the Singularity are more mainstream and with AI becoming a threat to national security, it’s hard to see this “bubble” ever bursting?

Governments can’t stop pouring liqiduity into AI since the cost of losing is too great. Currently China and the US are the only ones who can afford to wage this war… While everyone else just watches….

---

Real Intelligence vs Today’s AI?

My problem here is that we’re not optimizing for intelligence, we’re optimizing for consumer AI and somehow this is still a geopolitical war.

The labs themselves aren’t even leveraging their own intelligence to get ahead. They're chasing benchmarks and user growth instead, leaving it to everyone else to figure out what AI is actually good for.

My friend insists that if the models just keep getting better at math and coding, recursive self-improvement will be possible “AI training itself”. And only then will deployment truly begin.

Crack intelligence → Crack everything.

And I’d believe him, I partially believe him already but OpenAI and Anthropic aren’t even buying infrastructure… They’re buying media companies?

---

What Makes Intelligence?

Maybe I’m just too stupid to believe that math and the internet alone is enough to build Superintelligence. In fact I refuse to believe it.

I’d argue the greatest breakthroughs in both understanding intelligence and its deployment came about from tinkering with small problems and games. Google’s DeepMind is actually the best example of this.

Chess → Go → StarCraft → Protein folding??? 🤯

They trained a system on every human that ever played on Go, made AlphaZero which put two models against one another until they became unbeatable. And then created Alphafold.

A powerful AI system that predicts 3D protein structures from its amino acid sequence. Helping scientists understand how proteins work, speeding up biology and drug discovery.

---

The Two Futures I see…

The nerds will say this is more machine learning than AI but it’s still the best example we’ve got of this field actually materializing into something. And I know Mythos and GPT 5.6 are too powerful for us normies but then again…

While there have been breakthroughs. Where are the AlphaFold level breakthroughs?

I’m not going to drag this out… There are really only two futures.

• Future 1: Recursive self-improvement (RSI) happens. Intelligence compounds on itself. Whoever owns the thing; infrastructure, compute and leading models wins big.

• Future 2: RSI doesn’t happen, at least not anytime soon. AI continues improving, but mostly as an incredible tools for experts. And companies with distribution, engineering talent and cash flows slowly integrate it across their businesses.

If Future 1 doesn’t happen… There’s a big chance of a major corrections or even a crash coming but at the same time, I can’t imagine an AI winter taking place. And if it ever did, it would probably be breif.

---

So What Will Happen?

I don’t know.

I'm no longer trying to predict whether AGI arrives in 2028 or 2040. I'm trying to own companies that benefit across a wide range of outcomes. Which sounds stupid against the tradtional metrics: Cash flows, and MOATs etc.

I think both Future 1 and Future 2 are equally real so it’s a mistake to bet against AI nonetheless. I don’t care what Burry says or does, it’s just dumb. The AI capex is the problem, it looks directionless especially from Zuck the Schmuck.

But in case you haven’t noticed, the big companies are trying to diversify away from their big claim to fame. Google is trying to pivot away from Search, Meta is trying to escape facebook’s shadow and Amazon is barely about e-commerce at this point.

So it’s either full AI accelerationist investing that wins, trying to time:

Chips →Fabs →Utilities →Cooling →Memory → Cloud →Grid → Power →Metals (Copper & Aluminum) → Water.

Or trying to perfectly capture the future macro of super proficient tools.

---

My Move?

As I’ve said before, the irony is that the companies building the intelligence aren’t the ones making the most money. The companies powering them are. That can’t remain true for Future 1. You want both the iPhone and its charger.

I have to play the game, we all do. And these are the picks I’m monitoring: