How FinSights AI Knows Which Stocks Will Beat/Miss Earnings

Welcome in. It is 11 o'clock Central, so I'm gonna go ahead and get this kicked off with a quote that will lead into our live podcast today. Rather than rely on flawed GAAP or Street earnings, we use Finsights, our AI agent built by Google Cloud, to find companies most likely to beat Street earnings. That'll be the topic of conversation both on earnings and Finsights today. So I wanted to kick off with that, but welcome in. This is our live monthly podcast. Myself, Kyle Guskey II. We're here with David Trainer. David, welcome back. And I missed you last month.

Yeah, I'm sorry to miss and I'm glad to be back. We've got some exciting stuff to share today.

Agreed. As the topic of discussion, how Finsight's AI finds stocks most likely to beat and miss earnings. So we'll get into that and much more today. Glad to have you guys here. If you have any questions or anything, you can always put them in the chat. We'll look at those as we go through the podcast today, but start off with the disclosures and disclaimers. Most importantly, we are not providing advice. We are publishing research and you can read through these in the full replay. But again, we are not providing advice. This is a monthly podcast. Our next one will be August 14th. We'll probably talk to you guys before then, but this specific podcast is monthly. If you're signed up for this one, you're signed up for the next one. Congrats. We'll see you then, hopefully. On our latest research, put out a bunch of stuff since we last talked. Some new long ideas, some new danger zones, stuff on FinCEN. This week we put up our earnings preview because we've got earnings season kicking off next week. Talked about quantifying SpaceX valuation before its IPO. There's a replay of that podcast from last month that myself and Garrett, one of our analysts here, did. And then a lot of stock picking outperformance, model portfolio performance for 2Q26. So if you haven't seen all that, it is all on our website right now. The agenda for today, we're going to talk stocks most likely to beat and miss earnings, some upcoming training sessions we've got, and then we're going to jump into Finsights AI. I know you guys have probably heard about it a lot. We've talked about it, we've teased it, but we want to actually jump into the tool today. So we're excited for that. But before we get into that, we'll jump into earnings season, stocks most likely to beat and miss earnings. Longtime readers, listeners, users of New Constructs, Know this, but it's worth reiterating: GAAP earnings, despite popular opinion, don't accurately measure profits. Companies can and do manipulate GAAP earnings through accounting rule loopholes. That is the entire crux of New Concepts. We take those loopholes, we close them, and we calculate true profits, not the GAAP earnings that you see. Unfortunately, Wall Street does not like our approach. They would prefer their better version— non-GAAP whisper numbers or street earnings. But those numbers are also proven to be misleading. You've got times where they miss adjustments, they make the wrong adjustment, can't find certain things. We've seen all sorts of different instances. There's been academic studies that find that a lot of the Wall Street numbers simply don't find the accurate numbers and make the correct adjustments. So rather than rely on those GAAP and Street earnings, we use FinSights, our AI agent built by Google Cloud, to find companies most likely to beat or miss Street earnings. And the results of that work, we published them this week on our website. We've got earnings where Street estimates are too low, who should beat, and where Street estimates are too high and who should miss. Those reports are available to Pro and higher members right now on the website. We're going to do a little bit of some previewing of that later in this podcast here. We're not going to show the actual report from that, but its focus is specifically on the S&P 500 and the companies most likely to beat and miss. We'll look at some smaller cap stuff today and show you how to find the information for yourself using Finsights. But these reports specifically, Show frequency and magnitude of understated and overstated earnings, how Street earnings are flawed, and importantly, the 5 S&P 500 companies most likely to beat and miss. And so I think they're a great way to, again, look at what Wall Street's putting out, what the actual real profitability numbers are, and where the disconnects exist. Because when you can find disconnects in reported numbers, Wall Street numbers, and actual economics, that's where opportunity lives. And so that's what these reports Aim to do, but we do have a few announcements first before we get into Finsights and earnings preview. First up, we are proud of this. We always talk about it because it happened. It's not something that you see every single day. We've had 61 consecutive months our stock picks have ranked number 1 on SumZero. We are extremely proud of this and we like to share it whenever we can. I think David's often said, no one's going to say that your research is good unless you tell people yourself. And in this case, we don't have to point to our own metrics. We can point to third-party metrics. But proud of that. And number 1 in consumer discretionary, 2 in industrials, number 4 in large cap. And these are out of 16,000+ people. So, you know, those are just the top 10. There's some rankings that are outside the top 10 as well, but we only include the top 10 here just to put a nice cap on it, but extremely proud of that. And it's a testament to the work of digging into the footnotes, correcting those loopholes, correcting those, making those adjustments and getting the accurate profitability and valuation of companies. If you want to learn more, we do have some upcoming trainings. And we just got just here, July 16th, we have got a training called the $2 Million Edge. The registration link will come out via email soon. If you got emails about this one, you'll also get emails about that because you're already on our list. And then the Ideas Dinner for pro and institutional members is coming up on July 21st. Those are great sessions. Again, to pro and institutional, talk long ideas. If people have short ideas, it can be anything. We've had some actual new ideas come out of it. New ways to use our research, a whole bunch of stuff from the kind of heavy users of the system. David jumps in and talks with them, talks specific ideas. So it's a good time there and very insightful. And speaking of insights, we'll jump into the main topic today. David, I know we've talked about it. We've previewed it, we've showcased it, and it's finally here. Finsights AI has arrived. It is live on the website. And we're gonna jump in and do a lot of work in it today.

Yeah, we're really excited. The, some of you've already seen it and have been working on it. I think one of the things I want to start with here is a quick explanation of what's going on. And because we, we started building this last year and it was actually built like last December. And in typical big company fashion, it took us as long to get the storefront version of FinSights up as it did to build the whole thing. That's not what we wanted. And it was really interesting to see how the people building products inside of Google were really good at that. They handed it off to the folks pushing it on the front line. They outsourced most of it. Now, they wrote the check for most of that outsourcing, which was good of them. But these were folks who clearly were not in the business of getting things done. They were clearly in the business of making sure it took as long as possible so that it could look as expensive as possible. And what I'm talking about is literally just taking the agent from being in the lab at Google to the Agent Marketplace. And once it's on the Agent Marketplace, we still had to do a bunch of work so that we could actually make it user-friendly and have individual token accounts and things like that required to make sure, like, it could work for individuals. So that is all fully available. Now, before we even jump in to showing how FinCEN works, I want to share an amazing resource we have to support FinCENsights, and it is our community. A lot of you are already members. If you're not, it's free to join. And Kyle, if you want to drop the link to make it for where people can join for free, I want to show that. But you go to our community and you can click here. And you get basically a whole community of clients who've been using it for a while because some of our power users have been signed up. We've had some folks as beta users. Some of you on this call might have been a beta user and been sharing experiences. And there's been some really hardcore questions from some folks about how to use this thing, which I think is cool. And how much, how do you maximize token productivity, how do you minimize cost, all that kind of stuff. And some great ideas on different kinds of prompts or questions you can ask to do screens. So this is, I think, a great resource that you're not really going to see from many other products out there, which is a community here to support you and help you get up to speed quickly to get the most out of what I believe to be the best AI in the world. For what it does, so for that, why don't you guys like throw any questions in here or you want to know about any particular tickers? Why don't you tell me which companies you want to hear about or what kind of questions you might want to ask? You'll see there's some questions I already got loaded in here. For example, the difference between core earnings and net income for the S&P 500 since 2010—that's pretty cool. Who else can give you that? See, I just asked it for materials companies that are likely to miss. Earnings this quarter? Looks like no companies in the material sector. Let's see, what about the industrials sector? So far, no questions from anybody for the Finsights AI? What you might like as part of Finsights, when you first come to the page, we'll give you some questions to ask. You can see all my, all my charts here, or all my prior prompts are saved as well. So if I had one here, I could put this in. This is where it answered the question wrong, and I told it to answer the question right, and it learned. So this is an unusual type of AI in that it is not trained to give you an answer no matter what. It will tell you when it does not know. And in particular for this one, for this question, I told it like, hey, you gave me the answer. You should just format it in this way, and then it got it right. So it was interesting. It said at first it couldn't give me the answer, and then it gave me the answer. And I'm like, you gave me the answer, so don't tell me you didn't give me the answer. And it said, okay. So I thought that was interesting. What about— yeah, here's some stocks in the industrial sector that are likely to miss earnings. How about, can you show me the difference between core earnings and net income for those companies.

I think it might be worth explaining too, when we're talking most likely to beat and miss, we're looking at the difference there you can see between Street earnings and core earnings. And so we're looking at those Street numbers, the ones that miss the adjustments, make wrong adjustments, and our core earnings number and taking the largest distortion capture percentage, making it most likely to miss earnings there.

All right, so we got some questions here from folks. I appreciate that. One point that, that I want to make here is that when we're— when it's giving out lists of companies, it's not going to give them all, whether you're an institutional user or a discovery user, because we can't give out the whole database, folks. Thank you, Mr. Finsights, you got the question right. Sometimes When you have an AI that's not designed to give you an answer, that's not designed to be a sycophant, as most AIs are, you're, you gotta work with it a little bit. He's not as friendly, right? I worry about my dad who like spends a lot of time talking to Grok. I'm like, Dad, do you need to have friends in the real world? And he does. In fact, he's a professional counselor, so he meets with people every day. But he's gotta realize like Grok is designed to give you the answers you want to hear. And you guys, if who, who, who on this, in this meeting have been in other meetings where we've talked about why AI lies so much? And the number one reason, or one of the most consistent reasons across every time we ask this question across every AI tool is, I'm trained to give an answer no matter what. I'm trained to please my audience. FinSights is not trained to please. It's trained to give you answers. Let's see. All right. So let's go through some of these great questions. All right. One of them is GNTX. So let's just ask a question about his ticker. What is New Constructs? Rating on GNTX. There we go. Oh yeah, Charles Thompson. He was in the last call. Yes. That was my favorite line from the webinar about why AI lies. Yeah, it lies, but without the intent to deceive. Like, that just, oh, any of you who've got teenage kids, that's just, it hurts your heart a little bit because it just reminds you of how crazy kids can be in terms of The things they will say. Yeah. All right. Let's go to some other— so there are the ratings. It does that stuff easily. Let's go. What are the top 3 undervalued SaaS companies in your opinion? Let's see. I don't know if it can distinguish SaaS companies or not. I don't know if that's in our system, but we'll try. I don't know. We have a lot of sector definition stuff.

I know we've got the sectors and the subindustries, but industry, subindustry.

So I don't know if it's going to be—

they want a specific sub-industry.

Yeah, we might need to think about maybe a better way to say SaaS would be, let's see.

I think ours would refer to it probably software and IT services.

Yeah, let's see. That might be the way we have to go with it. Software and IT services. Yeah, let's see. All right, there's an answer. Looks like it's okay with SAS. So there you go. There's 3. That's a pretty good screen. And look, one of the real benefits of the way we're offering FinSites right now, because you can get it with Discovery and Insights memberships, this is typically— and so you're basically getting access to what Pro users can do and easier, right? Pro users and institution users can do screens Major access to screening tools, which you saw me just click through. With Finsights at discovery level, you're going to get able to ask questions just like that. How about Finsights bases to determine the moats that exist in the industrial sector, any sector? Well, a great question, but what does that even mean? Like, even as a human, I wouldn't know that. And machines are not good with intuition or gray area. And so I'm not going to pretend that our AI is going to be able to intuit what your answer, what you're really an answer to what your question is, but I'm going to put that into terms that I would use to describe moats. So let's say which industrial sector companies have the largest moats as measured by Market implied gap. In other words, which— and this is where the community can come in helpful, and this is where these meetings come in helpful. How do you ask the questions, the really hard questions? But you got to use our terminology because we had to give the machines an ability to measure moat in a quantifiable mathematical way. Go into like ChatGPT. We may even do this in a second. And tell me what the company with the largest moat is. It doesn't know what that means because it's like a strategic subjective thing. And actually, the largest moat doesn't even matter that much. What matters is which companies has the market given credit for having the largest moat? And which companies has the market given credit for having the smallest moat where I might think the moat should be larger? Because really, at the end of the day, What matters is how my opinion about the moat of a company differs from the market's opinion. That's how you pick the stocks. So understanding what an AI thinks about a moat doesn't matter because we care about what the market thinks and then we care about what we think. The largest market implied gaps. So that's how I'm going to ask this question. And I should have done that little rant after letting that go. Yeah, intent to deceive. That was like the greatest line ever. It was funny, the buddy Greg Lucan who kicked this whole webinar and an article off because he's a friend here in town and he's an investor in our product, our financial product, Core Earnings Leaders. And he just said, he sees, I gotta send, I gotta send this to you. It's unbelievable because he did all this work and then AI made up all these names and email addresses for what he was looking for. And he was like, what? And then he checked. He was like, what? I did all this work and you gave me this fake stuff? Like, why did you do this to me? Why did you lie to me? He said, I'm sorry I lied to you, but not in the way a human does because I didn't have an intent to deceive. And he's just, this is so silly. And so anyway, but yeah, so here we go. There are a lot of companies with a market implied growth appreciation period of more than 100 years. That's what 101 means. That's like the machine version of More than 100 years is 101. And yeah, a higher gap earns— means that the company expects to earn a high return on invested capital for a longer period before competitive forces erode those returns. This suggests a wider and more sustainable economic moat. In this case, all listed companies have a gap of 101, indicating very strong expectations of high, sustained high returns. Everybody clear, like, what we're talking about here? This is the size of the moat that the market has awarded these companies by virtue of their high valuation. And let's see, let's again ask this question. How does New Constructs measure GAAP?

Yeah. So I think what you were getting at really hits at the expectations investing theme we talk about a lot. It's what the market has assigned to the company. And then where you might believe there's disconnects is where you can find opportunities there. Yeah, here it is.

A reverse discounted cash flow model. And so we back into what the market— we solve for GAAP based on estimates on— for revenue and cash flows and WACC. And then we just back into what the GAAP has to be. I think we do this really well. And there's explanations. I think we do better than anybody in the world. And we've got clients who have spent Hundreds of thousands of dollars, millions of dollars and years trying to build models like ours. And they come and they see what we've done. Oh my God, you've done it. And we do it incredibly. And you guys have all seen, we've done a bunch of reverse DCF case studies. Kyle, if you wouldn't mind dropping a link in there to some reverse DCF case studies and—

Yep.

Even if we have a private link to a YouTube unlisted or something, I know we used to have one of those. We could drop this in here for people to see. We used to share a lot of those and we, we took those down off the public websites, internet, because let's face it, we didn't want to give them away to AI.

This will sound like a pitch for our community, but they are in the community. So I'll share that link first.

Yeah, that's right. No, that's the way to go. They are in the community. In fact, wait, why don't I show— gosh, I should remember that. Thank you, Kyle. Reverse DCF case studies. This will show you exactly how the reverse DCF model works. Done it 100 times, and there's some really cool stuff in here. But one of the comments on the community earlier today, or text to me or something. Oh yeah, it was a Q&A. It was like, oh hey, Wall Street Journal today talked about how Netflix remains the industry leader amongst subscription streaming services, but shares are down more than 40% over the last 12 months. Yeah, we've always said it was overpriced. And look, now they're going to have to spend big bucks to get, get into live media. And we predicted that, Kyle, how many years ago? 10 years ago?

Yeah.

We talked about how, like, the one good thing Verizon had going for it was proprietary access to certain live events because everything else has become commoditized. And then, and then, and like Fox and Disney had an advantage in that they had the ability to create new content. I think one way that Netflix has surprised me for sure has been its ability to raise prices, but also to get some decent original content on there. Like, some of these shows aren't bad. Some of them are terrible. Don't get me wrong.

Yeah, I don't know what their actual hit rate is, but the ones that do hit good—

That's right. That's what all they got to do. I think there's a lot of people who got way lower expectations than I do. Like, I was in a meeting yesterday and there's not many shows I can watch. I think a lot of shows are just— there's so much negativity, violence, killing. I just not— I'm not going to subject myself to that. I don't even go to— I don't go to movies hardly anymore. And if I do, I go late because I feel like you just, you're there, you're captive to like these previews that are absolutely, can be horrible. I went to a movie like last, over the holidays with my kids, and I think I maybe told you guys about this. I know I told Kyle, like after the previews, I wanted to cry and leave. Like it was just, I was getting these horror movies are so scary now. Anyway, and you're captive, you're just sitting there. We're literally like— Anyway, but I think a lot of people like that stuff. All right, let's do some more Finsight stuff. Moats, historical performance of stocks that have very attractive credit versus no credit. I'm not sure what you mean by that, DL, but I think that that's a great question to try with Finsights. Let's see, let's maybe narrow it down a little bit because I think if we give Finsights too big a question, sometimes it doesn't work so well. Have the stocks with very attractive credit ratings in the industrials sector performed over the last 5 years compared to stocks with unattractive, attractive, or worse credit Ratings in the industrials sector. So I recommend you guys, if you want to be specific like I'm being here, so that it's just— don't leave a lot to chance because again, REI is not going to try to infer what you mean and give you an answer just so you have a quick answer no matter what. It's going to be giving you real answers. So you got to give it real questions. Let's see here. Reverse DCF, neutral, neutral. Neutral, neutral. What is that deal? List of companies that have fallen in these credit ratings. However, directly comparing the performance of these stocks over the last 5 years, which involves calculating average returns or other complex times. The other thing to keep in mind here is that this sits on top of Gemini. So when the people at Gemini, like, are changing the model, like, our stuff's going to break. It happened all the time when we were building this thing. And the people had no idea. The programmers were like, hey guys, yesterday this, the system asked this question because I gave them all kinds of prompts. And I gave 'em what the right answers were. I gave 'em hundreds of those. Me, Lee, and Cheslin, the team, and Kyle and Garrett too. We were like testing the heck out of it and then it would work one day and not the next. I'm like, what's going on? It would happen so much that I made them create a tool that would just check and make sure certain prompts worked every day before they told me to test it some more. 'Cause we're always rolling out new stuff. I'm like, guys, the more sophisticated stuff isn't working and the unsophisticated stuff isn't working. Okay. All right. How about, can you give me the same list for companies 5 years ago today? It was broken and they didn't, and they were like, oh, and it took 'em like a week. Oh yeah, the guys in the AI lab changed something and it broke this thing and we reported it as an issue. And then another, like in a few days it was fixed. But they had to submit a ticket and say, hey, by the way, what you did broke. And they explained to me that Google had moved toward a model away from a model where everybody was building their own sort of AI tool to like one central AI group. And those guys would just push out changes. They don't ask for permission. They see something better, they push it out, go, and then you deal with it to hell with the consequences. And then people come back and suggest, and so suggest fixes. That's the way it works. That's gonna happen to us. All right, let's see here. All right. Now the list of companies with very attractive credit ratings 5 years ago. What we're hoping is once it has both lists, we can say, can you compare the performance of the stocks in the unattractive list to the stocks in the very attractive list? See if we can do that. All right, please see if it works. Say, please compare the performance of the stocks that got the very attractive credit, credit rating 5 years ago to the performance of The stocks that got the unattractive or worse credit rating. Let's see if that works. I know that I think Garrett has told me he's done it some access from backtesting questions. Oh, all right. I guess it can't do that. It doesn't do backtesting. It's not a backtesting tool. New concerts can't do that either, and I don't think too many off-the-shelf tools could do that. But if they could, you could copy these lists into into another system and ask them to backtest performance. Performance over the last 5 years and get that answer separately. Sorry, we're not a backtesting tool. Maybe I was totally wrong and that's never worked and nobody's told me it has worked and I was off by that. It'd be super powerful if it was. Backtesting was something we— back when we had our hedge fund, like when we started it 15 years ago, it took us like 6 months to build a backtesting system. So that's a pretty sophisticated capability. All right, let me see. I'm gonna do one more question. I can't end on a— on not being able to answer. The top 3 red flags in Adobe's latest earnings. We don't have adjustment level data. Let's see, what are the— okay, let's see, what's the difference in core earnings and reported net income for Adobe's latest earnings, earning latest report? And then we can ask it what creates those differences. So we do not put the individual footnotes, adjustments, and data into FinCEN because we did not put that into Google. That's still something that's an institutional-only type of capability, my friends. We did not put that into the Google Borg. I think you know why.

Yeah, I think to get this question I asked FinSights to try and simplify it. Just what are the largest adjustments in Adobe's TTM earnings? Yeah. And it gave me the adjustment and the 2 main adjustments basically that make up that earnings distortion. So that might be a way to shorten your query there.

Got a way to do it? Is that the question? I'll try it. Kyle, did I get the prompt right there?

Yes. It gave you a different answer.

Why don't you show your screen real fast?

Yep, let me pull that up here. Didn't need to type in all caps, but I guess Insight thought I was yelling at it. But one of the largest adjustments in Adobe's TTM earnings, and then it gave you the same info on core earnings, earnings distortion, and then the 2 largest adjustments being non-operating expenses hidden in operating earnings and non-operating tax adjustment there.

There we go. Yeah, that's the level of detail we're going to get. Anyway, we hope that was helpful. Check out the community. Check out the new Discovery membership. In fact, let me show you what I'm talking about.

Yeah, I put a link in the chat to the FinSights page as well.

Yeah, there we go. Look, so memberships for individual investors, there's a version of Discovery that has FinSights and Insight and Portfolio Insights. So Pro users also get access for free if you're already Pro, and you get a great deal more access. Look at FinCites. It'll show you exactly how many questions and how much you get, but you get a lot and you get to get more tokens at a cheaper rate. But this is how it works. But yes, you get in there and you get access and get things at Discovery that previously would be available only on the pro side. We have to meter it down so you can't get everything because we can't give our database away for that little. But otherwise the hedge funds would be just robbing us. They probably already are with the pro subscription. But yeah, that's how you guys can get this. All right, thanks very much. Enjoy and have a great weekend.