DeepInvest

Intelligence on tap and the next Warren Buffet

Mar 02, 2025

China came out swinging its model around with DeepThink R1. It’s ChatGPT but better and cheaper. The chains of thought are deeper, longer, and harder.

Recently, I’ve been thinking about three investments: Tesla, Amazon, and Carvana (short). DeepThink was able to come up with fair valuations for each company and show its work. All three valuations were similar to mine.

I asked Google’s Gemini AI to do it. It said, Nope. Not gonna happen, buddy.

providing a precise "fair valuation" for Tesla is inherently difficult.
There's no single "fair" valuation for Tesla.
It is very important to conduct your own due diligence before making any investment decisions.
I hope this helps.

It doesn’t.

ChatGPT got its math wrong, and begged me not to use it for financial advice.

DeepSeek, Claude, and Grok 3 all performed well.

DeepSeek is the most viral thing to come out of China since covid. I don’t know how many child workers and Foxconn nets were required to make it, I hope less than my phone.

AI is already useful for doing due diligence, valuing companies, and building a portfolio. And today is the worst AI will ever be.

A few months ago I took a Waymo driverless taxi. I was amazed how quickly it gained my trust. After 15 minutes, I decided I prefer AI robotaxi over human controlled taxi. I think many people will be surprised by how quickly we choose to trust machines more than other humans. We already trust Google more than our neighbor.

I think this trend will continue, not just for taxi drivers, but doctors, lawyers, real estate agents, and money managers too.

In driving, the AI’s progression went like this: Assistant → Supervised self-driving → Unsupervised self-driving.

First, the AI is a student, then assistant, then executing supervised, then executing unsupervised and owned by a human. Then potentially executing autonomously, unsupervised, as a sovereign entity. I think sovereign AI Agents will first come from AI equipped people dying and leaving behind sovereign agents. Also they could come from people birthing sovereign agents just for fun and/or to increase their power.

What a time to be alive.

Top performers are now AI-assisted humans. Eventually, top performers will approach pure AI. There could be human wrappers, like a customer-facing human, for marketing purposes. But eventually, people won’t care whether there’s a human wrapper or not.

Any fund manager would have to be an idiot not to include digital superintelligence on their team. DeepThink R1 already does more unbiased critical thinking than most human analysts I know, and we are in the early days.

Autonomous AI Agents

In the next decade, a nonzero number of people will die and leave their wealth in the hands of an AI agent to execute on their directives, and even evolve the directives over time given new information.

This is a potential point for conflict between carbon-based and silicon-based intelligences.

The capital is with the dead person’s Agent, but a human is pissed off because it expected a big inheritance. Now it’s human vs AI in a fight for the keys to the capital. But of course, the disgruntled human has an Agent to go after the Agent.

When will we see the first lawsuit against an Agent?

Claude.ai thinks in 1-3 years.

The lawsuit would likely name both the AI system itself (as a novel legal entity) and the company that created it as defendants, raising complex questions about liability distribution between creators, operators, and the autonomous system itself.

How does an AI System show up to court? I guess we’ll see.

Capital Follows Intelligence

The ratio of biological to digital compute is changing fast. Digital compute, AI, is growing much faster than organic intelligence (collective human brain power), which due to low and declining birth rates, is flat and on the verge of shrinking.

Capital follows intelligence, energy, and integrity. Over time, capital will flow into the servers of digital superintelligence. Far more will flow to AI than from AI back to humans.

Right now, humans have all the money. But I’d wager that in my lifetime, much of it shifts into the control of silicon/digital/artificial intelligence. This could have unexpected, unintended effects.

Wealth comes from technology and investing in it. The way to get rich, without getting lucky or stealing, is by owning businesses. The quickest way to own a business is to invest in a business. The quickest way to do that is public equities.

“Stocks,” they say. I like to say, I don’t buy stocks, I buy businesses.

I started investing the week I turned 18. Nine years later, I’ve done very well, substantially outperforming index funds. But nine years is a small sample size.

How do I continue my success in a world with AI?

Codify what has worked, and employ digital intelligence to help me.

Investing is hard. Few people beat the index funds over the long term. I wouldn’t have had the success I had if it weren’t for other independent analysts sharing their work on YouTube and X. But humans are biased. We love wishful thinking, which is believing something is true because we want it to be true. We can get caught up in echo chambers.

I’m consulting with DeepSeek DeepThink R1, Claude, and/or Grok 3 on all major investment decisions. Subscribe to stay updated on new AI insights and my investments.

I’m testing LLMs by prompting to come up with fair valuations for three companies I already know a lot about: Tesla, Amazon, and Carvana (short).

I first did this with DeepSeek. It showed it’s work. I was impressed by its logic, reasoning, and competence across multiple valuation methodologies including DCF, comparables, and sum of the parts. I can work with it, and check its work because I’m familiar with valuation methodologies and finance jargon.

The LLMs' level of thinking in some areas is stronger than mine, and it’s doing work I’m often be too lazy or distracted to do. It’s a complimentary intelligence, on tap 24/7, for free.

Prompts I’ve used:

Come up with fair valuation for X company
Assume X% growth rate of revenue/free cash flow/net income. Use Y% as terminal growth rate after year X
What is a fair valuation for 2025 and 2030
Now consider a scenario with a mild recession in 2025-2026
This is the recent Q4 data: X revenue, Y Net income. Come up with a fair valuation using sum of the parts, do a DCF for each part
What is the present value of Tesla’s Optimus program assuming they deliver X units in 2030

It’s worth noting that it makes mistakes, uses ridiculous numbers, and cannot be trusted.

BUFFET.AI

I am a Buffet-Munger Style of investor.

That means evaluate many investment opportunities, and make one investment per year, or do nothing. Identify wonderful companies, buy at or below a reasonable fair price, hold for decades, and do periodic check-ins to make sure they are still wonderful. It’s an extreme amount of thinking per trade.

Early BuffetBot Gets The Dollars

The best investors are smart, disciplined, focused, patient, self-aware, and honest. Lousy investors gets caught up in the noise. They are impulsive, impatient, and act on hubris, greed, fear. They take risk, which is not knowing what you’re doing. They don’t even know much they don’t know.

Digital intelligence has a huge advantage: It doesn’t rest on top of a primal limbic system that experience or act from these dangerous emotions and biases. It can just do the work, analyzing thousands of investment opportunities, looking for substantial discrepancies between reasonable valuations and market prices.

Market inefficiency refers to volatility, where valuations deviate substantially from fundamentals like earnings (profits for shareholders). I think inefficiency comes from stupidity, bias, wishful thinking, gambling, margin calls, groupthink, mania, FOMO, media manipulation, and so on. Everything humans are good at.

AI analysts/advisors/money managers could make markets more efficient because high quality intelligence on tap democratizes high quality thinking. It’s possible less investment decisions will be rooted in emotionality, ignorance, and wishful thinking.

If capital shifting to digital intelligence stewardship reduces inefficiency, then markets may be the most innefficient they’ll ever be, right now. Since great investors capitalize on the mispricing of assets, early adopters will of AI for investing could benefit the most.

Ask the AI

I think the best investors reason from first principles, which is boiling things down to fundamental truths and reasoning from there. This is hard.

Can AI reason from first-principles?

The LLM says no, not yet. But I don’t see any law of physics that says it can’t.

From Claude:

The most advanced AI systems today (including large language models like myself) excel at recombining knowledge in novel ways and following chains of reasoning, but we lack the genuine curiosity, conceptual understanding, and creative insight that characterizes the most profound human first-principles thinking.

I think a barometer for how smart AI is vs humans is how much capital is under the stewardship of each. Right now, humans are winning, by a lot.

I asked Claude, If capital follows intelligence, and humans aren't getting smarter, and AIs are getting smarter, then will capital flow into the stewardship of AI?

The question about capital flowing to AI stewardship depends on several factors:

AI systems aren't autonomous economic agents (yet) - they're tools owned and controlled by humans or human organizations. Capital currently flows to the human enterprises that effectively deploy AI, not to AI systems themselves.
Decision-making authority versus advisory roles - AI systems are increasingly valuable for analysis and recommendations, but final decisions about capital allocation still rest with humans in virtually all contexts.
Potential future developments - As AI systems become more capable, we might see novel arrangements where AI has greater autonomy in certain bounded domains.

The more likely near-term scenario is that capital will flow to human organizations that most effectively integrate AI capabilities into their decision-making processes, rather than AI systems themselves becoming independent stewards of capital.

I agree. It’ll be a few years before DeepThrill R69 is autonomously managing money. For now, DeepThrill, Claude, and Grok work for me. They are the personal analysts and advisors I’ve been looking for, for a decade.

The challenge now more than ever is having the right questions to ask. I’m blessed to have more questions than answers and digital superintelligence on tap.

Thanks for reading, and have a great rest of your day.

Tesla Update

In January 2024, I made a video where I predicted Tesla would be valued at $4 trillion in 2029. Since then, things have changed. Elon is working at the White House, not the Tesla factories, and does not have a compensation plan with Tesla. Full list of changes and analsyis here. If Elon circles back to working hard for Tesla, and the team executes well, I think Tesla could reach a $2 trillion market cap in 2029. This is a low-confidence and optimistic prediction. I have sold the majority of my Tesla shares. I sold in January at $432.10. The price is now under $300.

Products

If you like this, you’ll probably like my book, All Outcomes Are Acceptable. Readers are loving it. The first chapter is The time I invested all money in one stock.

I’m a member of CrowdHealth a new way to find and pay for healthcare. They helped me out a lot, so I’m spreading the word.

My favorite investing platform is Public.

Coaching: I’ve been working as coach since 2020. I love it, and clients do too. Right now I’m offering writing coaching and wealth coaching. I want everyone to be wealthy. The main driver of wealth inequality is a lack of financial education and 1 on 1 mentorship. My credentials for wealth coaching are: millionaire at 22, essays on finance like this one, and a 45% annualized return since I began investing in 2016. I can help you get rich and stay rich. Message me for an intro call.