AI for Smarter Investing: New Memberships & Insights | December Podcast

Welcome in everybody. Quote we've got on the screen will tie into what we talk about a little bit later today. But AI may feel inevitable, but many of the problems plaguing real-world AI applications are not new. We'll talk about that topic in depth today. And so we are glad to have you here calling this monthly podcast today. AI for smarter investing and new memberships. Get into some exciting news there. I am Kyle Guskey. I'm here with David Trainer as well. Start off at the disclosures, disclaimers. You can read these in the replay, but the main takeaway is we are not providing advice. We are publishing research. On the agenda today, this is our monthly podcast. The next one will be next year. This will be our last for the year. It seems crazy to think that 2025 is coming to an end, but next one will be January 9th, 2026. Since we last had our podcast, we published quite a bit. Some new long ideas, danger zone picks. We've closed a few positions, whether they've been long ideas, danger zones that either don't look attractive, don't look dangerous, or been acquired. There's a bunch of different reasons there. You can check those out on the website. We did finish up the earnings watch parties for the recent earnings season. David did a presentation at Harvard Business School. We've got a blog post on that. And then also a Podcast appearance on the transformative role of AI in investing, published some of the best and worst sector ETF data, updated on the Footnotes season. We're ranked number 1 in stock picking by SumZero again. That is the 54th month in a row. So we're very proud of that. And we've got lots of model portfolio updates and featured stocks we've published as well. The agenda for today, we'll talk about some upcoming training sessions. We'll go over the new memberships that are coming soon. I'll talk about machine learning and AI for smarter investing, and then get into some of the posts out in our community and some trade ideas there as well. And if you have any questions, please send us in the Q&A or the chat. We'd love to answer anything that you guys have today, or future topics, ideas you'd want us to look at. We definitely love to hear from you there. On the upcoming trainings, we do have an Ideas Dinner next week, December 16th. That is for Pro and higher members. The Pro members, if you've already signed up, you're still registered for that. If you haven't, you should get an email with a registration link for that. And then we also have a training on December 17th called Beat the Pros with the Only Proven AI. I will walk you through the Alpha Stock methodology that doesn't rely on expensive fund managers or biased Wall Street research. You should also be getting emails for that as well. And we'll have some links in our In Case You Missed It email that we send out every week. And we'll also have some registration emails going out. So definitely be on the look for those. But I think the big news of today is the new memberships that we're working on. We've always believed that high-quality institutional-grade research should be available to all investors. That's why we're developing new memberships. We can't wait to show you. We're calling them—

Good. This is a long time coming too, right? People have been telling us that they wanted something between Stock Tracker 50 and Pro, and we've been listening. Look, we really wanted to shift to more of a pro focus for a while. It was part of the strategy going back a couple of years ago, and that's why we did it. We really wanted to lean into that. And we found that while that's been a great experience for us, we really want to come back to more of our roots, is what Kyle said, which is making high-quality research available to everybody. And we thought that Stock Tracker 50 was a good way to do that, but we think coming back with some of these in-between memberships Is a great idea as well. And we're really excited to bring back the traditional Gold, Platinum, and Pro type models. We think that's really cool.

Yeah, I'm gonna actually show where you can join a waitlist to get information as well. Like I said, we're calling them Discovery and Portfolio Insights. If you go to the homepage here, and then the Get Access page, individual investors can actually get some details on them. Discovery, Portfolio Insights right here. And join the waitlist, put your email in, you'll be the first to get info as soon as we have more details, launch dates, everything. I'm gonna share that link in the chat here as well. So anyone that wants to join the waitlist, go right ahead. But yeah, as David said, it's really building on Stock Tracker 50, adding in some screening capabilities, some more data at the different levels there, and gives you a more comprehensive toolset with our research. So we're very Excited about that.

Yeah, that's the sweet spot, folks. We're going to give you a lighter version of it. And got to remember, this is the best research in the world on the companies we cover. And we say that with confidence because it's been vetted by Harvard Business School, MIT Sloan, Journal of Economics, Financial Economics, and now live traded portfolios that are dramatically outperforming the market. It doesn't get any better than that. With respect to proof. So we're excited to, to get this out there to more folks.

Definitely.

Are we going to give out discounts to people who sign up early, Kyle?

Don't give away all the good news.

Maybe we will, maybe we won't, but you only find out if you sign up early.

Yep, that would be something we would share for those waitlist people. We'll probably share with anyone on the email list, but the waitlist, you'll hear about it first. But not a bad idea. Jump back into the Slideshow here. Again, if anyone questions anything, you can put them in the chat here. You can also email us at support@newconstructs.com if you got any questions on memberships and your current membership, anything like that. Be glad to help you out there. But yeah, we're very excited about the memberships we've got coming here. But beyond that, we did want to talk a little bit about machine learning and AI for smarter investing today. As I said at the top, AI may feel inevitable, but many of the problems plaguing AI applications are not new. And you may ask, how do we know that? We actually published an ebook in 2018 with that same title, Machine Learning and AI for Smarter Investing. Put the link to that in the chat here. But we've been talking about AI long before it was the topic of discussion in the markets, because it's something we've been working on for many years now. In that ebook, we highlighted quite a lot, but some of the headlines and takeaways, the smoke and mirrors of AI on Wall Street at the time, Why AI needs to be transparent. AI's big data problem. Many of these issues still persist today. In the slideshow here, I wanted to pull some quotes from our ebook, some of the more now things that are happening, quotes from more recent articles. As the volume of data grows, most of that data lacks the structure needed for machines to analyze it. You've heard us talk about this. Teaching a machine to actually understand the data is much harder than just telling the machine to read the data. We looked at this article, Beyond the Hype: How Structured Data Can Save AI Financial Analysis, found there's studies on even the best LLMs achieved only 17% accuracy when linking numbers and text to correct concepts in US GAAP taxonomy. So you've got 2018, we're talking about data lacks structure. 2025, data still lacks structure. That's the key to building out that superior dataset. You can't just tell the machine to read it and hope that it knows what it's talking about. You've got to be able to actually endow it with that information, with that subject matter expertise, which is exactly what we've done with our roboanalyst, what we're doing with FinSights, our AI agent. But not a new problem and one that still hasn't been solved outside of New Constructs, really.

Because it's not a popular thing to do. I just did a couple interviews this morning and talking about what it is we do. People do the dirty work. They don't want to have to go through filings. They want LLMs to magically do all the work for them, just like everybody wants machines to magically do all their work for them. I had an interview or meeting yesterday with a portfolio manager. He's like, yeah, I want you to do this. And then I'm like, yeah, I bet you do. How about I do all of your job? And I also said, look, when I was at Harvard Business School a few weeks ago, I told those students there's not going to be many portfolio manager jobs left. Why are we paying people, 80% of whom Underperform their benchmark every year, net of fees. Why are they making money? They're losing you money. They're taking your money, their fees, and they're underperforming the market, right? There's so much money going to these people who actually lose money. They're going to go away, and it's because they have attitudes like that. Oh, you do all this work for me because I was showing him FinSites. I'm like, look, here's what we do. We do like this fundamental stuff. I want you to add this and this to it. I'm like, you know what? I'll take this list that we pulled out of FinSites. Really great stocks, like the best stocks in this sector. And then of those, which ones are most likely to beat earnings? And then we'll take it into ChatGPT and we'll ask which ones have the best technicals, which ones have the best insider trading. So I showed the guy like, here's how you do all that. But will you do this for me? You know what? You can do work too, man. But they don't want to do work. And we saw when we wrote the ebook, something like 70, 80% of all the capital from machine learning firms, because it wasn't AI yet, it was still machine learning. Was going into cleaning up the data. Maybe my numbers are off on that. It was a big number. Yeah, going into chart for cleaning up the data, folks. Somebody give me. I'm gonna give a. I'll give you a free most attractive stocks report to anyone who can come into the chat and tell me how do you clean data? How would you do it? Anybody want to raise your hand? We'll give you the mic. Can you tell me how you want to clean data? You just take a hose pipe out with a mop, right? That's my joke. I love that one. People don't even know what hose pipes are these days. Yeah, how do you want to clean the data? The point is you don't clean it. You go collect it right from the source. Cleaning it is comparing the stuff you don't know about, right? I got to clean it. I got to check. What do I check? I don't know. They don't tell you which data is dirty because then if they did, they could do that, then it wouldn't be dirty. They take out the dirty stuff. So they throw the dirty in with the regular. So to fix it, you got to go back to the source and validate it all. Well, that's more expensive than just going to the source and getting it right to begin with. Right? Would you rather try to use some crappy data and then go, oh shoot, I got to go to the source? So guess what? Nobody's going to the source. They're not really cleaning it. They maybe identify some stuff that's blatantly wrong that they don't even need to look at the source to know that it's wrong. But what does that mean? You just don't have data for certain companies then? Point is, there's no such thing as cleaning data. There's only one way to do what's been done with respect to fundamental data. And that's the way we did it. And we created proprietary technology to do it. And the fact that our indices outperform the market by as much as they do is proof that what we do, no one else can do. Otherwise, we wouldn't be outperforming. Someone else would have good data. They could measure core earnings as well as we do. Anyway, This is what we were saying 7 years ago, and it's still true. And because it's hard, it's hard going through filings. How many of these did you read today? How many of these did you read in the last few months? We did a couple thousand.

I think the number was 2,900.

Yeah, 2,900 with a level of detail that probably matched by nobody else in the business. Maybe Warren Buffett when he still read filings. And I think they estimate over his life, Warren Buffett did 100,000 filings, right? We've got over 200,000 expertly marked up. The machines testing on themselves, right? That's another several hundred thousand, right? Where we say, hey machine, here's a filing we know is perfectly parsed. All right, now we're going to let you do it. Pretend like you don't know anything. Go. You missed these 3 things. Try again. We've probably done that an infinite number of times. Who knows, Kyle, right? That's training too, right? And we can do that because we've got the expertly marked up filings. So folks, data, it's all about the good data. And I think we have a skill that most people don't have in figuring out how to get good data. All right, Kyle, sorry, I'll stop there.

No, I think that ties in perfectly. The next slide here, back in our ebook, there was a quote that AI startups spend at least half their funding on data cleanup and management. Now the funding levels may have increased since then when we see OpenAI raising Billions. But we do have still some evidence that Anthropic spent millions of dollars simply physically scanning books to build out Claude. That's because they need that data that is not just internet unstructured, unknown. Actually physically scanning those books to help and try and build out some of the Claude functionality. I think it ended up with a lawsuit, which is where that article came from. But still the whole matter of spending millions and millions just to get data. Back then it was cleaning it up. Now it's just to get some better data than what is already available on the internet.

Yeah, that's how desperate they are, right? And that's— they'll do the easy stuff. Obviously get a bunch of books and throw it in there. That's media. And I do think AI is good at media. Why? Because there's like an endless supply of it and it doesn't have to be accurate. It just has to be entertaining. In fact, maybe less accurate is more entertaining. When it comes to sophisticated activities like picking stocks, a medical diagnosis, a prescription, Are you going to trust AI to tell you how to do that? No. You'll trust it to rewrite an email for you because you get to review it first. You'll trust it to make a dinner reservation or give you an itinerary for travel or something like that, or summarize a document. Sure, it's great at that. Yeah, those are tedious activities, and yeah, machines can do that. But can they give you really expert advice? Can they glean data out of filings and organize it in a way that gives you a truer measure of earnings so good that it outperforms the stock market? No, machines aren't there yet. Sorry, it's hard. And until they are, you're stuck with everybody else except new constructs to give you bad information.

And that is best summarized with the next slide here. We talked about how when a computer reads financial statements, how is it supposed to know that equity method investments and investments in non-controlled affiliates are the same? What about retained earnings versus reinvested earnings? These are all different names for the same thing, but the machine wouldn't know immediately that those are the same thing. So that was one of the issues back in our original ebook. And then today, more recently, we still got quotes like, while AI models are good at locating and classifying facts in documents, they struggle dramatically with the precise semantic understanding that tagging and financial analysis requires. We saw the same issue back in 2018, still the same issue in 2025. Most of these large language models struggle with semantic understanding, which is highly important when you're dealing with financials that have nonstandard language, things we wouldn't normally say and talk about in regular conversation. But you've got to actually, again, endow those machines with that expertise to know the difference between retained versus reinvesting or reinvested earnings, or if they're different at all. And so I think, again, mentioned at the top, but sometimes what's old is new. Sometimes what old hasn't changed at all. AI still doesn't perform well if it is fed poor training data. It's something we've been talking about for a long time, something we're still talking about, and it's still true today. And so the question becomes, how do we imbue machines with experience and expertise so they can perform tasks with comparable skill to human experts? Got a nice little chart here. Researchers teaching computer to beat humans. Sounds like pretty awesome at teaching a guy, huh? Sometimes being able to teach the machine can be more important than just having the machine.

Yeah. And Kyle, I think these quotes you pulled together are really great because we spend a lot of time here talking about data. It's not just getting data, right? It's not just like taking the filing and getting all the numbers out of it. It's context. It's meaning. It's nuance. It's like, how did this data point affect Operating profit? How does it affect operating assets? How does it affect liabilities? How does it affect compensation? How does it affect these things? And that's complicated accounting jargon, legal jargon. And that all has to be translated into economics, meaning like core principles of, am I making money, losing money? Am I requiring more capital or less capital? That's complicated stuff. People go to business school for a couple of years to begin to learn that stuff, and then they study it and they look at it for decades before they know it. That's the kind of subject matter expertise that we bring to bear on every single filing. And we've developed institutional expertise around different disclosures and disclosure scenarios for all kinds of different accounting and economic events. 25 years. That's why the data's so good. And that's why the moat is so wide for us because of the expertise that went into collecting the data and the expertise that we developed from looking at so many filings. And we know what right is. We know what good is. We know what signal is, right? That's because it's outperforming. So that's a really big advantage that we have on folks. And guess what? It didn't come easy. It was a lot of work. A lot of late nights pouring through filings, pouring through code, checking models, checking data, creating data checks to check data, creating very sophisticated algorithms to automatically check certain things are right by looking at different parts of the filing to validate a data point, another part of the filing. Everything we could possibly do to get machines to take care of the filing analysis, we did. It started with the really easy stuff, like an income statement, balance sheet, cash flow statement. We were almost fully automatically doing those within a couple of years for most companies. And then that gave us the ability to cover more companies. And the coverage of more companies gave us the ability to understand more disclosure scenarios. And then the software got even better. And then that gave us the ability to get after the notes with more and more speed. Then that gave us the ability to cover even more companies. And you can see as we've Gradually and carefully expanded the scope of the technology. And we did that way so that we knew it was right before we expanded on it, before we scaled. So we made sure that we scaled on only the most reliable data. And that's the key to success.

Yeah, and what David's talking about there is exactly what's on the next one here, taxonomy and ontology, the answer to the question of how do we give the machines that expertise to Taxonomy in the context of an investing AI agent refers to the rigorous organization and classification of financial data. Ontology is the process of mapping the taxonomized data into relationships and patterns that can inform decision-making. That's exactly what David was just talking about there, being able to categorize it and then make meaning out of those categorizations and drive insights and drive alpha in the market. For the investing business, the ideal ontology is one that can produce idiosyncratic alpha. We've got—

And that's a big deal. Idiosyncratic alpha, folks. People throw that term around. It's rare. Let me show you how rare it is. The percentage of money managers that outperform in any given year, maybe 20%. Okay, guess what? That's luck. Because the number that it goes down to in 2 years is like half of that. And then by the time you get to 4 or 5 years, it's getting close to 1%. So, idiosyncratic alpha means that you beat the market for non-lucky reasons, right? It's not explained by some other factor like growth or value or quality, which everyone can participate in, or beta for that matter, which is just the overall market. Idiosyncratic alpha means you got an edge, something special that no one else has. That's what all the studies show that we have. And that's what now we see in live market performance shows that we have. So that's a big deal, folks. Don't take that for granted. Anytime anyone's talking to you about idiosyncratic alpha, make them show you. Make them show you. Say, hey, show us. They say, what do you mean? Show us like New Constructs shows us. Show us how you can pull data out of a footnote in a filing. We've demonstrated 100 times on these demos. are in these webinars. Show us how your products are outperforming your indices. Make them show you. We show you all the time because I want you to get other people to try to show you, because they can't. Maybe there's a few that can. More power to them. I can't say we're the only unicorn in the world. We are unicorn. We're pretty rare. I've never seen another unicorn.

I'll show you right here, the Core Earnings Leaders Index beating the market Year to date by 1,000 basis points, at 27% versus 17% for the S&P. Over the past 5 years, 3,000 basis points, 119% compared to 88% for the S&P. This is data that, as David mentioned, drives that idiosyncratic alpha. We've got our core earnings weighted S&P 500 index beating the market by 2,000 basis points over 5 years. And then I think one, the Very Attractive Stocks Index, which again only allocates to stocks with a very attractive rating, beating the market by 4,100 basis points over 5 years. So you've got short-term, year-to-date, over past 5 years, all of it showing strong outperformance, which further validates what we've got going on with our taxonomy ontology and starting with that superior dataset, not trying to clean data after the fact. And we put it all together, as we've talked about a little bit before, in FinSights, which is our AI stock picking for all investors. FinSights is an AI agent built on and by Google Cloud based on our data to drive real insights. We've talked about it before, we've done some demos. If we've got time, maybe we'll do a little bit of a demo today.

But do a demo. We got a brand new version here, man. You want me to share my screen?

Yeah, go for it. So yeah, you can Ask it all sorts of questions. And here, took a screenshot while David's getting that going. What are the best stocks in the consumer cyclical sector? And it's going to tell you top 10 based on overall rating, gives you an explanation of why they have that overall rating. I'm talking about economic versus reported EPS, return on invested capital, free cash flow. But then we can even go deeper than that and start asking about which stocks are high risk of missing earnings. We've got stocks in a portfolio here. Which one's most likely to miss? It's going to give you that Street earnings distortion score, what the Street distortion per share is, what we're looking at, and whether they're strong miss, beat, strong beat, all sorts of insights from the data that we've got put together. Do you have a demo video? But we'll save that one. I can share that one out and then—

Yeah, we can put that into the chat today. We are in the final stages of rolling this thing out into production. There we go. Let's go. So yeah, I just asked it, what is the ROIC for Apple? I think it really hates it when I misspell. Don't blame it. How about, what's the difference between GAAP and core earnings for PL over the last 5 years, including the TTM? Yeah, you can quickly Whatever ticker you want, obviously. And one of the other things I really like about this is the ability to look at multiple portfolios. So I'm going to put together a list of tickers here that I can copy and paste in because I think it's also fun to evaluate. I'll go ahead and start writing the prompt. Which of the stocks in my portfolio are most likely To miss earnings, and then I'll just put the portfolio in here. We just say KR, NVDA, AVGO, LULU, BDX, AMR, CGS, JPM. We could say those are our portfolio, and then let's see here. Here we go. Boom.

Okay.

All right. And it told us the gap distortion, which is pretty cool. We'll do the next prompt. Which of the stocks in my portfolio— so you can throw in your tickers and get quick analysis. That's another super cool thing. But this is great analysis here too, folks. This is what drives the outperformance of the Core Earnings Leaders Index of the Enhanced 500 S&P 500 Index that we have. A better measure of earnings. Gap earnings, core earnings, distortion. Apple doesn't have a lot of distortion typically. You can see it's ticking up here. So yeah, almost no distortion. It was a little bit more in the past. They were overstating a little bit. But yeah, when you're that good a business— So this is Street distortion, folks. Street distortion is different. Street is how much distortion there is in consensus, right? So GAAP distortion is based on the past because once they report, then it becomes GAAP, right? And then Street distortion is like the number that the Street's coming out with. Of those tickers that I put in the portfolio, you can see these are the ones that one is a miss, three neutral, five strong beat. And the cool thing to do from here, as I mentioned before, you can quickly get this list of tickers, download that. I'm gonna copy and paste this, and I can go into now next I could go into ChatGPT and say, okay, which of these stocks has the best?

Technical.

Oh, why am I looking at AMR? So let's just go to Gemini and ask the same question. The point is you can replicate what Two Sigma, Citadel, and multi-factor portfolio managers— these multi-factor portfolios, those are the like most incredible portfolios in the world, right? These things are generating so much money for these guys. What do they do? Well, they get the best fundamentals. How do we know? Because sometimes maybe they get that data from us.

I don't know.

I'm not naming any names. But they, hey, you start with what's the best underlying fundamentals? What's the best thing position to do? Oh, how good's the technical data? How good is the insider trading data? How good is the sentiment data? Same thing. Which of these stocks has the best sentiment? And we can copy the same list in. Bam. Come in here. Technical rating, right? It's all over the place because It's not as organized as ours, but you get the idea. You come to New Constructs for the proprietary, really difficult to get stuff. There we go. Look, monthly momentum. That's a pretty good answer. But you come to New Constructs only. We have a monopoly. We're the only source of this super good fundamental data. That's where you get your good core list. And then you can go figure out which ones also have good technicals, good sentiment, good insider trading. You can even ask for good option strategies. What's a good option strategy for my bullish opinion on GS? There you go. Buy a call, the bull call spread, the bull put spread, whatever it is, man. So all these other things these other research firms can do, they're great, but it's commodity. We only have— we're the only ones with the great data, and that's what makes FinSights and what we do special. And why we're glad to have you with us today. So we're going to wrap it up here, Kyle.

Yep. I think that's it.

Take a request. Anybody got a request for— you usually got to get these things in early if you want them. Oh, here we go. That was quick. We'll do one. What's the rating on SGHT? See how it does, because we didn't say credit rating. We didn't say stock rating. We didn't say earnings distortion score. rating. So a jerk question. Oh, the overall rating for it, I think that's the stock rating. There we go. Yep, very attractive. You chose a good one. Okay, Kyle, let's close it out.

Yes, I just wanted to thank everyone for being here. We'll get a replay up in our community for this, and we just want to say to everyone, have a wonderful holiday, Happy New Year, and we will be back with you again. Our next month's podcast here will be January 9th. So thanks for being here, and we'll see you guys next time. Thank you.