Edition 10: August 31 to September 14, 2026

His argument then was that the gloomy narrative had skipped a few steps in the journey. SaaS was being priced as though AI had already replaced it, when most of us knew the likely path would be messier, just like other major technology (r)evolutions. His hunch was that AI would quickly get absorbed into the tech stack, Systems of Record would keep handling the boring, critical work companies must do, and many incumbents would harness AI's intelligence rather than get run over by it. Of course, many software companies would either ignore the Ghost's premonition or fail to execute.

For much of the spring, his position that enterprise software would domesticate AI looked wrong. As of recently, while you can still find a headline to convince your priors, whatever they may be, the IGV software ETF finished August up about 7% on the year.

This is a long article (estimated 22-minute read), and the conclusion is far from AI being much ado about nothing to software valuations. It's nuanced and open-ended, but to try to summarize a few of the key points:

The closing image: legacy Systems of Record are a handmade scabbard that most incumbents have let rust, with no sword to put in it. AI-natives were born holding the sword but have nowhere to sheathe it. "AI will not destroy enterprise software. It will destroy a great many enterprise software companies that fail to exploit the position they're in."

In the article, his firm also released its AI Resilience Scorecard and a companion Claude Skill publicly, scoring a company on Source of Truth, Workflow Depth, Uniqueness, Pricing, and AI Stance.

For a more alarmist companion on exactly where erosion (or replacement) could bite, OnlyCFO catalogs the 10 biggest threats to SaaS, as told through his seat as a CFO overseeing spend.

Consequential (is it a system of record whose data takes precedence, and what breaks if it is wrong or unavailable?), Coordinating (how many people across how many departments depend on it), Codified (is compliance, security, and audit embedded, is it named in written procedures?), and Compounding (does its value grow over time through data, network effects, or domain insight?).

The headline finding, in the author's words: "At this point in the cycle, the market is paying more for durability than for headline growth. Durability explains 63% of the variance in ARR multiples, vs. 29% for forecast growth."

The median 'higher durability, higher growth' company is down just 3% year to date, whereas the median low durability company is down 27%.

Interestingly, they also conclude that growth and durability appear to be statistically independent of each other.

For a decade, "openness" in property management software meant how many third-party tools you integrated with and how good (and how expensive) your API was. Yardi has historically charged providers upwards of $25,000 for access, RealPage has offered a marketplace many find lacking, and AppFolio has kept fewer than 70 integrations so that they can own their core workflows rather than rent them out.

AI is changing what "openness" may mean. As property operators start doing actual work inside models like Claude, openness becomes less about access and more about agency. In the AppFolio CTO Matt Baird's own words: "The most open platforms are not the ones with the most connectors anymore. They are the ones that give customers the most freedom to shape how this new intelligence layer operates within the context of their business."

AppFolio's approach is a bidirectional MCP connector that puts Claude in conversation with AppFolio's own agent, Realm-X, agent-to-agent. The customer drives it with their own model, and the harness executes inside AppFolio under its rules. The analogy provided in the piece is that it is the difference between handing a contractor the keys and every document versus calling the property manager, describing the outcome, and trusting them to execute it within the rules.

As AppFolio sees it, this allows operators the best of both worlds: the intelligence of a best-in-class generalist model combined with the sector-specific understanding and guardrails of the property management platform. While a generalist model can read a database and reason an answer, harnesses add value by actually understanding the core business.

This technological shift is interesting in the context of AppFolio's position in the market. Much of the company's traditional customer base sits in the smaller and middle-market tier, comprised of multifamily operators who do not employ a CIO to hand-pick point solutions and who cannot bankroll the multimillion-dollar systems integrators that large institutions lean on to stitch a stack together. AppFolio is betting that its technology can help smaller operators keep pace. By drawing on shared training data and capability, a middle-market multifamily manager can effectively rent a data-and-automation moat it could never build alone, which turns the consolidated platform into less of a threat to the small operator than a lifeline against the giants' in-house advantage.

The article ends with an open question that the above point begs. If high-quality reasoning with industry context becomes a commodity, it remains to be seen how managers will continue to differentiate. "That is a serious question, and I do not know the answer yet."

The trait that separates the smartest people he has known is refusing to stop at an answer they do not really understand, no matter how many others accept it. It takes energy, a "compulsive unwillingness to lie to yourself", and being unafraid to look stupid.

As famed scientist Richard Feynman says, the first rule of science: do not fool yourself, and you are the easiest person to fool. As others more commonly say, this is why writing is important. Writing forces clarity, as it's harder to fool yourself when you sit down to write about it and it comes out all disjointed and confused.

Unfortunately, many people don't actually try to understand topics in a deeper way. Many schools almost train us not to question too deeply, else you fall behind. Just learn the algorithm, plug in the numbers, and pass your exams. Speed is of the essence. In this way, school kills the "will to understanding" in people.

His countervailing advice to people trying to understand something is: go slow. Read slowly, think slowly, really spend time pondering the thing.

For example, Bill Gates structures his famous "reading weeks" around an outline of important questions he's thought about and broken down into pieces. e.g. he'll think about "water scarcity" and then break it down into questions like "how much water is there in the world?", "where does existing drinking water come from?", "how do you turn ocean water into drinking water", etc., and only then will he pick reading to address those questions. This method is far more effective than just reading random things and letting them pass through you.

Relatedly, there are some mantra-like questions that can be helpful to ask as you're thinking through things. Some examples: But what exactly is X? What is it? Why must X be true? Why does this have to be the case? What is the single, fundamental reason? Do I really believe that this is true, deep down? Would I bet a large amount of money on it with a friend?

Personally, I am concerned that with AI, many people will skip to the answer rather than leverage AI to work through such questions more easily than ever before.

Investment hygiene. With no third-party capital to answer to, Vlad Barbalat argues Liberty can hold to what he calls "investment hygiene": doing the right thing rather than the expedient one. The caveat he adds is that permanence cuts both ways. It can breed complacency, and "we can make long-term decisions others can't" too easily becomes an excuse for results that simply are not working. Which brings him to...

One year vs. three to five. Long-term decisions still need to be built from sensible short-term results. Every business is stuck with an annual calendar, and one-year results matter to stakeholders, even though we all recognize anything can happen in a single year. His answer is to hold both truths at once: stay honest about the calendar year, but put yourself and more of the organization more explicitly on the hook for three-to-five-year targets.

To get what you want, deserve what you want.

Attain fluency on the big multidisciplinary ideas of the world and use them regularly.

Learn to think through problems backwards as well as forward. Invert, always invert.

Be reliable. Unreliability can cancel out the other virtues.

Avoid intense ideologies. Always consider, and know, the other side as carefully as your own.

Avoid being part of a system with perverse incentives.

Learn to maintain your objectivity, especially when it’s hardest.

You’ll be most successful where you’re most intensely interested.

The highest form which civilization can reach is a seamless web of deserved trust.