Factor Capital Update - March 2026

When you look at the market right now, the one thing that is certain is massive volatility wrapped in uncertainty. People are pretty confident that AI will disrupt and destroy some businesses and labor someday. They are just split on the realistic severity. Fear pieces are proliferating and rocking markets as new industries are written down each day based on the latest advancements in Claude. And underneath all of it, something structural is breaking that most people cannot see because they are looking in the wrong place.

I certainly do not have all of this figured out. Nobody has a perfect map for what comes next. But this environment feels familiar to me.

Cues From History

In early 2008, I was an equities trader at Merrill Lynch. It was the last few months before I left. The reinsurers were starting to implode. Everyone noticed this, but it did not feel like a critical part of everyday life. It was not treated as a clear emergency. Then Bear Stearns collapsed. I will never forget my boss standing in front of me as the stock hit $35, telling me to start buying BSC because there was no way they were insolvent. … $30 “buy more”... $25 “buy more”… $20 “cut it”, and we burned through a million dollars in about 30 minutes.

Over the weekend, JP Morgan absorbed them for a couple of dollars a share. That was supposed to be the big shoe dropping. Fast forward six months, and Bear was followed by Lehman and Merrill. The Great Financial Crisis was upon us.

This moment has that same feel. Not the exact same mechanics, but the exact same pattern. A growing emergency diluted by surface-level calm and a decent amount of denial.

The Indices Are Not Telling The Story

One of the main reasons this feels less severe than it is comes down to index math. The S&P and Nasdaq are near all-time highs. But they are overwhelmingly driven by the companies causing the disruption. Microsoft, Nvidia, Google, Meta. They are the upward leg of what is becoming a K-shaped market.

Most of white-collar society risks being caught in the lower leg of that K. The market is loudly recognizing that middlemen and brokers are being replaced. The trap is that none of us thinks we are the middleman. The consultants, the mid-level managers, and the knowledge workers all assume our specific judgment is irreplaceable. We think the disruption is overblown and the transition will be slow. We watch output get repriced by the week, but assume it only applies to our competitors. The individual pain does not show up in the Nasdaq composite. But it is real, and it is accelerating.

This triggers a brutal chain reaction. When white-collar jobs vanish, enterprise spending behaviors violently reset. Look at the panic around SaaS companies right now. The SaaS repricing has less to do with AI completely replacing software and more to do with a collapsing business model. Legacy SaaS is dependent on selling seats. When headcount contracts, seats shrink. If an intelligence tool allows a company to cut 40% of its workforce, 40% of those SaaS licenses evaporate instantly. The market is ruthlessly repricing the shift from seat-based models to outcome-based models.

Not all software will suffer. Headless systems of record are going to be incredibly resilient. As companies begin building their own internal apps and custom tools using AI, they increasingly need to invest in those core databases to anchor their new infrastructure. The application layer largely gets commoditized. The data layer becomes a utility.

Not all businesses will suffer. For some, they find that they have the right recipe of culture and leadership and talent to harness these capabilities to operate at multiples of their current output with their existing workforce.

Smart capital is already placing bets. Firms like Thrive Holdings and General Catalyst’s HatCo are actively rolling up traditional, human-heavy service categories. They buy the distribution and the customer relationships, then replace the human margin with AI leverage. They are seeing to be first movers, arbitraging the transition while most of the market strategizes and waits to see what happens.

This brings us to the private credit markets. Private credit is what investors have decided canary in the coal mine for this entire cycle. It’s the reinsurer market of 2008 as they underwrote those per-seat SaaS businesses at top-of-market, post-COVID multiples. Furthermore, the white-collar workers losing their jobs are the same people who hold the mortgages and consumer loans backing the broader credit system. As headcount contracts and enterprise software reprices, the financial system's underlying collateral degrades. For anything outside of the essential physical economy, the question is not if the pain arrives. It is when.

Where Transformation Occurs

You absolutely cannot observe this transformation if you are just using chat interfaces. For the average user, chat is still a nice-to-have efficiency driver. It creates a false sense of security because it makes the shift feel incremental, because 1. the model improvements are only every couple of months, and 2. the components the base model is mostly improving upon are not the ones that are apparent in standard ChatGPT. What’s actually happening daily is not incremental.

When you build, you actually see the future. You see just how many tasks you can offload and take on. How the agents work for hours to produce research or software. You enable the expansion of the base models through the tools and skills you architect on top of them. If you are accustomed to thinking in terms of the inevitability of technological progress, you can forecast pretty clearly where this is headed.

Technological progress comes in phases. You make advances, hit bottlenecks, and then your energy goes toward clearing them. Look at medicine. As humans, we want to stay alive. The primary bottleneck to a longer life is disease. So society points a massive amount of capital and human ingenuity directly at that bottleneck to develop cures. Out life expectancy extends

We see the exact same feedback loop happening in AI right now, but radically compressed. When AI labs realized power was the bottleneck to scaling their data centers, slower growth wasn’t an option. They started contracting nuclear power plants to solve the constraint.

We are seeing this bottleneck resolution happen in real time at the software level. People use a tool, hit a wall, and say it would be easier if the software could just do X. Then X becomes the next problem to solve. The pace of solving X used to be slow. Now it is nearly instant. The models are recursively improving. The technical or design bottlenecks a developer might hit are now being addressed by agents so that other agents can work more easily. If I am using Claude to build something and it says, “I need you to get me ____ to continue/configure ___,” I just say, “you do it,” and lo and behold, it breaks through the jam on its own. The level of human-level task completion is hitting crazy levels.

Claude Code is adding significant new features multiple times a day. Because of its own capabilities, the engineers at Anthropic building it have an unprecedented, continuous update cycle where they are no longer writing code at all. The friction between identifying a missing capability and deploying it is collapsing toward zero. While you use the tool each day, you can see the exponential change happening. This is completely invisible to people who only see upgrades every few months when a new default model rolls out in ChatGPT for standard writing tasks.

Forecasts to Take Seriously, If Not Literally

Two pieces have gone viral in the past few weeks that perfectly map to this acceleration. Both are directionally much closer to reality than most people I’m talking to seem willing to accept.

Matt Shumer's "Something Big Is Happening" used a COVID analogy, which I think is incredibly accurate for what we are living right now. He compared the current AI capability curve to February 2020. Everyone knew the virus was here, but the public was mostly ignoring the exponential math right before the entire world changed. His core point is right. You cannot grasp what is happening from the chat window.

Citrini Research went further with their "2028 Global Intelligence Crisis" memo. They modeled a 2028 scenario where agentic tools hit a step function and create a deflationary cascade. This is the death of the middleman. Autonomous AI agents optimize every transaction and bypass legacy software platforms. This creates a feedback loop with no natural brake. Companies use AI to cut costs, invest those savings into more AI, and displace more workers. Citrini coined the term "ghost GDP" for the result. Corporate profits soar while consumer spending collapses because displaced knowledge workers cannot redeploy their skills. Machines spend zero dollars on discretionary goods.

Critics have point out legitimate flaws in specific examples. They cite the marketplace power and logistical moats of businesses like DoorDash. They may be right in those specific cases. But even if Citrini is 50% wrong, the world as we know it is in for a huge awakening. I am personally preparing for a world in which the macro thesis and feedback-loop mechanics of Shumer and Citrini are overwhelmingly accurate. I am obsessively mining opportunities that can most gain leverage from this technological progress while remaining insulated from the coming dislocations. Specifically in things like manufacturing, industrial services and the like. But this might still be my own form of denial.

Block Is the First Domino

We are already seeing the first-order consequences of this theory in practice. On Thursday, Block (fka Square, the company whose iPad registers you tap at most coffee shops and local stores) cut over 4,000 people, reducing their workforce from more than 10,000 to just under 6,000. The stock immediately jumped over 20%.

They did not do this because the business was struggling. Block reported Q4 gross profit of $2.87 billion, up 24% year over year. CEO Jack Dorsey was explicit in his shareholder letter. He wrote, "A significantly smaller team, using the tools we're building, can do more and do it better. And intelligence tool capabilities are compounding faster every week."

Block is a pioneering tech company. Legacy, corporate behemoths with twenty layers of management will soon wake up and realize they need to fix their bloat or be crushed as a result. Jack said he believes the majority of companies will reach the same conclusion and make similar structural changes within the next year. He chose to act now rather than be forced into it reactively. Microsoft's AI chief recently warned white-collar workers have 12 to 18 months.

So, a profitable S&P 500 company cuts 40% of its workforce, the market rewards it with a massive pop. Every CEO has to be watching and wondering what they can be doing. Right?

The True Cost of Denial

Why are people missing this? Why aren't the markets in more of a panic?

Part of it is maybe the hope for a regulatory bailout, especially when you consider our election cycle. This week, the government banned Anthropic from the military supply chain due to their stance on safety. But Anthropic has a complicated relationship with Washington given its left-leaning roots and this was the DoD not wanting to be constrained in how they access and use AI.

You could see other types of attempts to regulate AI in the opposite direction when the unemployment rates spike, to slow the pace of disruption. But open-source models are on pace with closed-source ones. The genie is fully out of the bottle. Note that the Citrini scenario is dated 2028, placing the primary impact of this disruption right as we enter the 2028 presidential election in the USA.

The main reason people are missing this is much more human. It comes down to incentives. I have to check my own biases on this constantly. If your compensation, status, and identity depend on maintaining the current book of business, your brain will actively reject information that threatens it.

When you have a boss demanding a status report, a client flight to catch, and a mortgage to pay, you do may have the luxury of spending four hours a day testing frontier AI models. You are explicitly focused on the immediate task. Corporate IT systems often block these tools anyway. Most importantly, nobody wants to look directly at the thing that challenges the underlying value of their life's work.

I do not have a magical bunker to hide in. I am trying to figure this out just like everyone else and one of the ways I’m doing this and feel better prepared than most to navigate this scenario is because I’m experimenting with these capabilities and building constantly. I’m working with our portfolio founders to push them to reassess their priors and business models in this context.

But the bumpy road ahead is getting clearer by the week. If you rely on a W-2 income for a specialized cognitive skill, the ground is shifting. The only defense I see is leverage. We either learn to architect the systems that automate the middleman, or we become the middleman who gets automated. The people who make it through this transition will not necessarily be the smartest engineers. They will simply be the ones who aggressively embed these capabilities into their daily workflows.

The visibility gap is the real risk here. Not whether disruption is coming, but whether we see it in time to position ourselves on the right side of the K.

Thanks as always for reading,

Jake Dwyer

Founder, Factor Capital