Click on the image a PDF version of the letter.

We are awash in high frequency yack and CSC has determined NOT to comment on the wiggles and dashes d’jure no matter how omnipotent they may seem. Or completely moronic. (https://www.citrini.com/) So yes, a good chunk of the ensuing 3200 words (TGNTR) that will follow this sentence will reference the nonsense regarding AI pushing off a cliff all that has come before it, but we waited a few weeks.

Since no one really knows anything about the future, no matter how apparently dystopian, I don’t have to burden myself with any unease regarding my relative global standing in technology industry cognition. You have permission to throw rocks at that statement. But I would say 40 plus years of participating in and observing the behavior of humanity in financial and social markets, plus a lot of reading of history, leads me to certain mounds of conceptual wisdom. For example, how friggin dumb can “we” possibly be? How can a Tuesday show a miserable two-year trailing performance of anything consumer staplely due to the GLP-1 phenomena, and yet by a Thursday there is a mass rush to buy the maker of Oreos because apparently, they can’t easily be twisted or dunked into milk by an online agent? And Wednesday its “over” for the global payments system of Visa and Mastercard? And Thursday, a roughly $12 billion mark suffered by the holders of “transportation and logistics” stocks as a $6mm market-cap company put out a press-release trumpeting its pivot from karaoke machines to a new “something” that supposedly boosted a customer’s freight volume more than 300% without a corresponding increase in operational headcount. Do the words Pets.Com bring anyone to January 2000? There were 14 AI Superbowl commercials in 2026, which “rhymes” with what happened to the ad spenders from 2000, but there simply is not a “good enough” paper for me to confidently nod my head to with a smirk. And yes, there were actually Nobel prizes awarded to those who “proved” efficient markets.

What most of 2026 has shown is how little money on the margin is “investing” and how much is “whatever” leveraged in large pods of capital with apparently tight stop-losses where selling on the way down is the natural corollary to buying on the way up. And how quickly the corporate and investment worlds can develop a speech impediment where the phrase AI is inserted every 8 words like a bizarre Tourette’s manifestation. You don’t win on style points, despite the amount of assets still anchored to it, but mindlessly ping-ponging from one shiny object to another is bound to create some stupidity in markets and not go well for third party observers who think they can ping and pong themselves. Our generation’s least read book remains The Art of Not Doing.

This is relatively simple. In my best Monty Python voice, the software “business” is not dead, it’s just many of the stocks are. Yes, there are going to be “trades” in the right here and now. Even Buffett never lost his late-night degenerative glee at successful speculation. This isn’t figure skating; either we are on the podium with numbers, or we are skulking in the back with “well, I nailed it on Morningstar style points” which is no way to live life outside of a pension investment committee with three consultants watching you

But there has been nearly a decade of high growth with public and private buyers paying ever higher and higher multiples to compound and/or lever, and voila, something changed as it always seems to do on the way to the Unanimous Perfection Bank. Said another way by one Daniel Zwirn, CEO of credit firm Arena Investors: “Investors’ guard was at the lowest, the price competition was the fiercest and the rate of growth of the biggest firms at the highest. It was the perfect set up for the worst combination of risk and reward.” Now we all know he was referring to a few of the notable blow-ups in credit this year, but to which other asset classes can that be applied? Answers might include quantum computing vapor, “quality compounders,” SaaS investments both public and private, and, oh ….maybe the trillion dollar AI capex budget in 2026 of the ten largest companies in the world?

In the world of technology, and smaller cap therein, you can correctly imagine we see a lot of absolute speculative nonsense, that to be fair, looks exactly the same as huge successes in this space often do at the same point in their success curve. And until very recently, slower growth, “rule of only 30” companies with the same margins and free cashflow, just less pizzazz on the G variable, tended to be less than appreciated. That has changed a bit, and we have added there.  Yes, a pair have kitchen-knifed their way through a level we thought “almost silly,” but we have leaned in recent years to a “sizing is more important than entry level theory” that also has its limits of applicability. And have also leaned into a new name getting plastered on the freeway of “Oh, no, AI” and we have made an existing small position our second largest holding under the same premise. It all adds to roughly 12.5% of a portfolio.

While there is the immeasurable joy of being our client and seeing ALL the exact names, I know I hate to read soliloquy without an exact “buy or sell this conclusion.” So here is Clarivate (CLVT), which we own and insert all usual disclosures. And if you are asking, SPV world, the best way to work with us is to GIVE us the money to run and let us make the decisions on work we have compounded for years, rather than asking you to make decisions on $50mm in a week in the midst of all the other things you are doing.

1. High margin, high return, high FCF.
2. New and competent management took over from SPAC clowns.
3. Interesting ownership – PE, Exor, NY activist all of whom have been sandblasted over past 2 years with costs north of 6.
4. Will shortly announce the sale of 1/3 of business at double digit multiple – CLVT now 6.5 as a whole.  Simplify, take debt down a full turn enabling monster equity repurchase
5. Crushed by the AI trade.  I think “this” is a representative clip that is not just applicable to CLVT – but many software moats at large. Their IP product is a “better” – IQVIA is #2. And there 2 other business lines fit the same “closed moat” concept.

“I can tell you from our conversation with IQVIA, that IQVIA believed that defensibility and audit trail associated with the decision making process and its dashboard was not necessary. What we have learned is it’s absolutely necessary. Don’t forget, no matter who we are, where we work, we always have to answer to someone, whether it’s the board of directors, my CEO, potential health authorities, the VC group, etc. IQVIA never thought that defensibility is that critical for them.”

Zero growth is a 4 handle. I think the risk is “just no growth” and the market always questions the valuation of flat topline, “buy the crap out of your stock.” The AI move renders this somewhat moot in my opinion.

Back to the Letter.

But we hear an old phrase from the old days, and thanks to AI, I can prove it wasn’t part of Merrill Lynch legend Bob Farrell’s Top Ten Rules of Investing, but I still think it might have been #11: “Never underestimate how long a formerly favored group will stay out of favor in the next phase of life.” Or something. Translation: having watched XXXX et al go up 2000%, buying the first down 20% tick to feel like you are now part of the old club can be problematic in practice. It took 7 years to justify buying Cisco after February 2000.

So one wonders, is there legitimately a longer lasting “change” in stock market sentiment toward large vs small, public vs private, value vs growth, index vs active etc? We have hopes and dreams and private jet brochures that might reflect our greedy desires, but to repeat the above, we have no idea. Spending endless hours worrying about what others are thinking and doing with money is 94% wasted time vs just doing the work: looking at values, looking at business models, meeting C-level people and understanding their motivations and competence. Having a North Star in this world (and 40 years of mistakes) helps to provide context as to what if anything is really changing and what one should really do about it. If anything at all. And as Henry Singleton once said when asked about his strategy, “My only plan is to keep coming to work every day.”

Another thought of the year-to-date highlights some of the issues with “well, I just index to the S&P 500 at 5bps and call it a day.” As anyone who has remotely been paying attention, no, you have just increasingly bought yourself exposure to a cap-weighted bunch of technology. Just as any good idea in finance tends to get exploded into something that resembles something much dumber and riskier and more expensive than originally intended, indexing can be argued to be belatedly somewhat at that point. Commercial break: curated small-cap investments.

Back to the software mess of 2026. It seems very reasonable that growth on the margin will be picked off by changes from AI, terminal values in excel sheets will need to come down and the McNealy argument of the even more “reductio ad absurdum” of the 20x revenue multiple breaks back down into more breathable space.(Which is NOT good for everybody’s favorite until a month ago Private Credit bet, more on that below.)

Maybe the market is telling us that the multi trillion-dollar AI landgrab could have an uncertain outcome? Be wasteful? Whack at the return math on previously pristine asset-light cashflow machines? Uncertainty as to who benefits from this spend other than the current crop of suppliers — the spender or the user? “Obvious prospects for physical growth in a business do not translate into obvious profits for investors” so said one Benjamin Graham,

I will step out and say, “no AI is not going to be as good as HVAC” to mankind at large. It is an outstanding white-collar tool. It will assist good companies operating under a continuous improvement mindsight to get better. It will not ruin commercial insurance brokerage by enabling someone to do their own work with AI as they spend millions of dollars on risk management to avoid hundreds of millions of dollars of legal risk. It will take decades to truly “change” government and municipal services, healthcare, banking and the movement of large sums of money, or the military. The disastrous history of mergers and acquisitions will not be fixed by faster and prettier spreadsheets and PowerPoints, nor will it improve the accuracy of future projections created out of thin air by the whims of managing directors. (The hallucinations of AI are in fact man-made.)

On how humanity screws up good ideas, we read this from KPMG, one of the consultancy geniuses piloting allegedly hapless clients through the dystopian AI future or possibly using themselves as a prime victim. “The unnamed partner was forced to redo the AI test after uploading training materials into an AI platform to help answer questions on the use of the fast-evolving technology. More than two dozen staff have been caught over this financial year using AI tools for internal exams, according to KPMG.”

I read other things like this: “We have created a Value Acceleration Office. Here, we are using 80-20, smart AI and process engineering to streamline our portfolio, create powerful and simple internal systems, and manage costs more effectively. We expect strong contributions to our margins in the coming years based on this promising start in 2025.” Good companies continuously improve with the tools at hand. Tools change.

Here are a group of other thoughts that seem to make sense to us in thinking this world through, with the conclusion noted above: the business model of Saas is NOT going away, but the ridiculous premiums paid for the honor of investing in them will:

• “Converting to SAP” used to be a code word for “short the stock.” Anyone who has run a real business with tens of thousands of people understands committee decision making, switching costs, training, customization, the business disruption to enable massive systemic changes, cybersecurity risks and potential customer disruption.
• There are “costs” to AI conversion like there was huge fat costs to “let’s move to the cloud.” The people spending $4 trillion will likely want something more than the pure delight of enabling you to make videos of your dog dancing with Charlize Theron. Having an “agent” replace those annoying young people often seems like a good idea, but then there are issues of supervision, validation, integration, security hardening, model monitoring, failure management, and cyber security, as the very bad dudes in the world are equally in love with this technology as well.
• Let’s say 50% of US GDP is banking/payments/money movement, healthcare and government/defense/municipality. Now think hard about how quickly these will be vibe-coded away in practice. Cost and efficiency are only part of the equation in massive conversion to a new AI infrastructure and the ability of large organizations to restructure around it.
• From a call we had this week: “We can tell you from our conversation with XXX, that XXX believed that defensibility and audit trail associated with the decision making process and its dashboard was not necessary. What we have learned is that it’s absolutely necessary. Don’t forget, no matter who we are, where we work, we always have to answer to someone, whether it’s the board of directors, my CEO, potential health authorities, the VC group, etc. XXX never thought that defensibility was that critical for them.”
• I have the data and you don’t seem like a very viable moat. Ask any investment manager what the inflation rate has been for MSCI and Russell data access and is often mandated by institutional clients.
• Many of those that seem most threatened, at least by stock market reaction, are often the ones using AI the most. And as Microsoft’s market cap proves every day — the best product is not always the winner.
• “Yes, it’s — they’re using some AI models. But if you really peel a few layers back, all it is, is a really sophisticated algorithm, which is really just a decision-free on steroids and they layer a chatbot model on top of it. And the reaction of the world as a couple of my colleagues and I were talking about earlier, was like these people just discovered fire. It was just bizarre to me.”

We close this section of random verbiage with words from Mark Leonard, the outgoing founder of Constellation Software, whose stock really highlights the year-to-date kerfuffle. It was stupidly priced and lauded as the “compounders compounder.” Things have changed in the short run:

“…  software itself is not a monolith. Software, which it’s a system of record where you are integrated into the business processes of large companies. Our business processes are a big part of how companies operate; software is an enabler. And the best companies that we are seeing that are embedded in that position and have data moats and operating environments like health care and financial services with 0 tolerance for risk environments, regulatory limitations, what we’re seeing is they’re the ones that are the adopters of AI. They’re the ones that are then turning around and saying, “Here, I can offer you an agentic solution to replace some of your human costs, some of your labor costs by integrating these capabilities into the software, I am already resident in your system and fully integrated into your daily behavior.” 

So things change and securities reprice. Don’t be terribly wrong and terribly overcommitted on being right about the future; don’t overpay; be careful of the accolades of peers as investing successfully is often a lot lonelier than most understand; be wary of leverage; and really try to understand “baseline” — the world rarely changes as quickly as you think, and organized groups of people move more slowly than screens. Fiendishly simple.

All this leads to the canaries of credit and credit availability. Let’s all agree that the trailing performance of many asset classes and the present valuation of said asset classes has been enabled by massive fundraising and subsequent fund spending on said assets. We have seen that math that suggests “over 20% of private credit funds are loans to software.” In the current setup where institutional LP’s seem to be whining loudly they want their money back, and the GP’s aren’t liking the bids on what they might need to sell to answer their LP’s, who owns what paper and will have the largest problems? And on the margin, “we” only see what is publicly traded and what we see is impossibly messy. Private Credit Funds often own other funds from the same sponsor, so when say Blue Owl sells assets near par from “I need liquidity” Fund A to its own insurance company and a handful of other funds that may or may not have less than visible interest in a positive outcome, it’s not really clear what is being said about “marks.” To bring back thoughts from either an old S+L crisis or the 2008 fun, the pricing of illiquid assets in a market with no bids is an interesting exercise. All of which promises to be as riveting to watch as a 40-car pile-up on a Lake Tahoe mountain pass. Good luck.

The concern is becoming collateral damage. Less liquidity is simply not a good thing for financial markets at large and waltzing around with a proud chest saying “well I don’t own that crap” didn’t help much in 2008 or 2020. Not owning nonsense did help in 2000. The future remains uncertain.

Adding to this was the important change of the guard at the Federal Reserve, which was important for about a week until technology stocks got smacked and it disappeared in yesterday’s algorithm. But this “could” be a lot more important and material than whining about sector rotation changes.

Kevin Warsh has very smart and connected friends and is clearly a political animal as everything on the record in his past suggests he is the opposite of “Yes, Mr. President, I will lower rates, damn the torpedoes.”

In an airless vacuum of the financial world, the real job of the Fed is maintaining a steady hand on inflation, and not to play Polymarkets and try to time markets. It should not be trying to influence the pricing of X asset class over Y and it should provide some oversight over a properly functioning banking system. It is not for social experiments.

Warsh has tended to agree, and the obvious applicability of the previous paragraph is to shrink the balance sheet of the Fed and enable the market-pricing of risks and liquidity to have at least a dead-cat bounce. But as Jay Powell found out and backpedaled faster than Carlos Alcoraz chasing down a lob, changing policies like “we are going to shrink our balance sheet by 50% and let markets figure themselves out” is a very fat tail process. We have spent 15 years enabling large pools of capital on leverage to enjoy the hell out of themselves with a systemic put underneath most of their efforts. Changing the game threatens livelihoods and increases risk at large.

Contrary to our celebrity culture in all fields, there is a Federal Open Market Committee, which requires cajoling rather than just “because I said so.” Here is your Polymarket bet: which is worse: a Jay Powell “follow the data” Fed which has had as miserable a record of prognostication as any institution can offer, or a “conviction-led” policy focus?

So we close. Yup, we are having a nice start to the calendar year, some of which relates to actual work we have done and some of which is the inexplicability of timing. And of all things, some weather-related. As of this writing, we haven’t blown it by trying to be a hero in the aforementioned software space. But we like messiness and you should too: it creates investment opportunities.

Our “journey” here at CSC is related to this story, which I nicked from one Dan Shipper, the cofounder and CEO of Every. (X at @danshipper.) They do some interesting things about practical uses of AI, of which there are legitimately many to help one more quickly get to higher and better uses of time.

The story goes like this: At a company retreat in 2002, when Amazon managers wanted more communication, Jeff Bezos fired back that “communication is terrible!” A few weeks later, he restructured the company around small autonomous teams. If a team had more than 10 people — more than could be fed by two pizzas — it was too big. Twenty-four years later, two-pizza teams are now themselves too big for building software products. When each employee is armed with Opus 4.6 and Codex 5.3, the ideal team size shrinks even further. I call it the two-slice team. Two slices, to feed one person. (These are New York slices that you fold in half and eat standing at a counter.)

In investing and life, spending your time on critical variables and what is really important directionally pays dividends. AI can legitimately reduce “task time” in the investment business, but it does not replace experience and judgement. Not having double-digit bodies running around also helps one focus.

Operators are standing by.

Jeffrey Bronchick, CFA
Principal, Portfolio Manager
Cove Street Capital, LLC

*The opinions expressed herein are those of Cove Street Capital, LLC (CSC) and are subject to change without notice. Past performance is not a guarantee or indicator of future results. Consider the investment objectives, risks and expenses before investing.

You should not consider the information in this letter as a recommendation to buy or sell any particular security and should not be considered as investment advice of any kind. You should not assume that any of the securities discussed in this report are or will be profitable, or that recommendations we make in the future will be profitable or equal the performance of the securities listed in this newsletter. Recommendations made for the past year are available upon request. These securities may not be in an account’s portfolio by the time this report is received, or may have been repurchased for an account’s portfolio. These securities do not represent an entire account’s portfolio and may represent only a small percentage of the account’s portfolio. Partners, employees or their family members may have a position in securities mentioned herein.

CSC was established in 2011 and is registered under the Investment Advisors Act of 1940. Additional information about CSC can be found in our Form ADV Part 2a,