Things I believe that most people don't.

Education is a lagging indicator, assessment is the leading cause of change. The SAT shaped 70 years of American schooling, the bar exam shapes law school, and Leetcode, completely unofficially, reshaped CS education worldwide. Set the measurement, and the teaching reorganizes around it.

My past products were in the education space and I learnt the hard way that the way to change education and human capital training is by being outcome first. Polymath is me trying to control the outcome first.

IQ is in very large part a myth. It's mostly personality plus breadth of exposure, and curiosity and ambition usually play the big role in getting that exposure. Reading books plus personality traits like openness matter much more for outcomes than raw IQ does.

Almost every person who has done important things, Elon Musk, Jeff Bezos, Edwin Chen, has been self-taught and it made them very smart. Reading is the thing that makes the PayPal mafia smart.

I know 90% of my schooling and college was useless, and I studied at IIT Bombay, the hardest university to get into in India. I grew the most once I left school and undergrad and had the time to read and grow on my own. Many of the smartest people I know who went to IIT Bombay or Stanford say they did not learn much: the value was in the network and the credibility it gave.

Imagine everyone getting custom curriculums for what they're best at, being exposed to all the unknown unknowns they should know of in real time, and not wasting any time on things not useful or interesting to them. It will 10x the potential of humanity.

Cursor did not. Most B2B SaaS does not. SpaceX did, because space travel would maybe have been cracked 30 years later without it. Anthropic did, because it changed the dynamics of AI development.

I don't want to build something that someone else would build the same way I would. Let them do it instead. We have finite energy, finite youth, finite time.

Today Fox News and its counterparts influence people with partial takes, and people don't think too much for themselves when that's all they're given. No one puts in the mental compute to actually figure out which takes are right, because it's so expensive to do. AI could accelerate a lot of this.

People are often rational players when they're given all the information. Imagine every person having an AI that analyses every policy from all points of view, crowdsources new ideas from the people, and presents the top 3 options with their simulated downstream outcomes. It cuts through information overload and lets people choose without being wrong as often.

Too much wealth concentration. For critical thinking's sake, take the biggest dissent to the actual world we're in and play it out in the limit: AI does all the work. Tesla makes all the robots, OpenAI and a couple of labs make all the AI, and six mining companies hold all the resources. Those ten companies are responsible for producing everything in the world, so all the wealth accrues to them, everyone else has none, and capitalism breaks down.

Today, money is the proxy for how valuable the things people do are. Soon we will need an AI to attach that value to people instead, and allocate resources and status based on it, for people to still have meaning in the world.

This also ties into whether you measure people well, and whether you help them develop based on that measurement. This is what I want Polymath to become post-AGI.

It could very well be the reason for humanity not surviving. We don't stand the max chance of annihilation from a meteor, we stand it from too fast, unchecked progress. Maybe if someone showed the dangers of AI as viscerally as nuclear bombs, it would change how the world moves.

Ambition for the sake of ambition is not good either: in a post AGI world where people can't do anything to change how things work, ambitious people may see the most existential crisis.

Everything I've learned about how to have incredible taste comes down to this: you have to be very concrete. Never say things in a way that sounds generic, in a way where people see it and go "oh, I already knew this." It has to be concrete, and the conclusions should always be inferred, never stated. Being concrete evokes very vivid imagery, and that vivid imagery is what makes people so much more invested.

And taste is not about ability, it's about proclivity. Once you feel an affinity towards artistic things, you start paying more attention, you get more exposed, and you develop great taste. So pay attention to the writing that makes you feel deeply. Think more deeply about other people's emotions. Notice the art you love, and ask yourself why you love it. And embody it. Dress better. Make the things around you beautiful. Taste builds much faster when you live it, not just admire it.

Each universe may be a sub-simulation run by another universe: a way to simulate that same universe in a coarser, faster way and make predictions about its own future. If every civilization runs simulations like this, the purpose of each one may be to find a way to survive in your own universe before the death of your stars.

We still have an infinite regress problem here. I want to believe in a God, so let's just say at the top is God.

Pascal has this thing called Pascal's Wager. If you believe in God and you're right, you go to heaven. If you believe in God and you're wrong, you seem like a bit of a fool. If you don't believe in God and you're right, you don't gain anything. If you don't believe in God and God is real, you go to hell. Hence, you should believe in God. That's a stupid argument. You can say that about anything: "if I don't step on this tile in the next five seconds, I'll die."

So let me present to you my Pascal's Wager. Having read so many biographies, Rockefeller, Franklin, all these people, so many of them drew incredible strength from God and their faith. I've read a book on the neuroscience of loneliness, and there are only two known fixes to loneliness: making friends, and finding God. That's why so many old folks turn to God when they're old, and it gives them a lot of peace. Even just instrumentally, believing in God will greatly help you. I choose to put aside my epistemic honesty in this one sector of life, just because life will be so much better with it.

I've taken many courses on this, done computational neuroscience research, and built spiking neural networks that do speech classification. The non-invasive kind is what interests me.

I believe in ultrasound here. We're close to a breakthrough that makes reading the brain both localizable and instantaneous. Right now fNIRS is localizable but slow, and EEG is fast but not localizable. Something has to give, and I think it gives soon.

There have been experiments in mice and worms that extend their lifespan greatly. I do think aging is a biological clock type thing, something that can be reset rather than an inevitability.

Imagine a group of AIs receiving feedback and events emitted from all across the economy, where any person can voice their happiness, their sadness or their concerns to them straight from their phone. They reason completely rationally, read all the best research on how policies have actually played out before, implement a policy, track what it does, and update it the moment it makes rational sense to.

That's the shape of it: a government that navigates around the pitfalls of traditional democracies, the slowness and the incentive misalignment.

Libertarianism has traditionally been the best because the economy is too complex and the government is inefficient and can't handle much well: no competitive pressures, misaligned incentives, and they can print more money whenever they need. A superintelligent AI government that gets granular signals from throughout the economy could solve humanity's coordination problems better than anything else.

On voting: we don't actually need people to vote for decisions or for people. We just need them to give data to a centralized authority, on whether they're happy or not, what their biggest problems are, whether they're fulfilled or not. That way the computation of solving those problems sits with the centralized authority, and not with the people themselves.

Some things to keep in mind:

• Any system without mass feedback from people risks becoming an oligarchy, where the people or AIs at the top make laws that only benefit themselves.
• An AI as the sole arbiter is a single point of failure. One person hacks it and the whole system breaks down, and a system like this will be attacked by many power-hungry actors.
• The AI being misaligned will be catastrophic here.

But if we find a resilient way to do this, we may become much more resilient to x-risks.

Philanthropies are one of the best ways to make things happen in the world that the free markets would not bring into existence. Here is the shape of a philanthropy that could do a lot of good:

• It exists for the continued survival of humanity, and does so in a liberty-preserving manner.
• AI does the superallocation of the capital, with complete descriptions and cross-questioning of its reasoning traces for each and every decision, and people can give feedback to it.
• Perfect transparency and visibility into where every dollar went and what the downstream effects were.

If enough people support this, it could have a positive feedback effect where everyone comes to think this is the best way to allocate philanthropic capital, and they give money to it instead of to their own individual causes. Weighing all the options and deciding where to put your money is computationally expensive. It's better to have an AI allocator find the best use for it.