H1 2026: Loss of my mother, AI investing, AI amplifying what investors get wrong about tech, and an AI report on Kaspi (KSPI)

My mom passed away. As the son of divorced parents, with my mother as my primary caretaker, I can’t describe the sheer intensity of the pain. Being a new father, caring for a baby away from home (we live in Hong Kong, my mom lived in Toronto), and carrying the weight of her passing has been incredibly challenging. My wife has been an incredible partner and mother in the process. The primary lesson I take from her passing is to live your dreams. It’s so easy to forget to live your dreams. It's so easy to sleepwalk through life. As Buffett once said, don’t save sex for old age!

As a first order of business, you’ll note that I don’t have our portfolio published. I don't intend to start a fund. Instead, I plan to transition and delegate more of our capital management to AI (meaning our holdings may look very different a year from now). Furthermore, the portfolio contains some fairly illiquid names and will have more going forward. As a result, it seems like there is not really much upside to sharing the portfolio and some downside as I don’t necessarily want competition for the most illiquid companies I own. Going forward, this will serve as something of a diary for myself/my kids when they are old enough, as well as a springboard for conversation with those who read this. Finally, I have been writing software for ~24 years and it’s what I’ve spent my career on. It’s the industry I know best. As a result of AI fears, the sector has sold off because of a market rotation and/or because of AI fears. As of this year, for the first time, the correlation between semiconductors and software has broken down, appearing to be a zero-sum game for liquidity. Our portfolio is down around -9% to date. However, what it has given us is some of the best investment opportunities that I’ve seen since starting investing (at least for someone with my background). As a result, we are nearly 80% invested in tech/software as of the time of writing.

Rolling Correlation between SMH (Semis index)/IGV (Software index) [PortfolioVisualizer.com]

AI Investing

It’s of note that my family’s capital is in a transitional period. I am still managing it myself, by hand, the old fashioned way. But I do intend on transitioning the portfolio to being fractionally AI managed and perhaps even fully automated. My initial vision of a fully automated system is now essentially complete.

The irony about having gone through this AI development experience is that it has given me a stronger foundation in fundamentals value investing despite having the goal of eventually having AI become the practitioner. The base of my investing system is quantitative and deterministic, elements that aren’t debatable. I ran all manner of backtests looking for signals in the basics. As an example of something I was wrong about, I thought metrics like P/FCF or P/E were superior indicators of value and thus future performance. As it turns out EV/EBITDA and P/S outperform these metrics by a landslide in backtests. Intuitively, it makes sense since margins move and free cash flow/earnings are full of one time accounting items. P/S and EV/EBITDA tends to be more indicative of normalized value. The experience taught me the value of attempting to deduce a company’s normalized owner’s earnings, an exercise I previously did not spend a great deal of time on. There are numerous other findings that surprised me. Momentum investing works. In fact, in some backtests, just using single momentum factors, the backtest returned results higher than 30% CAGR. In addition, when analyzing the historical performance of momentum targets, I found that even though many were temporarily mispriced by overoptimistic markets, a surprising number actually grew into their valuations.

To me, AI represents the ability to traverse the vast space of investing at speeds previously impossible for the “small” investor. I think there are many ways a pure AI system can and will beat a human. That said, do I think all forms of investing are suitable for AI? No. At least, not today and not for a portfolio size under $50M USD. The galaxy brain, variant perception, high IQ, contrarian style of value investing, requires a large budget. AI can follow many/most frameworks you give it. If you ask it to validate a brain dead easy investment thesis, it’s something it can do relatively easily. If you ask it to, as a human would, evaluate a highly contentious, difficult, and uncertain investment thesis multiple times (ie. the exact same prompts),  depending on the prompt/settings you might get a different answer every time (albeit confident sounding answers). In many cases, if the investment is hard for a human, it is also hard for AI. Currently, my answer is to put the most difficult investment cases into the “too hard” pile, just like a human would (for a given budget). This current wave of technology is still young, I expect things to change over time. That said, at the time of writing, there is a fair amount of low hanging fruit that exists. Markets aren’t efficient across all geographies and thus, I think fully automated investing can provide superior investing returns.

It is of note that currently, to invest at a level I’d feel comfortable putting a fully automated system behind, would cost a fair amount. Likely $50K+ for the hardware and maybe $5K/year of opex. To be clear, this is much less than the salary of an analyst but the system would likely outperform a team of analysts. However, the cost of tokens is trending towards zero and we are still seeing exponential improvements in hardware. Likely I am going to wait a year or two before fully pulling the trigger. As an example, I am including a link to an AI generated report I generated so you, the reader, can make your own judgments as to how good AI can/will be.

Personally, I’m of the view that once more economical, the analysis part of investing will tend to go away. Personally, I find this sad as this is the work I personally enjoy. Algorithms (without AI) are arguably already better than humans at portfolio management, risk management, and so on. What remains in wealth management is the art, the emotional work (client management), and so on. Whether this remains an industry that continues to employ large swaths of people, I am unsure. Certainly, the writing on the wall has been visible for some time with respect to passive vs. active management. I suspect those who offer index beating returns in the long run, however, will continue to be just fine.

AI amplifying what investors get wrong about tech

I’m going to say something provocative in the hopes of adding a step to readers' processes (not because I want to ruffle any feathers necessarily). I don’t read most professional investors writeups on tech, especially generalists. It could be a fund manager running billions of dollars or a smaller investor, it doesn’t tend to matter. They get a lot wrong. They don’t understand the technology, the evolution of the technology, how software is made, the history of the companies, etc. Of all the mistakes tech investors make, the worst is focusing solely on the current state of the IP, moats, unit economics, and financials. The things that you’d look for in every other industry. To be clear, I think these things are important, especially in the short term. However, to solely focus on those factors is akin to building a house whose foundation is in the sand. Why? Tech changes fast. 15 years ago, Blackberry/Nokia were relevant, Uber expanded outside of SF for the first time, Instagram was barely a year old, mobile compute capability was ~1/2600 (CPU+GPU+NPU combined) that it is today. Imagine trying to put a DCF on anything for more than 5 years going forward in an era where we have AI. One of Buffett’s virtues is “I try to invest in businesses that are so wonderful that an idiot can run them. Because sooner or later, one will.” In the case of tech, the vast majority of businesses do not fit that description. They require competent management and more importantly, competent orgs to navigate ever shifting waters.

In my opinion, time and understanding should be spent not on the state of where things are today, but the first derivative with respect to time. In tech, unforeseen evolution, disruption, and so on are the rule not the exception. Engineering, be it software, hardware or so on, is an intense, full contact, team sport. Always has been. It is the teams, the people, product iteration process, collaboration systems, customer discovery, org structure, etc. that determine where the puck is going. The higher quality the team, the more likely the company is going to fare amid a constantly changing world. In investing, time is typically spent researching the management teams, incentive structures of the management teams, and so on. I have seldom seen this research extend to the same degree as the staff. In fact, most investors complain that stock-based compensation is high and tend to have the view that engineers are mostly a commodity. As if companies voluntarily pay engineers out of the goodness of their hearts. Nothing could be further from the truth.

If the rate of change of tech was f(t), the rate of change in tech with AI likely is f(t)^n. It takes a stellar engineer, product manager, etc. and amplifies their productivity by, potentially an order of magnitude or larger. What we are likely going to see, is an increase in the rate at which tech changes. Investors need to skate to where the puck is going to be, and that puck is going to be moving faster, much faster. Nothing seems to exemplify this point more than AI itself. Every other week, we see a new remarkable model released. Every other week, we see markets move dramatically as a result. This week it is Google, the next it is Anthropic, Deepseek the next. Within the industry, the value of star talent is understood. It’s why stellar tech workers tend to get million dollar (or more) salaries and why Meta (and the broader industry) is paying mega bucks for star talent. It is not inconceivable that with AI, the 100x engineer could potentially be the 5,000x engineer. In short, it's my belief that it is the quality of human agency and the systems that galvanize effort, that determine longer term success in tech.

So if there is anything I would say to add to a process amid a world that is undergoing yet another tectonic change is to focus on understanding what the culture/incentives of the company is and who is on the boat at ALL levels, not just the c-suite.

Kaspi

An AI generated report can be found here. It is a good starting point. Despite the length, I had to do a fair amount of supplementary research to cover my earlier point about figuring out people and processes. Kaspi has a culture that seems very aligned with how much of Silicon Valley thinks about building consumer tech businesses, the management team is very much aligned, and unlike many listed tech companies, SBC does not represent a large percentage of net income. But most importantly, for me, when I view the business 5 years into the future, there likely is a business that can be seen with some resolution. For the things that aren’t, it seems there is a highly adaptable, aligned team that executes at a world class level. That said, as we are talking about Kazakhstan and emerging markets, care is required.

Concluding Thoughts

Time is short, life moves very fast. So if there is anything I would say to add to a process amid a world that is undergoing yet another tectonic change, it is this: undervalue the static, and adequately value the dynamic. It is insufficient to model the world as it exists today because tomorrow, it may not exist.