Money For the Rest of Us

Speaker discusses AI productivity tools, token economics, and stock market valuation risks

25 min Listen to the original
Artificial IntelligenceProductivityStock MarketValuationToken Economics

Summary

The speaker describes building a portfolio analysis tool using AI, contrasting his past coding limitations with current capabilities. He characterizes AI as a general-purpose technology, advocating for human-AI collaboration where AI handles research and coding. He notes that his high-volume usage of Anthropic's API suggests the company loses money on his subscription. Finally, he warns that current analyst forecasts for earnings per share growth are historically high, suggesting potential mean reversion in future returns.

Episode sections

  1. 0:00 – 4:03

    Intro and 401(k) Tool Build

    Speaker describes achieving 10x productivity by leveraging AI to build a portfolio analysis tool he originally sketched in 2012. He contrasts his past inability to code with his current ability to use AI to create deterministic financial tools.

  2. 4:03 – 11:59

    AI as General-Purpose Technology

    Speaker defines AI as a general-purpose technology like electricity or the internet, citing Ada Lovelace and JCR Licklider. He argues for human-AI collaboration where AI handles background research and coding while humans provide direction and judgment.

  3. 11:59 – 18:15

    AI Adoption and Token Economics

    Speaker analyzes AI demand vs. supply, noting he uses 640 million tokens weekly on a $100/month plan. He highlights that API pricing would cost him ~$2,200/month, meaning Anthropic loses money on his subscription, and discusses the high cost of inference.

  4. 18:15 – 21:22

    Stock Market Earnings Risk

    Speaker warns that analyst forecasts for 21.6% EPS growth are four standard deviations above the historical average of 12.4%. He cites a study showing that after periods of high EPS growth, subsequent decade returns are typically low (~3%) due to mean reversion in P/E ratios.

  5. 21:22 – 25:28

    Listener Survey and Trust in AI

    A listener survey shows 77% use AI for investing, but 71% say it hasn't changed how they learn. Speaker introduces an

Do your own research

  • Verify the historical average EPS growth rate of 12.4% cited by the speaker using long-term market data.
  • Check current analyst consensus forecasts for S&P 500 or broader market EPS growth to assess the 21.6% figure.
  • Review Anthropic's pricing page for API costs versus their subscription plans to verify the $2,200/month estimate.
  • Look for academic studies on mean reversion in P/E ratios and their impact on long-term returns.
  • Assess the methodology of the listener survey mentioned regarding AI usage in investing.
  • Verify the speaker's claim about Ada Lovelace and JCR Licklider's contributions to AI history.
  • Check the current token usage limits and costs for major AI providers like OpenAI and Anthropic.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): This year, I am 10 times more productive than I've ever been, and I'm producing my best work ever. I'm going to show you what I've done and how you can do the same in this episode, episode 568 of Money for the Rest of Us is a personal finance show on money, how it works, how to invest it, and how to live without
0:19
Unidentified speaker (speaker_0): worrying about it. In April 2012, I left my investment advisory firm after 17 years, and I launched some websites. But about six months later, I drove from Idaho down to Las Vegas and I attended the Amazon Web Services Invent Conference. I think it was the first one they ever did.
0:40
Unidentified speaker (speaker_0): And in that conference, Jeffrey Bezos spoke. And let me share some notes that I took from November 2012. If you're watching the video, you can see my my handwriting is horrendous. But what did he say? He talked about a flywheel and that we should focus on things that never change. And he
1:00
Unidentified speaker (speaker_0): said that innovation takes really a pioneer mentality, a high rate of experimentation. And I've always been one to experiment. In fact, the reason I went to that conference is I wanted to launch a website, a service called 401k Second. Opinion, where I would analyze or our company would analyze
1:19
Unidentified speaker (speaker_0): portfolios that people had. They could take their 401k and it would break it down and look at their what the valuation is, their allocation, the expected return, just get a really good understanding of their investment portfolio. But I never built it. In fact, I I I worked on some prototypes, but I didn't have the
1:39
Unidentified speaker (speaker_0): skill set to build that. And I didn't want to go out and hire a developer. I I was had still wasn't quite sure what I wanted to do after leaving my investment advisor career. But I was I was doing some programming on the side. It wasn't very good, and it was well beyond my skill set. But then eventually I launched the podcast and I and I did
1:59
Unidentified speaker (speaker_0): other things. But this month, in fact, in the past month, starting in October, late September, I built this portfolio analysis tool, similar to what I sketched back in 2012. But today I can build it myself. We're currently testing these portfolio tools with our cohort members as well as some Select Plus
2:19
Unidentified speaker (speaker_0): members, but we plan on rolling this out to money for the rest of those plus, including it as part of the membership next month. But it allows individuals to upload their portfolio, all their accounts, and they can do it through a spreadsheet, they can do it through statements, they can do it by hand, they can set up different portfolios, it will combine them, and then it calculates the expected
2:38
Unidentified speaker (speaker_0): return, the expected risk, how it would have done in different scenarios, such as the 2008 financial crisis. It breaks out the asset mix and calculates its regional weights, how much growth versus. value, various portfolio statistics like the dividend yield and the price to earnings ratio relative to the market. The
2:58
Unidentified speaker (speaker_0): tool allows individuals figure out their risk capacity, their risk aversion. They can do analysis of spending in retirement or how much are they saving and how much will their portfolio grow based on various ways to look at it. Finally, the portfolio analysis tool
3:18
Unidentified speaker (speaker_0): compares your portfolio to the overall market so you can get a really good understanding of how you're invested in terms of your allocation, in terms of your concentration. It will spotlight things you might consider in terms of maybe you're you're too illiquid or too two overly concentrated in and one holding. It compares your portfolio
3:38
Unidentified speaker (speaker_0): to the broad. stock market. How where are your overweights relative to the market in terms of value, growth, small cap, valuation? How concentrated are you in in the top five names or the top ten names of your portfolio? What's your portfolio's duration or interest rate sensitivity relative to the bond market? How credit
3:57
Unidentified speaker (speaker_0): sensitive is it? And what are things you should consider, at least compared to our views?
4:03
Unidentified speaker (speaker_0): as outlined in the strategy report, what we've been recommending individuals consider as they invest. Now I mentioned I built this over a month and I used AI to do that, but I just didn't turn it over to AI. And then this is really important when we think about a general purpose technology like AI. A general
4:23
Unidentified speaker (speaker_0): purpose technology is something that can be used in many different ways. Electricity is a general purpose technology. And the internet, you think of all the different ways to use the internet, that is a general purpose technology. And that's what AI is. And as we uh have seen AI develop over the last four years, it you know sometimes we it gets
4:43
Unidentified speaker (speaker_0): overhyped. But the key is how do we use these general purpose technologies to become more productive? to do better work, the best work we've ever done? There was a recent piece by Walter Isaacson in the New York Times, and he he quoted Ada Lovelace and JCR Licklitter.
5:03
Unidentified speaker (speaker_0): And Lovelace wrote this back in 1843. She worked with Charles Babbage on his analytics machine, and she talked about, as she saw these inventions, that humans could work well. with machines. And Lick later put it as they could cooperate in making decisions. In other words, how instead of turning everything over to AI or the
5:23
Unidentified speaker (speaker_0): internet, how can it enhance our creativity? How can we work to magnify our impact and to do the best work we've ever done? The way you do that is you don't just turn it over to AI and say, write this blog post for me. You use it to do the background research. You use it to to code like I've been doing. But then
5:43
Unidentified speaker (speaker_0): you have to go back and forth all the time and say, well, that that's not really what I was after. That really wasn't what I'm we're trying to convey. And this is what's different is we we have these experiences we've had over time. I've been investing for 30 years as an institutional asset manager. I've been teaching individuals. And so I have a point.
6:03
Unidentified speaker (speaker_0): of view that I can work with AI to convey in terms of what we built with these portfolio tools. We leverage all the work we did on asset camp. And my approach to investing is very different to someone else's. And so the tools we built are not focused on stock picking, it's focused on asset allocation. It's
6:23
Unidentified speaker (speaker_0): based on expected returns for the stock market using our build. building blocks approach. And so it's unique to us, but we leverage AI to do it. And I think of all the other ways that I've been leveraging AI in terms of putting this podcast together, in terms of just researching what's going on with asset classes, the history.
6:43
Unidentified speaker (speaker_0): Because we build asset camp, we have 50 years of stock and bond data from MSCI and Bloomberg that we can leverage.
6:51
Unidentified speaker (speaker_0): to really understand the fundamentals of markets. And so that's kind of how we're using AI and leveraging it based on my personal experience. And so you need to think about what path have you taken and how can you leverage AI and leverage other general purpose technologies to enhance your work, to focus
7:11
Unidentified speaker (speaker_0): on you, not to focus on AI. And that's when it comes to new technologies, one of the things that we do is we tend to to put them on a pedestal. Like this is just so amazing. And we bow down to it rather than letting it fade into the background as we use it more. So here's some examples. So back when
7:31
Unidentified speaker (speaker_0): electricity came out, it was it was an amazing invention. And the terms used were electric light. And the and the term electric light at least based on on writings, peaked in 1895. Now no one says electric light. We we don't emphasize, well, that runs by electricity. It just happens. Same for the internet.
7:51
Unidentified speaker (speaker_0): When the World Wide Web came out, the the the terms information superhighway that peaked in 1997. We don't use that term anymore. Now we just say online to differentiate from the being in person. The internet is faded into the background. We use it, but we don't put it on a pedestal. Same for website. When technology first comes
8:11
Unidentified speaker (speaker_0): out, we tend to capitalize it. So website initially was two words and it was capitalized. Internet was capitalized with a capital I. We don't do that anymore. It's just, it's a lowercase, just a general purpose technology. The cell phone's the same way. When it came out, it was a cellular telephone. And then for a while it was
8:31
Unidentified speaker (speaker_0): a smartphone, but now it's it's a phone. And maybe we differentiate between cell phone, but it it fades into the background. That's what AI needs to do in our lives. These portfolio tools, they were built with the help of AI, but they're not Are run on AI. They're run on deterministic formulas based in the browser. The only time AI is
8:51
Unidentified speaker (speaker_0): used is if a particular investment isn't in our database. And then AI in the background will search for it and categorize it. That's it. This is not an AI first tool, and it fades into the background. And that's how we put ourselves in front. Our work, what makes us unique?
9:11
Unidentified speaker (speaker_0): We use the general purpose technologies to enhance our. humanity. And that's what we need to do with all these general purpose technologies. Before we continue, let me pause and share some words from this week's sponsors. If you considered what is the size of your digital footprint, think of all the stuff you share on social media, the
9:31
Unidentified speaker (speaker_0): apps you use, your emails, text, online forums, forum shopping. All that information's out there, and it's vulnerable for data brokers to buy. or scrape in order to sell it. That's why I use Delete Me. And Delete Me knows your privacy is worth protecting. I've used Delete Me since September 2023. I was just on and updated my
9:51
Unidentified speaker (speaker_0): profile. Delete Me has removed 106 listings that had me or a family member and removed it. They send me a personalized privacy report, and that's where I saw that. This year they haven't had to remember. remove very much because they've been protecting me for three years. So take control of your data and keep your private life private by
10:11
Unidentified speaker (speaker_0): signing up for Delete Me now at a special discount for our listeners. Get 20% off your delete me plan when you go to www.joindelete me dot com slash David20 and use promo code David20 at checkout. Only way to get 20% off is to go to www.joinete. www.joindeleteme.com slash David20 and enter code
10:30
Unidentified speaker (speaker_0): David20 at checkout. That's www.joindeleteme.com slash david twenty code David20.
10:39
Unidentified speaker (speaker_0): AI isn't just opening new business opportunities, it's changing the way businesses get work done. We're certainly seeing that in our business. And that's the idea behind Net NetSuite Suite Next. It's the next generation of the business management suite trusted by over 44,000 companies, with AI woven into every tool you already rely on.
10:59
Unidentified speaker (speaker_0): Netsuite Next surfaces insights when you need them, puts agents to work on routine tasks and problems, and lets you ask complex questions about your business in natural language, just like talking to a colleague. But technology is only part of the story. AI is changing how people and teams work together. And that's why I want to tell you
11:19
Unidentified speaker (speaker_0): about a new expert-created business guide that you should download today. It's called Aligning for the Agenic Era: How AI is changing everyday work. And right now, it's absolutely free. In this guide, you'll see how AI could reshape work across your organization and how you can prosper from it. If your revenues are at least in the seven figures,
11:39
Unidentified speaker (speaker_0): go to netsuite.com slash david and get your free guide at netsuite.com slash david. That's netsuite.com slash david, netsuite.com slash david. Including AI. One of our ongoing themes in the podcast and in our strategy report is trying to figure out well, is there
11:59
Unidentified speaker (speaker_0): overcapacity in the AI build-out? And will this AI bubble burst, impacting our investment in terms of a falling stock market or or or other areas we're investing? And one way I did that is like, okay, I'm a heavy AI user. How many tokens am I using per week? So I have the Claude 5X max
12:18
Unidentified speaker (speaker_0): plan. I think it's $100 a month. I use about 40% of my capacity. So that's $640 million tokens.
12:27
Unidentified speaker (speaker_0): per week. And a token is just sort of the base unit of of inference that's used for for AI. Yeah, I'm paying $100 a month. But if we actually priced out these tokens using Claude's API, it'd be about four to five hundred dollars per week. I'm using upwards of $2,000 of AI inference as
12:47
Unidentified speaker (speaker_0): I do research, as we develop new tools and do anything I can to allow myself more time for deep work. And so I have I have Claude can take over my computer and do manual tasks, such as organizing these slides for this video recording or formatting a page on our website. And so not giving away our
13:07
Unidentified speaker (speaker_0): humanity, but where's the drudgery and have AI do that? So again, we're focusing on doing better, deeper work than. that features our humanity. Now, Anthropic is going to do an IPO. And so we have actually got insight into their business. And when you think about it, I'm paying $100 a
13:27
Unidentified speaker (speaker_0): month for access to AI. Anthropic's losing money on me. So it probably costs $250 to $450
13:36
Unidentified speaker (speaker_0): for what, in terms of the number of tokens on I'm using. Yet at the API price, if I ran it all through the back end, it would be about $2,200 a month. And so anthropic is making a lot of money when you're running AI through the API, which is why we don't want to use too much AI in terms
13:56
Unidentified speaker (speaker_0): of our tools. We want to use it only where it's necessary because it's expensive, at least using some of their models. Now other models are coming along, but the point is, and I'm sure OpenAI is the same thing, for a hundred dollars, you can get massive computing capacity to build anything you want. What today
14:16
Unidentified speaker (speaker_0): feels like is right after I left FEG and I announced I was leaving, and I remember saying, I can do anything I want. That's what it feels like today. We can build anything we want. We're programmers now. I read a recent article by Pragmatic Engineer, and they were talking about how programming is different.
14:36
Unidentified speaker (speaker_0): Now, because you just tell AI to do something and then you check the work and you just iterate, but AI will take you step by step. And so if there's something you want to build, you can build it with AI now. SpaceX recently announced that they wanted to build a new data center, the largest data center ever, six gigawatts. It's in
14:55
Unidentified speaker (speaker_0): Cochise County, it's just southeast of Tucson. Huge. I mean six gigawatts is three times the amount of electricity capacity used in all of Cochise County. And so when we think about So AI, and just people like me, so I'm using 100 to 400 watts of AI capacity. per week.
15:15
Unidentified speaker (speaker_0): And if we think about 48 million developers worldwide, and one of the questions right now is there how much supply is coming online? How much demand is there? And so we look at right now in terms of the supply demand balance, is pretty close. And yet there's a lot more demand
15:36
Unidentified speaker (speaker_0): coming if more people act more. like me. So there's about thirty-three gigawatts of AI capacity today. In twenty twenty seven, it's estimated to be about sixty-five gigawatts of capacity. And then by twenty twenty eight, a hundred and two gigawatts of capacity. Now about two thirds
15:55
Unidentified speaker (speaker_0): is inference or the model actually doing, like working with individuals or companies versus another third is training. So when we think about is there overcapacity? I recently had a call with my sisters. We and my and my stepfather, we do a Zoom call once a month and we got to talking about AI. And my oldest sister was
16:15
Unidentified speaker (speaker_0): thrilled because she used it for the first time and she created a some pumpkin muffin And tags, just an image in AI, and she made tags out of it. And she was absolutely thrilled. Uh My other sister and her husband use it to build apps. So they're very involved in AI. Another sister's has used it a little bit in their construction business. And so it varies in
16:35
Unidentified speaker (speaker_0): terms of individual use. More and more are discovering it all the time. But there were a billion users just kind of chatting or occasionally just c A casual. That's not enough to fill the capacity. That would only be one to four gigawatts of use where. right now there's 20 to 25 gigawatts of AI capacity. Now, 48 million
16:55
Unidentified speaker (speaker_0): developers, if they're developing in sort of the same level I am, you know, that again, there's enough capacity today. If we get 100 million heavy users, so more people developing, just making things using AI, coding, because coding's the heaviest use of AI. It it really, really is. And so you need
17:15
Unidentified speaker (speaker_0): agents doing something. And when we look at this this AI capacity, it is agents. It's it's giving a task and having it do it and work on it and iterating and perhaps on all the time. And if that's the case, that would they would need 40 to 160 gigawatts of capacity, double what's forecast for 2028.
17:35
Unidentified speaker (speaker_0): But it it demands heavy use by a lot of people. And we don't know if that's going to work. And it matters because it it impacts, it's going to impact the return to the stock market. It it really is. But now you can use AI to build anything you want. And it's super cheap because of these all-in
17:55
Unidentified speaker (speaker_0): plans. So use it to learn, better your career, better your skills, and figure out what it is that you want to make. Now, when we think of the investment implications of this, I just spent a week writing a strategy report that we release monthly to our plus members explaining what happened, where are we now? Where
18:15
Unidentified speaker (speaker_0): are we heading in terms of financial markets? What are the risks? And the biggest risk right now, here's a chart. The bottom bottom-up up company-by-company company by company analysts that follow individual companies are forecast. forecasting that earnings per share growth over the next three to five years for the global stock market will be 21.6%. That's four standard deviations
18:35
Unidentified speaker (speaker_0): higher than the long-term average. They average over three to five years. Now, this is expectations. This isn't actually what it happened. The actual earnings per share growth is lower than the forecast. But the average forecast going back to nineteen 1995 ninety-five was twelve 12.4%. point four percent. We're almost double that. And that is a risk to the portfolio
18:55
Unidentified speaker (speaker_0): when you look at this chart. When we've seen super high earnings per share growth, like we've seen. So earnings per share over the past 10 years for US markets been 12.5%. Global stock market's been over 10%. And so when we've been at these levels, the earnings per share growth over the next 10 years has been.
19:15
Unidentified speaker (speaker_0): just over 3%. These are these are nominal returns. And this was this was a study I did with AI based on MSCI data. And we went back and we broke out a bunch of different markets, broke it into fifths. This is a kind of a study that would take me hours to do previously, and AI can do it very quickly. And then I just double double checked check the work. But
19:35
Unidentified speaker (speaker_0): when earnings per share growth has been this fast in the past. about 70% of the time over the next decade, the earnings per share growth has been less than 5%. Now, why does that matter? It matters because the long-term returns of stocks are driven by the dividends, the earnings per share growth, and what investors are paying for that earnings per share
19:55
Unidentified speaker (speaker_0): growth as represented by the price to earnings ratio. And so if earnings are only going gonna to come in at less than 5%.
20:03
Unidentified speaker (speaker_0): then the price to earnings ratio of stocks is going to fall and the returns for the stock market are going to be much less than they've been in the past decade. And it's all kind of linked to AI. How much will people use it, not only to use the capacity that's being built, but actually to increase the productivity and come up with new and better
20:23
Unidentified speaker (speaker_0): ideas that people are willing to pay for? Because the forces that led, and we talked about this in the strategy report, that led to this big earnings per share growth, a lot of those are reversing. Interest rates are higher than they were. The earnings per share growth jumped because corporate income tax fell. And now there's pressure to increase it, especially when we think about the Social
20:42
Unidentified speaker (speaker_0): Security Trust Fund running out of money in the US. When we look at profit margins for companies are at their highest since 1970. 1966. And labor's share of output is the lowest it's ever been. That could also reverse. And so the reality is we need greater
21:02
Unidentified speaker (speaker_0): productivity from AI in order to offset the secular forces that could put downward pressure on earnings per share growth. And that's going to impact the stock market. Now that's a little aside from what we're talking about today. But when we think about, well, how can I be more productive? How can I leverage what I already have? And let me give you another example. First,
21:22
Unidentified speaker (speaker_0): let me share. We just did a listener survey from our Monday for the rest of us listeners, and we'll look at how they're using AI. 77% use AI for investing. And more than half, they use it on a monthly basis. About 34% are using it to research funds and stocks. 86% check, you know, they check what it
21:41
Unidentified speaker (speaker_0): tells them. But interestingly, 71% say AI hasn't really changed how they've learned, which I find fascinating because it's definitely changed how I've learned because I can just synthesize much more data than I ever had to come to conclusion, my own conclusions, using it to leverage what I've done. This trust
22:01
Unidentified speaker (speaker_0): issue is really important.
22:03
Unidentified speaker (speaker_0): because what do we trust what AI says? Recently, in the last, well, this year, we started getting emails from companies wanting to buy our archives to train AI. So they wanted all of our transcripts, they wanted anything we could produce, and they were going to pay upwards of $100,000 for this. And I
22:23
Unidentified speaker (speaker_0): realized those transcripts and those archives, all the plans. plus episodes, our premium podcast episodes that have been never been made public, all the strategy reports that we've written, all the trades I've made in my portfolio, our model portfolio, and the changes, everything I've said in our Money for the Rest of Plus forum, that's incredibly
22:42
Unidentified speaker (speaker_0): valuable. How can we leverage that to help more people? And again, here's a tool I developed in a couple hours. We indexed our entire archive of money for the rest of us content. And now, as plus members, and we'll roll this out, we're rolling this out this week, you can ask it anything. And it's just like a
23:02
Unidentified speaker (speaker_0): chat bot, but it's linked to things that I have said in the past. And so, what was the last trade I made in my portfolio and why? And if you ask it that, it'll say I made, I reduced my small cap value expensive. exposure. And it it footnotes everything I've said. And you can ask it anything. And it pulls from the archives and it
23:22
Unidentified speaker (speaker_0): answers in English. In fact, one of our members I was testing it said that the answer the Ask the Archive gave was better than the original forum post that it was pulled from. And and this issue of trust is key. My daughter's a ceramicist. It's a very, very niche area. And if she just asks AI, she's not
23:41
Unidentified speaker (speaker_0): confident the answer is correct. What she would love is she has the ceramics magazines where that she trusts, if they would make the archive available so she could query ceramics questions knowing the source is trusted. And that's what we've done with this tool. You can ask investment related questions and it
24:01
Unidentified speaker (speaker_0): pulls from the archive. And again, that's available to plus members this week. So that's what we've been building. I put it out there as really encouragement. Start using the general purpose technology that's out there and figure out how you can use it to better your life, to better your career, become more productive, and
24:21
Unidentified speaker (speaker_0): just do your best work because the tools are there. And let the tool fade into the background. Don't aggrandize AI. Just use it in a way that helps you be a better person, a smarter person, more productive, more creative, and do your best work ever. This is episode
24:41
Unidentified speaker (speaker_0): 568. Thanks for listening. Everything I've shared with you in this episode has been for general education. I've not provided investment advice. This is simply general education on money, investing in. the economy. Have a great week.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights