Every week the news tells you AI is transforming the world. Every week you open ChatGPT and use it to fix a paragraph in an email.

If you’ve noticed the gap between those two signals and quietly wondered whether you’re the problem, you’re not. The gap is the phenomenon. It’s the most underdiscussed dynamic in technology right now, and it’s what this piece is about.

The standard story is that AI is incredible but the tools still need to mature, the use cases will get clearer, and we’ll all figure it out once the product gets better. That story is comforting because it puts the responsibility on the technology. I’ll start using it seriously when it’s ready for me.

I think the story is incomplete in a way that misses the most important part. The thing that takes a generation to mature isn’t the product: It’s us.

Let me show you what I mean.

The Law

There’s a pattern across every transformative technology in human history, and once you see it, you can’t unsee it.

The more a tool can do, the more skill it takes to operate, and the more imagination it takes to actually find it useful. Function, skill, and imagination move together. They’re not three separate variables. They’re a single curve.

A plow has one job. You can learn to use it in an afternoon. The imagination cost is essentially zero, because the function is bounded. You don’t sit around brainstorming creative applications for your plow. You plow the field, the move on to the next task.

A spreadsheet has many jobs. You can learn the mechanics in a week. But most people who’ve used Excel for twenty years still use less than five percent of what it can do, because the imagination cost has outrun the skill cost. The function space is vast and they can’t picture most of it. If you need a function, you’ll learn it. But if you don’t know it exists, and more importantly can’t imagine that it should exist, then you won’t know to ask: this is the paradox.

AI is the first tool in human history where the function is effectively unbounded. There is no list of things it does. There is no manual you can finish. Every domain of your life is potentially in scope, and that’s exactly the problem. The skill ceiling is real but reachable. The imagination ceiling is wherever your mind can reach. Just like you thought the issue with Excel was skills and AI is literally why you go to learn those skills, the reality is much deeper: the spark of an idea that the skill is even possible is where it actually comes together. AI isn’t a linear progression, it’s an inversion of what we’re used to, and yes, it might be the final boss.

This sounds abstract. It isn’t. It’s the thing creating the dissonance you’re feeling right now.

Why This Matters Now (And Didn’t Before)

Here’s the question that should be asked: who cares if skill and imagination are different things, as long as they move together?

In past technology transitions, the answer was: nobody. They moved together on a slow enough timeline that the distinction didn’t sharpen into a personal problem. The factory owner in 1895 wasn’t lying awake wondering whether he was personally behind on dynamos. The clock was slow, the gap closed over decades, and the cost of being a slow adopter was diffused across a long career.

AI is the first technology where the timeline has compressed enough that the imagination gap creates real-time personal dissonance.

Every week the news says it’s eating entire industries, replacing knowledge work, generating billion-dollar companies in months. Every week your own use of it feels modest, and there is always someone saying “omg, you have to use this new AI tool - it’s perfect for you,” and yet you feel an internal resistance - not because you’re afraid of learning a new skill (after all the AI makes that nearly seamless), but because you’ll have to do the real work: restructure your thoughts and processes to get the most out of a new tool - this requires imagination and creativity. It’s what makes us human, but it’s also tiring.

Like I mentioned in last week’s essay, we have been tricked into thinking that productivity is pushing paper (even if it’s a sophisticated version of it), but the deep human work is actually a different beast altogether. Not being able to hide behind paper pushing to feel productive is intimidating. Routine operations feel comfortable. Creativity is daunting.

The internal feeling and then inbound messaging don’t seem to match. And the gap between them isn’t being papered over by time, because there isn’t enough time.

So people start to suspect they’re the problem.

They’re not. The mismatch is the phenomenon. The clock is faster than imagination, for the first time in history. Naming that is the first step out of it.

The Pattern, Three Times

If this is the first time you’ve encountered the imagination-gap idea, you might (reasonably) want some evidence that it’s real rather than just a convenient story. So let me show you the pattern across three technologies. One that nobody alive has used, one that everyone uses without thinking, and one you remember learning.

The Telegraph (1844 → ~1880s)

Morse sent “What hath God wrought” in 1844. The function was complete on day one. Near-instant communication over any distance you could string wire.

You’d think the use cases were obvious. They weren’t.

It took roughly forty years for society to develop the actual social grammar of the telegraph. What’s worth sending versus what should still go by letter. How to write a message when each word costs money. The genre of “breaking news.” Wire services. The convention that you telegram for deaths and write for births. The US State Department didn’t integrate telegraphy into actual diplomacy until after the transatlantic cable in 1866, twenty-two years after the function existed and was sitting there, waiting.

The technology was instant. The social knowledge of what to do with it was not.

The Telephone (1876 → ~1900s)

Then there’s the most famous artifact in the history of technological misjudgment. A widely-quoted internal Western Union memo from 1876 (possibly apocryphal but consistent with documented attitudes at the company):

“This ‘telephone’ has too many shortcomings to be seriously considered as a means of communication. The device is inherently of no value to us.”

Western Union owned the telegraph monopoly. They had every commercial incentive to recognize the future. Bell offered them the patent for $100,000. They turned it down. Within two years it was worth more than $25 million.

It wasn’t because they were stupid. They genuinely could not picture what a telephone was for. Even Bell himself initially imagined it as a sort of broadcast device - early operators ran “théâtrophone” services piping live opera into subscribers’ homes. The idea that the killer use case was two ordinary people having a casual conversation took years to crystallize.

David Sarnoff lived through the same pattern at the other end. In a 1916 memo, he proposed that radio could be used to pipe music into people's living rooms. His superiors at American Marconi sat on the idea for years. They couldn't picture what a radio was for beyond point-to-point wireless telegraphy - exactly the same imagination gap, exactly the same generation later.

Email (Late 1980s → Ongoing)

And here’s the one you remember.

Function arrived on desks in the late 80s and early 90s. The mechanics were trivial. You typed, you sent, it arrived.

But the social grammar took twenty years. Anyone over forty remembers the awkward decade where nobody knew whether an email was formal or casual, whether you needed a subject line, whether silence meant assent, whether you should CC the boss. Should this be a meeting? Should this be a phone call? Should I have just walked over? Email etiquette is still being worked out, decades later.

The function arrived overnight. The imagination of what to do with it took a generation.

These three are the easy examples. Economists call this the productivity paradox - Paul David documented a roughly forty-year gap between Edison opening his first power station in 1881 and electrification finally showing up in factory productivity statistics in the 1920s. Factories had electricity for decades; they just kept using it to power the old shaft-and-pulley systems instead of redesigning around individual motors per machine. The function arrived in 1881. The imagination of what a factory could look like took until the 1920s. Same gap with personal computers: Robert Solow’s famous 1987 line that “you can see the computer age everywhere but in the productivity statistics.” PCs had been on desks for a decade. The productivity gains didn’t show up until the mid-90s.

The pattern is so consistent that the only honest expectation for AI is that the same thing is happening right now.

It Was Never the Product

So why does it take so long?

The standard explanation - the one most people accept by default - is that the product needed to get better. Early telephones were scratchy. Early computers were slow. Early AI hallucinates. Better product, more adoption.

This story isn’t wrong. It’s just radically incomplete. And the part it leaves out is the part that matters to us - the users.

The imagination gap doesn’t close through product improvement. It closes through social exposure.

People learn what a technology is for by watching other people use it. By hearing a friend describe what they did with it last weekend. By reading a profile of someone running their life through it. By trying something themselves, failing, and trying again. The technology can be objectively identical on Tuesday and Wednesday - what changes between Tuesday and Wednesday is that on Wednesday, you heard your sister-in-law explain how she’s been using it to plan her meals, and now you can picture yourself doing something similar.

The factory owners in 1900 weren’t waiting for better motors. The motors were fine. They were waiting - without knowing they were waiting - for someone to be the first manager to put a motor on every machine and redesign the floorplan, and then for that idea to spread by word of mouth from factory to factory until it became thinkable everywhere.

The telephone didn’t take thirty years to mature because the sound quality was bad. It took thirty years because people had to learn what a phone was for by watching their neighbors use one.

Adoption isn’t imagination. Adoption is the first inch. Imagination is the mile.

This is the part of the story that doesn’t get told, because it doesn’t flatter anyone. It doesn’t flatter the tech companies, who’d like you to believe their product is the bottleneck. And it doesn’t flatter the user, who’d like to believe they’re just waiting for the right moment. The truth is that the bottleneck has always been the slow human work of figuring out what to do with new capability, and that work has always required people watching other people.

The Four Phases

You can actually watch this happening right now, in your own conversations, if you know what to look for.

Six months ago, when I’d ask someone what they were using AI for, the answer was usually some version of “oh, everything.” Which means nothing. Everything is what people say when they have access to a tool but no working picture of its use cases. Enthusiasm without specifics. Awareness without imagination.

In the last few months something has changed. Now when I ask the same question, people give me real answers. “I used it to figure out whether my landlord’s rent increase was legal in my state.” “I used it to plan a week of meals around the protein target my trainer gave me.” “I used it to draft a hard message to my sister.” The shift from everything to specifics is the imagination gap closing in real time, in your social circle, on a timeline of months rather than decades.

I think there are four phases everyone passes through, and recognizing which one you’re in is the first useful diagnostic:

Phase one: Nothing.
You haven’t tried it, or you tried it once and didn’t see what the fuss was about. Most of the people who signed up for ChatGPT and never came back live here. Awareness without use.

Phase two: Everything.
You’re using it, you’re enthusiastic, you can feel that it’s powerful - but you can’t articulate what for. This is the phase where people post “AI is changing everything” without ever saying what they did with it that morning. It feels like progress, and it is better than phase one, but it’s still pre-imagination. You have access to the function but not yet a working picture of the use cases.

Phase three: Specifics.
You can name three things you used it for this week, and they’re not “writing” or “research” - they’re concrete and they span domains. This is what imagination actually looks like in practice. It’s specific. It’s portable. You can teach someone else what you did, which means you can spread the imagination forward.

Phase four: Invisible.
The tool stops being a thing you talk about. It becomes infrastructure. You don’t say “I used email to send that” - you just sent it. A small number of people are starting to reach this with AI. The rest of us are not close yet.

If you can’t give a specific example of how you used AI this week, you’re still in phase two. That’s not an insult. It’s a description of where the social imagination process has carried you so far. And it tells you what to do next.

A How-To For Closing The Gap On Yourself

If the gap closes through social exposure and personal experimentation, then the prescription writes itself. You don’t have to wait for the rest of the population to drag you along. You can run the process on yourself, deliberately, faster than it would otherwise happen.

Here’s how.

1. Start where the stakes are low.

The reason most professionals are bad at AI is they only try to use it for high-stakes work, where being clumsy feels expensive. So they don’t try, or they try once and bail. The fix is the opposite. Use it for things that don’t matter.

My own entry point was personal fitness and nutrition. Nothing was at risk if the advice was mediocre - my alternative was a sea of YouTube videos and arbitrary google search or anecdotal advice. I just kept asking. What should I eat before a workout? How do I structure a week of training around this goal? What’s actually wrong with this rotator cuff? Over months, I developed a feel for what kinds of questions get useful answers and which ones don’t. By the time I started using it in higher-stakes settings at work, I had hundreds of reps in.

The point of low-stakes use isn’t the output. It’s the reps.

2. Talk to it like a person.

The single weirdest thing about the current moment is that most people still talk to AI like a search engine. Three-word queries. Stripped of context. As if you only get to type once.

Talk to it like a colleague. Tell it who you are. Tell it what you’re trying to accomplish. Tell it what you’ve already tried. Tell it the constraints. The amount of context you give is the strongest single predictor of whether the output will be useful.

Instead of sending your friend a 2 minute WhatsApp voice note describing your situation to ask a simple question, do that with AI. Context is the super power and unlike WhatsApp voice notes to your friends, the AI won’t get annoyed with you.

3. Push past the first answer.

The first response is almost never the useful one. The skill is in the follow-up. Make this more specific. Apply it to my actual situation. Argue against this. What am I missing?

Most people stop at the first reply and conclude the tool is mediocre. The tool is not mediocre. They stopped too early. It is not a search engine, but it knows how to use one. If the answer doesn’t feel right, keep pressing.

4. Collect use cases from other people.

This follows directly from the social-imagination point and almost nobody does it deliberately. When a friend mentions using AI for something that hadn’t occurred to you, write it down. Try it that week. The imagination of the people around you is one of the most valuable inputs you have access to right now, and it costs nothing.

Better yet, when someone says they’re using AI “for everything,” push. Like what, specifically? Not as a gotcha. As genuine intelligence-gathering. You’re trying to harvest specifics from the ambient social field. The more you collect, the faster your own phase-two evaporates.

5. Move it up the stakes ladder.

Once the patterns are familiar from low-stakes use, port them into work. The fluency transfers. The confidence transfers. And you’ll find yourself doing things your peers don’t, because you’ve practiced in a domain they haven’t.

The muscle is unpacking a task or a query in the work setting into the quanta that AI can properly handle. Workflow is mastery.

6. Notice what you can’t imagine asking.

This is the meta-skill, because imagination is the actual ceiling. Pay attention to the moments when someone else uses AI for something it never would have occurred to you to use it for. That is the gap. Those moments aren’t accusations. They’re maps. They show you exactly where your imagination has further to go.

The Acceleration Caveat

One thing in our favor.

Generations are shorter now than they used to be. The telegraph took forty-plus years to mature socially because information moved by ship and rail and newspaper. The telephone took thirty. Email took twenty. Smartphones took five. Each cycle of social learning is compressing, because the channels through which we talk about technology have themselves accelerated.

So the AI imagination gap will close faster than the dynamo gap did. Probably much faster.

But, and this is the part the adoption numbers obscure, it will not close as fast as the headlines suggest. ChatGPT hitting 100 million users faster than any product in history is not the imagination gap closing. It’s awareness, not imagination. A hundred million people have tried it. Far fewer have any real picture of what it could do for them. The number who use it routinely for anything beyond writing and search-replacement is smaller. The number who use it imaginatively - in their domestic lives, their finances, their health, their parenting - is smaller still.

The gap between adoption and imagination is where you live for the next several years. It’s also where the leverage is. Everyone who’s only counting users is missing the actual race.

Everyone wants a hack - my friend is getting tons of value for the same $20 per month. What am I missing? Just tell me what to do…

The hard, but consistent answer: put in the work. Experiment. Think. Imagine.

Coming Back Around

Every generation tells itself that its technology is the one that will finally be different. That this one will meet people where they are, that the use cases will be obvious, that the imagination gap won’t apply because the product is just that good. Every generation has been wrong, and the people writing breathless trade-press coverage of AI today are wrong in exactly the way Western Union was wrong about telephones.

The work has always been the same, and it has never been the product’s work. It’s yours.

The good news is that the work is doable. It doesn’t require special access or technical training or a particular kind of brain. It requires reps, attention, social curiosity about what other people are doing, and a willingness to look slightly silly trying things that don’t pan out.

If you’re reading every breathless AI headline and feeling like you must be missing something, you’re not behind. You’re early. The hype-vs-personal-experience gap you’re feeling is the most honest signal in the conversation right now. It tells you the social imagination of the population hasn’t caught up yet. Including yours. Including mine.

That’s a feature, not a bug. It means the territory is still open.

Take the reps. Talk to it like a person. Push past the first answer. Steal use cases from your friends. And pay attention to the moments when somebody does something you’d never thought to try - because that’s exactly where you go next.

Subscribe now

There’s an old parable about a businessman who meets a fisherman on a dock.

The businessman tells him he should work harder. Scale his operation. Buy a fleet. Build the corporate infrastructure. Take it public.

The fisherman hears him out and asks, “and then what?”

“You retire a rich man,” the businessman says. “What will you do with all that spare time?”

The fisherman replies: “Sit on the dock and go fishing every day. Exactly what I’m doing right now.”

I love this parable because it captures the hidden cost of an idea that’s run our economy for the last hundred years: the long short path. It looks like the most direct route to what you want - but the route around (the short long path) is shorter when you measure what actually matters. The fisherman skipped the empire and went straight to the destination. That's the short long path. The businessman wanted him to take the long short one.

The classic image: climbing a mountain straight up the cliff face is technically shortest. The snaking switchback path takes longer on the map but gets you there faster, with less suffering. Both reach the top. But you’d better know which one you’re on and have the right equipment.

In business: someone sets a goal of being a CEO. Option one - start a company from scratch right now. Goal achieved. But is it actually what they wanted? Or is it a lot of annoying, risky, exhausting things that take them away from the reasons behind the goal? Option two - join a company, work their way up, move diagonally if needed, until they’re promoted or hired as a CEO. Same destination. Completely different journey.

For the last hundred years, the global economy has been the businessman on the dock, pushing everyone toward the long short path. We invented increasingly complex corporate structures, management layers, and workflows largely to solve problems our own complexity created. We built massive, exhausting fleets when all we really needed was the fish to eat and all we really wanted was to enjoy our days.

Here’s why this matters right now: AI just deleted the need for the fleet.

The path that looked safe for a century - build skills, climb the ladder, scale headcount, accumulate prestige - has quietly become the long short path for almost everyone. The path that looked slow - deep relationships, judgment, taste, contemplation - is now the short long path. If we do it right, I think we can have the best of both worlds, and more importantly stay in control of our own unique path.

This piece is about how we got here, why the rules are flipping right now, and what to do about it. It’s about a 250-year contract between employers and employees that’s being rewritten in front of us, with major case studies already on the public record. And it’s about why the most valuable thing you can do this decade is also the simplest thing: take the suit off.

Let’s walk through it.

The 250-Year Arc of Leverage

Every era of work is defined by a simple question: what does the employee bring, and what does the employer control?
(This is textbook people leverage, as I outlined in my 2020 piece “Leverage Frameworks”)

The answer to that question determines who holds the leverage. And the leverage has shifted - dramatically - five times in the last 250 years.

There’s a pattern across these eras that I want you to watch for. Each one teaches a surface lesson at the time - “learn this, do this, you’ll be fine.” And in every era, the surface lesson eventually misses the mark. But the side effect of having learned it - the second-order skill or dynamic people picked up almost by accident - is what actually carries forward. History doesn’t repeat, but it rhymes. The rhyme is the side effect. The further back you look, the more obvious the side effect is. The closer you get to today, the more it’s a working hypothesis.

Pre-Industrial (Pre-1800s): Craft & Local Trust.

The employee brought specialized physical skill - craftsmanship, agriculture, localized customer relationships. The employer (really the patron or guild master) provided sustenance, tools, and trade access. If you were a skilled artisan, you had real leverage because your craft was scarce and your reputation and brand was local currency. If you were unskilled labor, you had nothing.

The side effect: The literal craft skills depreciated - nobody needs a blacksmith now. But the underlying dynamic - your reputation and brand in your community as the actual asset - never went away. Mass scale temporarily made it look irrelevant, but it was always the substrate. AI is bringing it back, full circle.

Industrial Revolution (1800s – Mid-1900s): Capital & Scale.

The employee brought muscle and time. The employer brought the capital for the million-dollar factories, supplier networks and customer lists. This is where the power dynamic got brutal - and where the structural change was most violent. In 1900, 41% of Americans worked on a farm. By 1970, that number was 4%. An entire way of life was absorbed by the factory in 70 years. The employer owned the means of production - full stop. Employees had to unionize just to create any collective leverage. Compensation was hourly cash. You were a unit of output. People were cogs in a machine and worked in factories with easily measurable productivity metrics.

The side effect: The lesson everyone took home was “show up, work hard, the system rewards muscle.” But the people who actually captured the era’s wage premium were the ones who understood the machines - the mechanics, the foremen, the engineers - not the ones who just operated them. The pattern: learn the system, not just the surface.

Introducing Corporate America (Post WWII to early 1990s): Welcome to the Office.

Desk-filled offices similar to what we have today with one very notable difference: no computers. People used slide rules to close accounting books. Reviewed handwritten notes and contracts. Typed dictated letters. An analog world but on steroids. Think about Don Draper’s office in Madmen. The secretary became a communication enabling staple. Pan-am flights across the country to close big deals. It was faster and more connected than anything anyone had ever seen.

The side effect: The promise was “the company is family, give them 30 years and they’ll take care of you.” That promise died with pensions and the 80s-era restructurings. But the relationships built inside those offices outlasted the promise. The Don Draper / Roger Sterling friendship was worth more across a career than any title or benefits package. Titles are temporary. Relationships compound.

Technology 1.0 (Late 1900s – 2009): Knowledge & Processing.

The employee became the “knowledge worker.” Cognitive labor, specialized technical skills, middle management, information routing. The employer provided the software infrastructure, salaries, benefits packages, and early stock options. Capital was still expensive, but companies couldn’t scale without specialized human brains. Leverage started shifting toward talent. Everyone got a laptop and a cell phone. Connectivity abounded but the office dynamic stayed strong. Similar specialization, but increasingly at light speed.

The side effect: “Learn to code” and “learn to use Excel” were the slogans. What it likely taught, for those who really took the classes, was structured problem decomposition and how to navigate large swaths of data. The skill of breaking a fuzzy real-world problem into pieces a machine can execute against. We thought we were learning syntax, but we were learning to communicate with machines and solve previously unsolvable problems as a team.

Technology 2.0 (2009 – 2020): Networks & Agility.

The employee brought niche expertise, personal brand, and networked relationships. The employer countered with equity, platform access, and flexible work models. This is where the split first appeared: massive leverage for “10x” talent who could demand RSUs and remote work, near-zero leverage for gig workers who were algorithmically managed and completely commoditized. The new path to getting rich became 20 years at a FAANG company. New business models, deeper technology reach, instant connectivity.

The side effect (still emerging): The advice was “get a job at FAANG, vest your RSUs, build your resume.” The actual side effect was surviving in the fast lane. The mechanisms to separate the wheat from the chaff became ruthless. It was much harder to hide if you couldn’t keep up. Payroll mistakes didn’t allow the “digital collator” to keep his job, even by accident. Influence, audiences, thought leadership spread quickly and accrued tremendous value - even more than $250k+ average salaries.

The AI Era (2020s – Present): Judgment & Orchestration.

The employee brings abstract problem-solving, personal networks, strategic taste, empathy, and the ability to direct autonomous agents. The employer provides compute, proprietary data, and what I’ll call the “suit” - more on that in a minute. Enter the era of the 100x employee.

The side effect (predicted): The slogan today is “learn to prompt, master the AI tools.” If the last five eras are any guide, that’s the surface lesson that won’t carry. The thing that will carry is judgment, taste, and trust. Bet on the side effect.

When you zoom out across 250 years, the trajectory is clear. The core unit of human work evolved from Muscle → Mind → Machine Orchestration → Whatever comes next. And compensation evolved from sustenance → hourly wages → salary → retirement plans → equity → something we haven’t fully defined yet.

But here’s where it gets interesting.

The Collapse of the Middle

In the Tech 1.0 and 2.0 eras, companies relied on a massive middle layer. These were the human routers. They synthesized data, moved projects between departments, wrote boilerplate code, formatted slide decks, and summarized meetings. They were the connective tissue of the organization. Massively inefficient but also somehow highly functional (we’ve seen this movie before).

AI doesn’t just do this work instantly. It does it at near-zero marginal cost.

I remember the first time I heard the term “product manager.” We all know that a great one makes all the difference, but most are paper pushers. They’re the “people person” from Office Space who handles the requests between the clients and the engineers. And then it really started to creep back in: the layers that exist just to interface with the other layers. Peak inefficiency. We say it now as a joke but “I’ll have my people talk to your people and coordinate” was a real thing, and one we certainly don’t need in the age of phone-based calendar invites and text messaging.

And here's the part I want to make sure people sit with: we have watched this exact movie before.

While the product manager as paper pusher reference may have triggered some of you, remember that for most of the 20th century, the largest single occupational category in America was office and administrative support: typists, secretaries, clerks, coordinators, filing staff. The connective tissue. By the time the personal computer started landing on every desk in the early 1990s, this category was the spine of corporate America. Then word processors arrived, and managers started typing their own letters. Calendars went digital. Spreadsheets replaced the bookkeeping clerks. The typing pool (a fixture of office life for 80 years) simply dissolved over a generation. The Bureau of Labor Statistics has tracked office and administrative support employment falling steadily ever since, and they project the decline to continue through at least 2034.

The middle layer didn’t disappear all at once. It transformed while quietly halving over a generation.

We are now watching the same pattern start with the analyst pool, the coordinator pool, the synthesizer pool. Different category. Identical structural dynamic.

The result is that the middle layer is structurally collapsing. And it’s forcing a rapid, brutal bifurcation of the knowledge workforce into two very different paths.

The middle layer always collapses. The low level jobs are an obvious cost cutting target and the senior ones are high-risk, high-reward (and not for everybody). It’s those people in the middle who “play it safe” who are at the biggest risk. The ones whose parents tell them “just do this and you’ll always have a job… after all, every company needs a typist or a scheduler or a bookkeeper or a lawyer.” The safe bet from last generation is just that: a thing of the past. Winning yesterday’s game is losing today’s.

This is micro economics followed by innovation economics:
A functional need across all industries grows. Salaries are great. Many people flock to that job function. It becomes a major cost center. Innovation cannot ignore it any longer and comes after this tax on the economy. Prices get driven down by competition and then ultimately to zero when the innovation takes its place in one fell swoop. Capitalism.

The Downward Path: The Model Feeder.

If your job can be mapped into a standard operating procedure, your leverage is evaporating. Your previous output - writing, basic analysis, routine code - is now automated. Your residual value to the employer is generating edge-case data that the AI can’t yet produce on its own. You’re wearing sensors. You’re doing RLHF (reinforcement learning for human feedback). You’re capturing video and labeling images. You’ve become a biological API. And your compensation will look a lot like factory floor labor in the 1800s: piece-rate, low-wage, no equity, no upside. And then it will be gone - as soon as it starts becoming too big of a cost center, the innovation cannons will fix their aim, and then it’s only a matter of time. This is actually capitalism doing its best work.

This isn’t theoretical. Time magazine reported in early 2023 that OpenAI used Kenyan workers, paid between $1.32 and $2 per hour through an outsourcing firm called Sama, to label graphic content (child abuse, torture, self-harm) that ChatGPT needed to learn what to filter out. Multiple workers reported lasting psychological damage. The total contract OpenAI paid for the work was around $150,000. That’s the new piece-rate factory floor: traumatic, invisible, low-wage, and entirely upstream of a product that now generates billions. The model feeders aren’t a future risk. They already exist.

The Upward Path: The Orchestrator.

This group absorbs the leverage that the middle layer left behind. They don’t do the work, they direct it. From 5 direct reports to 0. They define the problem, navigate ambiguity, and apply strategic taste to AI output. The headcount-to-output ratios in the orchestrator class are unlike anything we’ve seen in the modern economy. Their compensation reflects it: significant equity, profit-sharing, high base, and real ownership of outcomes.

Many of you reading this hope or believe you’re in this layer, and you very well might be. But think hard about the linear progress of AI and make an argument for why you matter. It’s a bit harder to do than you may feel comfortable with.

But even the highest level ADHD orchestrator needs a break. Slowing down is starting to become the superpower. You can have 20 AI agents performing tasks for you at the same time, but if you can’t keep track of what is going on and why it matters, then it just becomes a slop train. It’s always obvious when this is happening right now. In a year? Maybe it will only be obvious half the time. Having your interns do your work for you never worked out well if you weren’t over their shoulder double checking to make sure that the right dose of judgment was sprinkled into the grunt.

The economic ethos around work that I grew up with: get rich by getting as much operating leverage as possible and then never sleeping. I think this breaks now. I saw people trying to do this in investment banking: a team of VPs, associates and analysts below you. Access to the world’s data. As many clients as possible. 1-2% fee chunks out of multi billion dollar deals. Leverage in almost every capacity. But if you look at their faces, their bodies and then judge their work on what really mattered, you could see something different: misery, middling wealth, and no body or time to get the most of it. Then there was the legendary banker who spent a day a week in the office. The rest of the time he was on the golf course building relationships. But he was also doing something else: giving himself the mental, emotional and physical space to contemplate what actually mattered most in his job. And having run the compensation model, I can assure you that his approach paid better. He had leverage where it mattered most.

The bottom layer was already gone, the Industrial Revolution took care of that. The top layer still matters. But the middle is where the action is right now. The middle is splitting, and every person in it is facing a choice: get elevated, or get pushed down.

Klarna: A Parable in Real Time

If you want to see this thesis play out in a single corporate case study, look at Klarna.

In February 2024, Klarna announced that its AI customer service assistant (built with OpenAI) was doing the equivalent work of 700 full-time agents. The numbers were stunning. 2.3 million customer conversations handled. Two-thirds of all customer service chats automated. Resolution time dropped from 11 minutes to under 2. Projected profit improvement of $40 million in a single year. The company froze customer service hiring. Workforce went from 5,500 to 3,400 over the period.

This was, for a moment, the canonical proof point of the AI workforce transition.

Eighteen months later, Klarna quietly walked it back.

CEO Sebastian Siemiatkowski’s public admission, reported by Bloomberg in 2025: “We focused too much on efficiency and cost. The result was lower quality, and that’s not sustainable.”

What happened? The AI was excellent at routine work. It still handles two-thirds of customer chats. The 82% reduction in resolution time held. But on the cases that actually mattered - disputes, fraud claims, hardship situations, anything requiring empathy or judgment - customer satisfaction cratered. The AI couldn’t carry the trust load. Klarna started rehiring humans, specifically for cases that needed empathy, expertise, and real conversation. New positioning from the company: “in a world of automation, nothing is more valuable than a truly great human interaction.” And the skills they screened for also shifted. The jobs are shaped by the technology very soon after the technology is shaped by the jobs. Functional output is the metric that matters: capitalism 101.

Read that arc carefully. This is not a story about AI failing. It’s a story about AI succeeding at the routine middle and then hitting a wall at the trust and empathy layer. The most aggressive customer-service AI deployment on record validated the framework I’m describing in real time, with real money, in front of public markets. AI ate the middle. Trust stayed human. The CEO admitted it on the record.

Compelling vs. Unique

Here’s where most people and most companies are missing the mark.

The corporate world is panicking. Their instinct is to pour money into “AI Upskilling,” teaching employees how to write prompts, connect APIs, build workflows. Every Fortune 500 company is buying some version of an AI training program right now.

This is a trap. And it’s a specific kind of trap we’ve watched play out before.

I went to Stanford 2007-2011. During that time computer science became the #1 major and CS106a became the most well-attended class. “Learn to code” they said. “You’ll always have a job” they said. And that was true (especially for Stanford alumni) for a while. But then Claude Code came and ate every “coder’s” lunch in 2025.

With one small but absolutely critical caveat: what they said (”learn to code”) was a thinly veiled version of “learn to solve complex problems by thinking like an engineer.” And what is an engineer? A well-trained creative puzzle solver. And that is why those people still have jobs. If you actually took the classes, you would know that the core competence coming out of Eric Roberts’ book on Java was to solve puzzles using language that computers can understand. Well now we have the greatest translators in human history at our fingertips, but the puzzle solving part - the real puzzles - the ones that require creativity - that is the part that’s left. And it turns out that’s what mattered all along.

AI skills are the same thing. Technical prompt engineering is a commodity racing to zero. The models are explicitly designed to remove the need for you to know how to talk to them. In two years, knowing how to prompt an AI will be as irrelevant as knowing how to write command-line code to use an iPhone today.

There’s a frame I keep coming back to that I picked up from Pat Grady: compelling vs. unique. Compelling technologies are easy to sell. There’s real demand, real revenue, real growth. They IPO. They get hyped. They’re real businesses. But they don’t have pricing leverage, because anyone can build the same thing. Unique technologies have the missing piece. Often harder to sell at first. But once you’ve proven they work, you have pricing power, because nobody else can replicate them. To be a great company, you need to maintain both.

Compelling and unique might look the same in the early days. Both have growth. Both have customers. Both feel important. The difference only shows up later: when the thing that was unique stays unique and the thing that was merely compelling gets eaten by competition. AI didn’t kill SaaS, SaaS killed SaaS.

Quick history lesson. Nobody remembers the companies that got rich helping enterprises “move to the cloud” in the mid-2010s. The systems integrators that defined that era were intensely compelling. They got hyped. They IPO’d. And then the multiples compressed. Cloud became infrastructure. The integration layer a commoditized service. The cloud migration darlings of 2014 are mostly still around, they just don’t trade at growth multiples anymore. They became utilities. Important. Critical even. Unloved.

Same pattern is starting with AI integration today. AI integration companies are doing genuinely useful work: wiring AI into enterprise workflows. That’s compelling. But is it unique? Ten competitors will appear in the next two years. The integration patterns will standardize. And the multiples will compress, exactly the way they did in every prior wave of integration plays.

Here’s the part that should worry every founder, every operator, every employee: things are becoming not-unique faster than ever. What used to take a decade to commoditize now takes 18 months. The half-life of “unique” is shrinking. Truly compelling things won’t stay unique for long. That’s a feature of capitalism, not a bug, but it changes how you think about what to bet your career on. The compelling skill you’re learning today might already be a commodity by the time you’re paid for knowing it.

Cal Newport has been making a related argument I think is right: instead of rushing to learn AI tools, people should go read the literature. Become a deeper, more thoughtful person. Build real judgment. Because that’s what will drive value when the tools are commoditized.

Companies are spending billions trying to teach humans to act like software at the exact moment that software no longer needs our help. That money is going to the wrong side of the ledger.

The example we use in both finance and engineering is “garbage in - garbage out.” This describes any system or model and says that the quality of outputs is defined by the quality of the inputs. For many years, that meant clean, accurate, formatted data. Now it means something different: the information inside the data is the cornerstone. The formatting doesn’t matter any more. But today there is an odd expectation that putting nothing into AI will somehow magically result in good outputs when the AI is good enough. That my friends is how we define AI hallucination. You have to give it something. Then it can go to work. I have said before “If the AI could read your mind, the output wouldn’t be any better… because there’s nothing going on up there. Clarify the idea for yourself first, then ask AI to do something.” There is a valuable skill in typing a prompt, it just may not be the one you think it is.

So Where Does Value Actually Live?

If execution is infinite and practically free, what is the most expensive asset left on the board?

It’s not orchestration. Even that will commoditize as agents get better at managing other agents. The “technical” skill of directing AI will flatten, just like coding, just like prompting.

What’s left is the thing that can’t be automated, can’t be replicated, and can’t be compressed into a model weight.

Trust and Judgment.

Human trust. Real relationships. The energy between people.

Judgment to know on a knife’s edge what is right and what is wrong. Not to get too religious, but that divine spark that makes us uniquely human. Unique from animals and unique from each other.

Think about it this way: in a world where every company’s AI can pitch a client flawlessly, write perfect copy, build a flawless financial model, and spin up a product demo in minutes, who does the client actually buy from?

The human they trust. Klarna learned this at scale, in public, with their CEO on the record.

We are completing a 250-year circle. We’re going right back to the pre-industrial village dynamic where your value was entirely based on your reputation, your handshake, and your relationships. Where trust was earned on judgment and through judgment.

I think of it as going back to sticks and stones. Back to man-to-man combat. Back to the fisherman on the dock, who understood all along that the fleet was never the point. No more easy fall backs. No more “if I learn this, I’ll earn this.”

But there is one massive difference between then and now.

We move at light speed.

The sticks and stones of human connection are now powered by a supercomputer exoskeleton. Think Iron Man. Every individual now has access to a suit that gives them the productive output of an entire department. Midjourney is reportedly running at around $500 million in annual revenue with somewhere between 40 and 100 employees, depending on whose estimate you believe. The company has never taken a dollar of outside capital and reportedly spends nothing on marketing. Even at the high end of headcount estimates, that’s roughly $5 million in revenue per employee - a ratio that simply did not exist in the modern economy a decade ago. And Midjourney is not a one-off. There’s a growing class of AI-native companies operating at headcount-to-revenue ratios that would have looked mathematically impossible in 2015.

The question isn’t whether you have the suit, everyone will have the suit. The question is what you do with it and if you have the judgment to know when to put it on and when to take it off.

And what you should do with it is stop trying to out-compute the machine. Use the suit to handle the execution and accelerate your known workflows. Free yourself to spend 100% of your time on the thing that actually creates and holds value: unique human thought and human-to-human connection.

The Counterargument

Let me steelman the opposing view, because it’s the strongest argument against everything I just wrote.

The bear case is that AI agents will eventually develop persistent identities, durable reputations, and long-term relationships at scale, and that the “trust capital” I’m calling a moat will itself be eaten the way every prior moat has been eaten. Agents will remember you, learn your preferences, build track records, recover from failures, and become trusted entities in their own right. It will know you better than you know you.

I think this is the strongest counterargument and it deserves to be acknowledged.

Here’s why I still don’t buy it. Trust between humans isn’t just track record. It’s embodiment, consequence, mutual vulnerability, shared time. When an agent fails you, there’s no human to call, no relationship to repair, no ongoing reputation that suffers in any way that matters to anyone. The agent is replaceable, ephemeral. The human standing behind it is not. As long as economic transactions ultimately involve humans on at least one end, the human-to-human trust layer holds value.

The day that stops being true, this entire essay needs to be rewritten, and so do most other essays about how the economy actually works.

My best proof against this: cryptocurrency. The promise of trust-less transactions. Humans want trust. Put another way, they want someone to call and yell at or to personally verify their identity to approve moving their money for a down payment on a house. The extreme case doesn’t work. (Though I am still waiting for the true smart contract for everything in between.)

The Prediction

Here’s where I think this is going.

The Employee Side:

The people who win the next decade won’t be the ones who mastered the most AI tools. They’ll be the ones who built the deepest trust networks and developed the sharpest judgment. They’ll read more, think harder, understand people better, and leverage AI as pure infrastructure, the way we use electricity today. Nobody brags about being good at electricity. The people who matter are the ones doing something meaningful with the power.

The Employer Side:

Companies are going to stop being defined by the size of their workforce and start being defined by the leverage of their workforce. Look at the Midjourneys, Ramps, and Perplexitys of the world, micro-enterprises where tiny teams of orchestrators serve millions of users, generating massive revenue per head. They automated the middle layer internally, equipped their top people with the best possible tools, and trusted them to drive outcomes, not output. That’s the model. (Sorry junior VCs with LinkedIn Sales Navigator… headcount growth or open job recs are no longer a proxy for the kind of growth that actually matters.)

The employers who win will be the ones who understand that their new moat isn’t their software, their data, or their org chart. Their moat is the trust capital of their best people. Their job is to arm those people, buy them the best suit they can find, and get out of the way. I heard a great line today in an interview with Brian Chesky: spend more time recruiting so you can spend less time managing.

The Compensation Side:

We’ve been trying to map value to compensation for a hundred years and everyone knows it doesn’t work. Annual reviews are theater. Salary bands are political. People game the system in both directions. AI for the first time gives us the telemetry to actually align value creation with economic reward in something close to real time. The companies that figure out this alignment will attract the orchestrators. The ones that don’t will be left managing an army of model feeders wondering where their talent went. Scary enough, this probably means fortunes will be gained and lost faster than ever before.

The Big Picture:

We don’t need the fleet anymore. The 100-year cat-and-mouse game of creating problems and then building departments to solve them is over. The fisherman was right all along.

Take The Suit Off

We can now bring our metaphor back to the fisherman. The businessman wants the fisherman to put on an Iron Man suit. The fisherman has found human contentment in the singular activity of fishing. We now have the tools to explain precisely where their outlooks are completely at odds. And neither one is wrong or right. It’s about knowing which mode is the right one for that situation.

The dangerous mistake of this era is going to be the people who put on the suit and never take it off. They’ll mistake constant orchestration for constant value creation. They’ll have 20 agents running, three Slack channels, two AI meeting summarizers, one personalized newsfeed, and zero quiet hours. They’ll be efficient and irrelevant. The output will be enormous. The judgment will be threadbare.

The deepest competitive advantage of the next decade is going to be the willingness to take the suit off.

Sit on the dock without a goal. Take walks without earbuds. Have dinners where the phones stay in coats. Read books that have nothing to do with your job. Have conversations with people who can’t help you with anything. Be still long enough that judgment can actually form. Stop running away from your thoughts and dive into them.

Trust is built in those moments. Taste is built in those moments. The judgment that the AI cannot replicate, the divine spark that makes you uniquely you, is built in those moments. The orchestrator who never unplugs becomes a routing layer with a name. The orchestrator who knows when to unplug becomes irreplaceable.

That’s the real Iron Man lesson. Tony Stark wasn’t impressive because he had the suit. Plenty of people had suits in that universe. He was impressive because of the man inside it, and because he knew the man mattered more.

Sticks and stones. But at light speed. And critically, at human speed too.

That’s the future of work.

Share

Why productivity and deep focus aren’t the same thing, and how to find the middle ground where most of the work actually gets done

(check out the podcast here)

We live in an attention-starved world that has trained most of us to be addicted to engagement without ever realizing it. And the strangest place I’ve ever noticed it is on a mountain bike.

I’ll come back to the mountain bike. First, a confession.

As a millennial, I’d like to think I have my relationship with technology under control. I don’t. Most people don’t either, and the people who think they do are often the most captured, because they’ve stopped running the diagnostics on themselves. We’re not digital natives in the Gen Z sense, we got our first iPhones in college, and that generational lag gives us a false sense of immunity. It shouldn’t. The feeds came for us too, just a few years later and with less of a fight.

I think of the condition that this tenuous technology relationship has created as ambient restlessness, and I think it’s the most underrated variable in how we spend our days.

The trap with ambient restlessness is that it rarely shows up as a dramatic failure. It shows up as a slight inability to do the thing you sat down to do. You open the book, you read two paragraphs, you reach for your phone. You put the lecture on, you last ninety seconds, you get up to make coffee. Nothing has gone wrong in any measurable sense. You just cannot seem to stay put. And the reason you can’t stay put is that some part of your brain has been trained, over years of optimized feeds and tiny rewards, to expect a constant drip of novelty, and when it doesn’t get one, it goes looking. Of course this isn't constant. There are real moments of deep work scattered through the week, sometimes structured, sometimes discovered. But the steady state is a delicate balancing act, and it eats a good chunk of your daily discipline budget before you've even sat down to do anything hard.

One of my rabbis said something to me years ago that I didn’t fully understand at the time, but it’s been sitting in the back of my head ever since. You can’t focus if you’re not happy, and you can’t be happy if you can’t focus. Those two things are more intertwined than we give them credit for. Every time I’ve gone down an Instagram reel or YouTube short rabbit hole, I don’t just lose time, I lose capacity. I feel worse afterward. Not just tired, specifically depleted in a way that makes the next hard thing harder. Passive consumption is the tax that compounds.

Focus is a Skill, not a Habit

Cal Newport has been writing about the consequences of all this for a long time, and I think he gets a bad rap as the “quit social media” guy. That flattens his actual argument into something much less interesting than what he’s really saying.

What he’s really saying, and what he articulated well in his recent conversation with Arthur Brooks and in the ten-year retrospective he did on Deep Work, is two things.

First, the prescription to quit was never a prescription to throw your phone in the ocean. It was a prescription to put things on an active pause and then decide, deliberately, which ones earn a place back in your life and on what terms. I did exactly this with video recently. For over a month I watched nothing. If something popped up on the screen I closed it. No sports highlights, no nerdy engineering or geography content, no biking content, nothing. What that did for me was two things. It gave me back the quiet in my head that I hadn’t noticed I was missing, and more importantly, when I started letting video back in, I got to choose the terms instead of the algorithm choosing them for me. I’ll probably have to run the same fast again in six months. These things creep.

Second, focus is not a habit you flip on or off. It’s a skill, and like any skill, it has to be trained. You get better at it by doing things that force it out of you. In Cal’s recent podcast, the example is tennis. For me it’s mountain biking. These specific sports are gym for your attention, and in the same way you can grow your VO2 max, you can grow your capacity for sustained focus. But only if you put the reps in.

Once I had that frame, a hidden variable in every endurance activity started to become obvious, and it has nothing to do with heart rate. It’s how much of your brain the activity is actually using, and therefore how much ambient restlessness it absorbs while you’re doing it. I’ve been sorting my audio and my podcasts and my language apps around this variable for years without realizing I was doing it. Once you see it, you can’t unsee it.

The Hierarchy

Here’s the rough spectrum, based on a decade of doing all of these badly and well.

Notice that the two ends of the spectrum are both deep focus, but they're deep focus for opposite reasons. Mountain biking forces the focus on you whether you wanted it or not. Lap swimming hands you an empty room and asks what you'll do with it. The first trains you to lock in under external pressure. The second trains you to direct attention with no scaffolding to lean on. Most of life happens in the middle, but the training that makes the middle work happens at the edges.

Mountain biking sits at 100 percent.
Up or down, technical or flowing, it doesn’t matter. If your brain drifts for three seconds you’re on the ground. There is no bandwidth for audio. There is no room for a stray thought about an email. You are, in the most literal sense, present, because the alternative is a trip to the ER.

A few weeks ago I rode the Sugar Trail, one of the more famous trails in Israel, which starts in Ma’ale Adumim east of Jerusalem and drops down toward the Dead Sea to a town called Almog. You throw your bike in a truck at the bottom and get a ride back up. The ride itself is an Enduro-style descent, rocks, drops, switchbacks, running on tons of adrenaline top to bottom. At the end of it you’re physically and emotionally exhausted. It’s almost like a hangover (without the regret). But the thing I noticed afterward is that during the ride my head was completely quiet. Not managed, not suppressed, just quiet. Every email I was worried about, every small ache, every ambient piece of worry I’d been carrying around all week had been burned off by the sheer cognitive demand of not dying on a bicycle. That’s the training effect. Mountain biking doesn’t just tolerate focus, it forces focus, and when you force focus for long enough in one direction, the restlessness has nowhere to go and it leaves.

This is the Cal Newport point in physical form. An hour of mountain biking is an hour of deep work for your nervous system.

Road biking lives in the 60 to 70 percent range.
You’ve got enough processing power left for music, maybe a podcast on a long climb, but the second traffic gets sketchy or a descent opens up, the audio becomes noise and you tune it out without noticing. It’s a sport that lets you consume content, but only content you don’t need to remember. I usually ride with open-ear headphones so I can hear what’s going on around me, and that in itself tells you something about the bandwidth budget. If you have to leave a channel open for survival, you don’t have the whole brain available for the podcast either.

Running lands somewhere in 30 to 50 percent.
It’s the sweet spot for most spoken-word audio and music. Podcasts work. Audiobooks work if the plot isn’t too demanding. I do some of my best thinking on runs, and also some of my best not-thinking, which might turn out to be the same thing. The specific workflow I’ve landed on: I’ll listen to something for most of a run, and when a thought hits I’ll pause the audio without pausing the run, pull up a voice note, and talk the thought out for thirty seconds while I’m still moving. Half the posts I’ve written in the last six months started as voice notes on a run. The running occupies my body just enough to get out of the way of my brain, and the pausing ritual means the ideas don’t evaporate between the trail and the desk.

I ran my first half-marathon a couple of months ago, down at the Dead Sea. You’re running on the levy across the salt flats, which is about the most featureless running surface imaginable. Nothing to see, a few potholes to dodge, and a lot of time alone with your own effort.

I should be honest about something here, because the cleaner version of this story would be that I deliberately ran the first half in silence as some kind of test, and that’s not what happened. I hadn’t figured any of this out yet. It was my first race, I had no playbook, and I was making decisions in real time about a kind of effort I’d never sustained before. I started the race quiet, not because of a theory, but because I genuinely didn’t know what the right call was and silence felt like the safer default.

The first few kilometers, I was weaving around other runners trying to find my pace, so the silence was probably correct by accident - the focus needed to be on the people in front of me and the line I wanted to take. With the framework I have now, I’d have turned the music on around the third kilometer, once the field thinned out and the only thing I needed to do was settle into a rhythm. Instead I ran the whole first half quiet.

Somewhere around kilometer twelve the fatigue started to set in and I finally reached for a playlist I’d been building in Spotify for the last twelve years. (If you want the playlist, reach out. I’ll send it.)

The pace dropped from around 4:50 per kilometer to 4:35. The last kilometer came in around 4:15. The music wasn’t a distraction from the running, it was the scaffold that kept my brain from negotiating with my legs. Without it, my head would have started drifting to the pain in my quad, the cramp in my stomach, the question of whether I actually needed to finish this thing. With it, I had somewhere to send that drift, and the drift converted into pace.

First half: silent, mostly by accident, fine. Second half: needed the music as the driver, as the thing that absorbed the restlessness and left the running free. The framework came later. The race is where I learned what it was pointing at.

Lap swimming is 0 to 10 percent.
I’ll be honest, I don’t swim, and if you do and I’m wrong about this, let me know. But my understanding of lap swimming is that it’s close to being in a sensory deprivation tank with chlorine in it. You’re looking at the black line on the bottom of the pool. Experienced swimmers know how many strokes it takes them to reach the wall without counting. You could do it blindfolded. Which means lap swimming is the closest thing a Type A millennial is going to get to real meditation without actually having to learn how to meditate. (That’s the thing that tempts me to try it, honestly. It’s also the thing that’s probably going to push me into a triathlon eventually, which people have been asking me about for a while.)

Worth a quick shoutout here to an app called Olo, which uses high-fidelity natural sound to help people who find meditation impossible actually get there. If traditional meditation doesn’t work for you, try it.

The Twist

So far, so intuitive. More physical complexity means less bandwidth for other inputs. Simple inverse relationship. You could stop reading here and you’d have a useful framework.

But there’s a twist, and it’s the part that made me actually understand what ambient restlessness was doing to me.

When I moved to Israel ten years ago, I had a wacky schedule. I was spending part of my time in yeshiva, I was working a full-time job, I had an apartment in Jerusalem and another in Tel Aviv, and I was driving back and forth constantly. This was before the fast train existed between the two cities, so driving was the default. A friend of mine, who I remain grateful to, recorded every lecture given in the yeshiva that year. Every time I got in the car, I’d put one on.

Here’s the strange part. Sitting still at a desk trying to listen to those same recordings, I could not do it. My mind would wander inside of ninety seconds and I’d find myself reaching for my phone or getting up to make a tea. Twenty years ago we might have called this ADD. Now I think it’s just the baseline condition of anyone who’s been marinating in optimized feeds for long enough. Par for the course.

But in the same room, live, with the rabbi in front of me, the exact same content was effortless to absorb. The live experience is a four-dimensional thing. There’s the sound, there’s the physical presence of the teacher, there’s the body language, there’s the energy of the other people in the room, there’s the slight social pressure of being seen listening. All of that additional sensory load takes up the slack in your attention that would otherwise be colonized by restlessness. You don’t have to focus, exactly. You are physically pinned in place by the richness of the experience.

A recording strips all of that away. It’s one channel of audio into an empty room. For a brain already primed to expect constant stimulation, one channel is not enough. The restlessness floods in and takes over.

But in the car? I could absorb an hour-long class without breaking a sweat. I could do Pimsleur language lessons, the ones where the recording asks you to respond out loud and you’re essentially having a conversation with a robot in a language you don’t speak. I could retain things. I could focus. The driving was occupying just enough of me, physically and visually and proprioceptively, to substitute for the four-dimensional richness of a live lecture. It was adding back the sensory load that the recording had stripped out. The fidgety part of my brain had been given a job, and the listening part was finally free to listen.

Sometimes you need two channels.

This isn’t actually a new observation, by the way. Netflix figured it out years ago. They now engineer shows on the assumption that viewers are on their phones for much of the runtime. Scripts get written so you can miss half a scene and still follow the plot. Dialogue repeats key information. Big reveals are telegraphed. They’re not stupid, they’re responding to what people actually do. One screen is not enough anymore for a brain that’s been trained on fifteen-second vertical clips. And the question isn’t whether to fight this, because fighting it is exhausting. The question is how to use it.

The Framework, and the Two Failure Modes

Most productivity writing treats focus as a monolith. A single resource you protect and deploy. Deep work good, distraction bad. And Cal is completely right that for certain kinds of work you need to protect the room. When I’m writing, or working through a complicated analysis, I need silence. Every interruption is a tax, and the tax compounds.

But not every kind of work lives in that mode. Some kinds of focus actually require a baseline of physical occupation to unlock, and fighting that fact is a losing game. The right move isn’t to brute-force yourself into stillness you can’t hold. The right move is to recognize the level of ambient restlessness you’re walking around with, and then deliberately match the activity to the input. Instead of fighting the digitally-integrated life, you use the texture of it, the movement and noise and mild physical demand, as a scaffold for the things you want to absorb.

The two failure modes are symmetric.

Failure mode one: your activity demands more than you have, and you try to consume something anyway. You listen to an emotional audiobook on a technical descent. You play a podcast while debugging something hard. You try to have a real conversation during an interval. The consumption fails, the activity suffers, and on a mountain bike you actually increase your risk of getting hurt. Noise in, nothing out.

Failure mode two: your activity demands almost nothing, and you try to consume something demanding. You sit on the couch and try to absorb a dense lecture. You lie in bed and try to learn a language. Your brain, unoccupied and marinated in years of ambient restlessness, eats itself. The content bounces off. You blame yourself for not having enough discipline, when really you just picked the wrong configuration.

The sweet spot for most learning lives in the middle. Enough physical engagement to quiet the restless part of the mind, not so much that there’s no room left for anything else. Running with a podcast. Driving with Pimsleur. Walking with an audiobook. Easy spin on the road with a long-form interview. And then, periodically, the focus training sessions: the mountain bike ride, the tennis match, the swim, the walk without headphones, the sit-down deep work block with no inputs at all. The activities that don’t just tolerate focus but force it, and leave you calmer on the other side because they’ve burned off the restlessness directly.

Both configurations work because both are deliberate. The failure modes show up when you stop choosing the match and start letting the activity and the input collide by accident.

Knowing Where You Stand

The framework isn’t prescriptive. It doesn’t tell you to do less of anything or more of anything. It just asks you to notice, before you press play, what your hands and legs and eyes are already doing, and to be honest about how much brain that leaves on the table.

Most people are running the wrong audio for the wrong activity, and wondering why nothing sticks. A week of paying attention to the match will tell you more about your own attention than a year of reading about it.

And if it turns out your mountain biking isn’t a sport at all, if it’s coding or cooking or a specific conversation with a specific person, I’d love to hear about it. Send me a note. Tell me what forces your focus, and what absorbs your restlessness, and what the configuration looks like when you get it right.

Subscribe now

In 2007, the subprime mortgage collapse was driven by a simple, toxic combination: overinflated housing prices (driven by low interest rates) and zero-down-payment mortgages.

Because buyers put effectively no money down, when housing prices inevitably dropped, they found themselves completely underwater. They were essentially renters with a massive obligation hanging over their heads.

So, what did they do? The rational economic move. They just walked away. They let the bank take a house they bought for $1M that was now only worth $500,000.
Why pay the mortgage on that?

The exact same dynamic is playing out right now in Seed-stage venture capital.

The Setup: The “Zero-Down” VC Bet

Right now, mega-funds with billions in AUM are writing small seed checks at ridiculous, inflated valuations.

When a $3B fund writes a $3M check into an early round, that is not a core investment. It is purely an option bet. To a fund of that size, a $3M check will not move the needle even on the greatest venture return of all time.

A 1,000x return on a $3M check returns the fund 1x. That’s not just the best case, it’s the case of a lifetime.

They are simply paying the option price for the right to deploy a $50M check a few years down the road. And frankly, it means they lack the conviction to underwrite the Series A right now — something they have absolutely done at inception stage.

This isn’t about capability. When a multi-billion dollar fund writes a $25M check to lead an inception-stage round, that is a core bet. They are putting their money where their mouth is and doing the real work to underwrite a Series A. That’s what these funds are built to do.

But a $3M check from a $3B fund? That is the venture equivalent of a zero-down mortgage.

Founders beware.

The Trap: Going Underwater

You can’t fault founders for taking the money, just like you couldn’t fault homeowners in 2007. It’s a beautiful house. It’s a big, shiny fund brand and historically low dilution.

But here is how the math traps them:

Let’s say a company raises $3M at a $30M or $40M valuation cap. Two years later, they’ve ground it out to $3M in ARR. That’s incredibly solid growth for a seed-stage startup, but it is nowhere near enough to justify raising their next round at an up-round from that previous cap. Especially the way the market is behaving now - it just won’t generate the excitement for a meaningful up-round. Investors are expecting $10M within the first 18 months or it’s a bust. Patience is at an all-time low, driven by fast-moving AI takeover fears.

The company is now underwater. They need another $5M to $10M just to keep the lights on and try to grow into their valuation.

Having invested in 120+ companies in my ~13 year venture career, I am unfortunately speaking from first-hand experience. This is not just pie-in-the-sky theory on how it might play out. We’re seeing this happen right now with 2021 vintage seed deals. “This time it’s different” is a weak excuse. It’s never as different as people think it is.

The Walkaway

What does the mega-fund do? They look at an underwater asset they put effectively zero money down on. They didn’t put enough capital in to really care. Relationships are relationships, but money is money. And in most cases, even if they love the founder, they’d all rather just scrap it and start again with a clean slate. (Usually in whatever market is hot at that moment in time).

So, they walk away. They treat it exactly as what it was: an expired option bet.

What does the founder do? They look at their valuation hurdle. They look at their capitalization stack. They realize they have to climb an impossible mountain of ARR just to raise again, let alone ever see a personal return. They realize they were just an option bet, and now they are stuck.

Their motivation evaporates. So they walk away, too.
They already have the cachet, they’ll be able to raise again. Why tough it out? I honestly don’t blame them.

Subscribe now

The Pushback: “But It’s a Power Law Game!”

When you point this out, the industry pushes back: “It’s an outlier game anyway. The big winners are all that matters.”

But inflating the entry price on an option bet is like shooting the moon on every hand of Hearts before the cards are even dealt. By pumping valuations, you take what should be an outlier outcome and mathematically reduce it to a mere double in the best case. When your massive outcome on a game-changing company is an OK-returner, you’re in trouble. (But hey, if your game is getting rich on fees, then it’s probably fine.)

The Aftermath

There are sell-side opportunities that come out of these situations - acquihires, asset sales - where the VCs get their money back and the founders make a few bucks. (I’ve been involved in a couple). But usually it’s a scraps deal that doesn’t come close to clearing the cap stack. (I’ve been involved in many more of these). The founders end up busting their butts to return a small amount of capital to people who don’t really care about it and don’t meaningfully benefit from it economically. All so they can say they “had an exit.”

The other path is a pivot, which requires an ugly recap that most investors don’t want to touch and most founders would rather skip in favor of folding and starting fresh. Economically, it’s hard to argue with that logic. Emotionally, it might be a different story.

In most outcomes, it’s brutal for founders either way.

In the 2007 housing crisis, the homeowners walked away, and the banks were left holding the bag.

In the Subprime Venture Collapse, both the VCs and the founders end up being the homeowners who walk away. But there is a bank. It's the employees - the ones who took below-market salaries in exchange for equity priced at those inflated valuations. They were sold the dream and ended up with nothing. The VCs lose an option bet. The founders lose a company. The employees lose years of compensation they will never get back without the brand equity of a successful startup. Next time you hear someone say "I am joining a unicorn," give them the warning they deserve.

So... what’s the “Big Short” of 2026?

Share

I thought of this while I was gravel biking… check out the original video here:

Let me explain why…

There has been a fundamental shift in the types of companies that excite me in the last year. For a decade, I was an enterprise SaaS investor. Why? Because it was the best business model ever.

What does that mean? What even makes a great business model?

I have talked many times about how to define business value. And a great business model will drive meaningful business value - sustainably, reliably, and over a long period of time. It’s that simple.

The Mathematics of Value

So let’s define value. To do this we will enlist my favorite tool in all of finance: the discounted cash flow (DCF) analysis. What is this? It’s a quantitative way to understand business value through the lens of long term profitability.

It has 4 key variables which we must understand. I outlined them here years ago in a post, but I will review them here simply as well:

1. Growth (near and mid-term): Big G

2. Growth (long-term): little g

3. Free cash flow margin: FCF%

4. Discount rate (risk): measured by cost of capital - r

Not one, but ALL of these variables must be strong to build a truly valuable business.

Why WAS enterprise SaaS SOOOO good?

1. Big G: It can grow as fast as you want. It’s multi-tenant software that can scale up basically infinitely with no speed limit. And the demand for these products for ~20 years was insatiable.

2. little g: The market showed that it was MUCH bigger than people thought, especially as software became a layer of the economy rather than a sector.

3. FCF%: It costs almost nothing to deliver technically. It costs a bit more to deliver from a customer management perspective. Basically, you invested once big to build it. Then continue to spend a bit to maintain and expand it. Then your only other cost is sales and marketing, which fuels GROWTH rather than sustaining existing customers.

4. r: When the customers you acquired last year are expected to continue paying you next year without meaningful investment, the risk goes WAY down. Your ability to hit your numbers, especially at scale, gets WAY easier. In addition, when done right, enterprise software customers pay a year upfront. This way, even if they leave, you keep the cash. The risk to the business is very low.

Many of these characteristics are interrelated, but when they combine, they form super businesses. The amount of market cap accretion to software companies over the last 20 years is INSANE.

Just look at one example (who represents the industry pretty well):

• Salesforce Market Cap:

• 2006: ~$4Bn

• Dec 2024 (Peak): ~$350Bn

• Feb 2026: ~$175Bn

In addition, total spent on acquisitions by Salesforce has been ~$75Bn - and that is just on the ones where the price was disclosed!

But there were a few underlying fundamental assumptions that drove this:

• High switching costs, low competition, and product worthiness

• Variable costs remaining low

• Economies of scale being an advantage

The removal of these assumptions from the investment thesis recently is the explanation for why SaaS stocks have cratered over the last year. (Salesforce is down 40+% over the last year.)

What Has Changed?

• People can spin up custom apps to do exactly what they want almost instantly, and it’s only getting easier. The expectations around what an app will do for me is also higher than ever. These are not feature requests; there is a fundamental shift in the architecture of digital tools.

• Infusing your app with AI-powered tools to try to keep up is currently a losing battle. Even if you assume AI token costs will fall sharply, they’re killing software margins for now. There has been a more fundamental shift at the architecture level that is very difficult to keep up with using legacy teams and products.

• Economies of scale in SaaS is a mega power... or at least it was. The brand recognition. The compounding revenue. The contract terms. The contract scale. The cost negotiation power. The reference-able customers. Being a scaled SaaS company was like owning a bottomless gold mine.

So if we apply these changes to our DCF variables:

1. Big G: Churn baby churn. Companies are increasingly favoring bespoke solutions. The ability to transition and do the data mastering to make it seamless is easier than ever. Competition is fierce and pricing leverage is gone. It’s possible to keep your customers, but hard to grab new ones.

2. little g: This is all about future sentiment, and it is grim right now for enterprise software. The market doesn’t know where things will end up, but they’re confident that large and mid-size SaaS players are not in good shape to outpace GDP growth meaningfully for the next 10 years.

3. FCF%: This is probably the most interesting and most important. COGS (variable costs) have gone up. Less gross margin to play with from the top down means less margin for error. CAC is also out of control. There is just so much competition in this gold mine after 20 years that the investment required to get new customers has gotten out of hand. On a contribution margin basis, these companies are on thin ice.

4. r: Discount rates are super high because the foundation is being SHOOK. With churn questions, cost questions, and broader market dynamic questions, investors (justifiably) do not trust the ability of these mega software companies to adapt. (Contract terms are generally still good, but there’s a big push towards usage-based pricing - and billing - and away from standard paid-upfront seat-based pricing).

What does this come down to? OPERATIONAL INERTIA.

These companies have been built with thousands of employees with a culture underpinned by 20+ years of enterprise SaaS business model dominance. I often say it’s the best legal business model ever. The result: companies (and their shareholders) got both greedy and lazy. They dug their heels in to get the most out of these underlying assumptions just in time for them to be ripped out from under them.

Enter the New Normal

You have to build a good business model. You need real differentiation. You need to find fundamental revenue resilience.
The new normal sounds an awful lot like the normal of the last 200 years of business.
Companies with too many employees, static processes, and whose economies of scale are really just corporate bloat are going to be in trouble. You just can’t hide much anymore. The emperor is wearing a speedo.

The potential savior: Some of these SaaS companies are run by some of the greatest entrepreneurs ever to walk the earth. I expect we will see several of them pull more rabbits out of more hats.

But for me, as an early-stage investor with a small fund, I am excited about founders who can build strong, resilient businesses. Get to $10M of revenue and do it in a sustainable way, because there are simply too many questions beyond that horizon to possibly answer. It’s harder than ever to know what business or markets will support massive scale, so the founder comes back into clear focus as the sole driver of exponential enterprise value growth by navigating an extremely dynamic economic landscape.

Why legacy industries are the new greenfield opportunities

In the new game, going after legacy industries has become A LOT more attractive. Here is the math:

• Margin Flipping: Gross margin is getting crushed in the pure software world. Meanwhile, gross margins in legacy industries are going up, driven technology automation. This means gross margins in what were considered low-margin industries can be higher than ever.

• Opex Corrections: Opex in software has gotten lazily high. An AI-native operation today has the opportunity to scale with WAY fewer employees. Enter AI-powered super employees whose former 3 hours of meetings per day have been collapsed into 15 minutes of AI conversations. The quality of output is better and the time it takes to produce quality starts approaching zero.

• CAC Opportunities: CAC in some of these legacy industries is a greenfield opportunity. There simply is not competition and noise (yet). A few conversations, a killer product, lower cost, and you’re in. The software world is extremely crowded and the CAC advantages of the 2010s are long gone.

I think it will be very hard for legacy tech companies to fire 90% of their workforce, but a complete rethinking of opex is what is required right now to maintain a competitive edge. This opens the door for young, small, under-resourced companies to make real waves, and do so quickly. The advantages of the agile startup are stronger than ever.

Welcome to Business Model Logic

Switch the math around. SaaS relied on high gross margin and it resulted in high opex. Now the tables have turned: gross margin is crashing while opex stays higher than ever.

New AI-powered companies can have higher gross margins than ever, while the ability to sell an innovative product at a lower price has a renewed openness that we haven’t seen before on a global level. So as always, the more things look different, when we abstract it enough, we realize that we’re actually on the same curve as we have been forever.

It’s time to get back to the fundamentals for a new era of technology-powered businesses. I am psyched.

I have changed the name of this newsletter from SaaS Engineering to Business Model Logic. It’s not because SaaS markets are down, it’s because I’ve been thinking more broadly about business models for many years. The SaaS Engineering idea started in 2014 when my boss at NEA, Harry Weller, came to me and said, “you really like finance, and you’re an engineer. Everything is going SaaS, and I want you to learn this business model inside and out.” Thus the SaaS Engineer, Alex Oppenheimer, was spawned (and my career has been guided by this conversation ever since). The work I have done since then gave me a sharp view on WHY the SaaS business model can be so powerful, and as you can see above, why I can explain its current downfall. It was never about the jargon or the hype, it was always about the calculus. I am extremely excited when new technologies open new markets, but most of all, create new business models. Looking forward to sharing more here soon.

Subscribe now

P.S. check out the original DCF / Direction Valuation Modeling post here:

In the venture world, we live with a scary reality: you often don’t know if the emperor has no clothes until you are very far down the road.

We know if a company can be capitalized. We saw plenty of that in the zero-interest rate era. But we don’t always know if that company is actually accruing value. Just because a VC wrote a check doesn’t mean the machine works.

I’m an engineer by training. And when I look at how most founders (and many investors) approach their business, I see a lack of rigor. They build complex three-statement financial models—income statements, balance sheets, cash flows—before they’ve even proven the basic physics of their business.

If you are early stage, your financial model is fiction.

What you need instead is a Business Model.

Here is how to tell the difference, and how to fix yours before you run out of runway.

The CAD Model of Business

In mechanical engineering, you never build a physical object without first building a CAD (Computer-Aided Design) model.

Why? Because the real world is expensive. Materials cost money. Machining takes time. If two parts bang into each other in the real world, you have a disaster. If they bang into each other in a CAD model, you just hit “undo.”

Your business model is the digital representation of the machine you are trying to build.

A professor of mine, Marsh McCall, used to look in the mirror while writing papers and say, “Marshall, your outline is weak, and therefore you are weak.”

If your business model—your outline—is weak, your business is weak. You can’t out-hustle bad math. If you are selling a dollar for 75 cents (looking at you, WeWork), no amount of “community-adjusted EBITDA” or charismatic leadership will save you eventually.

The One Column Model

Founders often avoid modeling because they think they need a CFO to build a massive spreadsheet. You don’t. You need what I call the One Column Model.

Open Excel. Use one column. You only need three ingredients:

1. Independent Variables: The inputs you control (Pricing, Ad Spend, Sales Commissions).

2. Dependent Variables: The outputs (Revenue, Churn, Conversion Rate).

3. The Mechanics: The math that ties them together.

I recently sat down with a founder running a brokerage business. We opened a spreadsheet. We listed the deal size ($30k), the commission (10%), and the payout to sales staff. In about 11 rows—maybe 10 minutes of work—we realized the strategy was totally under-optimized. The volume required to just pay his sales staff a decent wage was impossible at his current price point. So he adjusted the initial target market, and began executing on that.

That 10-minute exercise saved him two years of grinding on a strategy that was mathematically unfeasible.

Financial Models vs. Business Models

This is where people get confused.

A Financial Model is accounting. It relies on GAAP standards (Generally Accepted Accounting Principles). Concepts like EBITDA are useful, but they are innovations designed to standardize reporting, not necessarily to tell you the economic truth.

A Business Model is about mechanics. It answers one question: What is the Unit and how does it work?

If you tell me your “unit economics” are good, my first question is: “What is your unit?”

• Is it a customer?

• Is it a contract?

• Is it a node?

The unit is the precise point where the Product Side (Revenue) meets the Cost Side (CAC & COGS).

Take Uber. It’s a complex marketplace. What is the unit? It isn’t just a driver, and it isn’t just a rider. You have to find the least common denominator over a period of time.

• Let’s say a driver does 300 rides a month.

• Let’s say a rider does 30 rides a month.

• The Unit is 1 Driver + 10 Riders = 300 transactions/month.

You have to calculate the cost to acquire 1 driver and 10 riders, and compare that against the margin on 300 rides. That is your unit economics. If that equation doesn’t result in a positive contribution margin quickly, you don’t have a business; you have a subsidy.

The Hierarchy of Revenue

Once you understand your unit, you have to look at the quality of the fuel you’re putting in the engine. Not all revenue is created equal.

When I was at NEA, Harry Weller taught me that Enterprise SaaS is the greatest business model in history for one reason: It accumulates and it pays upfront.

You pay to acquire a customer once. They pay you $100. Next year, you don’t have to pay to acquire them again, but they pay you another $100. Then you acquire a new customer. Now you have $200.

This is distinct from a “consumable” business (like hardware or e-commerce) where you have to re-sell the customer constantly.

When analyzing your revenue quality, score it on three vectors:

1. Reliability: Will it happen again automatically? (Subscription > Single Purchase).

2. Profitability: What is the contribution margin? (Software > Hardware).

3. Velocity: How fast does the cash hit the bank? (Unearned Revenue > Accounts Receivable).

You want high velocity, high reliability and fundamentally strong unit economics. If you have low velocity (18-month sales cycles), low reliability (high churn) and low gross margin (high variable costs), you are in the danger zone.

Financial Data is Blood, Not Oil

People like to say “Data is the new oil.” I disagree.

Data is blood.

When you go to the doctor for a blood test, they stick a needle in and take a sample. That sample tells them how the organs are functioning. Is the liver failing? Is the heart pumping? Is there an infection?

Financial data is the blood of your organization. It is the diagnostic tool that tells you if the organism is healthy.

It is the leading and lagging indicators. The independent and dependent variables. All in one place.

If you build the right business model (getting the mechanics right) and you monitor the blood properly, you shouldn’t have to make hard decisions.

I tell founders: “Great analysis results in non-decisions.”

If the data is clear, the choice should be obvious. “Oh, we lose money on customers under $10k ACV? Great, we stop selling to them.” That’s not a decision; that’s a realization.

Stop guessing. Do the math. If your outline is weak, you are weak.

Subscribe now

Takeaways

• Raising venture capital is challenging despite abundant funds.

• Understanding the sales process is crucial for founders.

• Different VC conversations require tailored approaches.

• Emotional dynamics play a role in early-stage decisions.

• Aligning with the right investors is key to success.

• Junior VCs can be strong advocates for founders.

• Market idiosyncrasies add complexity to fundraising.

• Strategic planning is essential for effective fundraising.

• Founders should focus on long-term investor relationships.

• Navigating VC layers requires understanding their roles.

00:00 Navigating the Venture Capital Landscape
04:29 Types of Venture Capitalists and Their Decision-Making
10:31 Strategies for Engaging with Different VC Types
18:53 The Importance of Timing and Economic Alignment
26:23 Building Relationships with Junior VCs
32:20 Final Thoughts on Fundraising and Investor Relationships

Subscribe and learn more at verytrue.fm and on spotify.

You can subscribe and hear all the episodes here:

https://verytrue.fm/

Subscribe now

After listening to the entire Acquired podcast on Rolex, it really got me thinking…

Tech bros love both, they are way cooler to talk about than actually do/have, and people who are involved in them have way less money than people think, but is there more?

Perhaps unsurprisingly as an early-stage investor, I have been drawn towards the upstarts and microbrands. But there is a lot to learn from the big names. The question always remains - which is better value? Or is some mix the right answer?

Here are a handful of things that struck me as oddly similar:

• Actual vs. perceived utility

  • They are both solving smaller problems than people think

    • Watches tell time worse than a variety of other devices at our disposal

    • Venture as an asset class has proven incredibly inefficient

  • The signal they send is stronger than the tactical purpose they serve

• There is tremendous opportunity in either 1) being creative and innovative or 2) doing it better and for longer than anyone else

  • There is very little opportunity in between

• You can be meaningfully involved in both without really knowing what you’re talking about

  • Having a bunch of money gets you there fast, but it doesn’t create staying power

  • It’s easy to look like you know what you’re doing, but there may be no rhyme or reason behind it… and it may never matter

• Owning a bunch of expensive stuff doesn’t mean you know anything about how they’re made or why it matters

• Knowing how to build great stuff doesn’t mean you know what they next wave of popularity will bring

• People make more money on momentum than on fundamentals

  • A big winner can become a big loser faster than anyone expects

• The media around the asset is often more valuable than the asset itself

  • Both industries have recently (last 10 years) attracted a type of deal-maker that has bombarded the industry and pushed away many of the OG aficionados

• People say it’s all about access, but access can be bought

• The best in the business just go for what they like instinctually and others follow on

  • Only years down the road do the masses see what the visionaries saw early (before it was popular or obvious)

• Real quality might be hard to describe concretely but you know it when you see it

  • This frustrates many who are momentum players

• Low interest rates blow out both markets

• Impressing other dudes is clearly the goal of both

There is a lot to learn from thinking about these similar dynamics and even drawing parallels between specific brands, i.e., Is Rolex more like Apple or Sequoia? Is Swatch Group more like Microsoft or Google?

And more importantly, is there space in the industry for another titan to be built? Or are all the major players in place, and any upstart that gains real momentum is either a competitive target or an acquisition target?

Does more value accrue with simplicity or novelty? Through longevity or growth? Does innovation really matter or does flash win the day?

One of the reasons I have gone down this rabbit hole (besides that it’s fun) is that there is a lot to learn from businesses outside the technology world. In the luxury business, micro economic principles break down and reveal some deep psychological realities. I also believe that ultimately every business is a consumer business - there is a human decision maker and there is a human user, directly or indirectly.

If you too are down the watch rabbit hole, feel free to reach out!

Subscription businesses represent a unique opportunity because of the expectation of retention - this, along with high gross margins and scalability, is why these SaaS valuations are so high relative to size and profitability. There is a general assumption that customers will stay on board and continue to generate revenue. Because the biggest expense - customer acquisition - only happens once, long term profitability looks great. There is some chance that they don’t renew, and we call that churn. There are two types of churn:

1. Involuntary churn

   1. Customer goes out of business

2. Voluntary churn

   1. Customer found a better solution

   2. Customer doesn’t need the tool any more

But the term churn is a bit deceiving. In practice, companies don’t rage quit from platforms like employees sometimes do, rather they choose not to renew. So what is the difference? Renewal timing. The 1-year contract (paid upfront as it should be) has a big impact here because it means that customers can only choose not to renew once a year. Even if customers decide to not renew a few months in advance of their renewal date, their cash is in the bank and their access to the platform remains, so they still generate revenue.

Investors and founders alike cleaved to this churn metric for two main reasons:

1. It is easy to calculate if you have a simple ARR waterfall model.

2. It is useful in the often-misunderstood customer LTV calculation.

The problem is that a simple calculation of Churn Rate (Churned ARR / Beginning ARR) does not take into consideration what portion of that Beginning ARR is even eligible to churn. In a fast growing business with annual contracts, a company may go from adding 5 customers per month to 40 in a 1 year span, but only those original 5 are eligible to churn after 12 months. As a result, a simple churn rate calculation is very deceiving. For example…

What looks like a 16% annualized churn rate is really an 80% non-renewal rate, which is reason for alarm. And this is just the tip of the iceberg.

There’s another part of the equation: what if we keep a customer and they start paying us more or less than they were before?

This is where upsell/expansion and downsell/contraction enter the picture, and along with them, plenty of questions of how to incorporate them into the projection and valuation process.

For several years, there was a trend to use a term called “net churn,” which still makes me gag a little bit to say. I am happy to say that investors and operators have come to their senses and stopped using this term. The reason this oversimplified, conglomerated franken-metric was used:

1. Because it further simplifies things

2. It can be even more mis-used in the often misunderstood customer LTV calculation

3. It hides how bad the actual churn is

Oh, but there are so many problems…

1. Anyone looking at these numbers should be able to take a churn rate, say 15% and subtract a net upsell number, say 6%, and get to a net churn of 9%. Abstracting away this simple arithmetic is fishy at best.

2. More importantly though, it doesn’t tell us much about the business or customer behavior because it blends together so many different time periods and customer behaviors into a single number. Building models and doing analysis is the art and science of taking intelligent averages to improve accuracy. If you’ve already blended a bunch of stuff together and now try to analyze it by slicing it up again, you are adding precision while decreasing accuracy.

But let’s get back to churn, and why looking at it from a retention perspective makes much more sense. In theory, they are just 1 minus the other: 10% churn = 90% retention. But when you’re dealing with anything besides monthly contracts, the customer behaviors diverge much more than the numbers allude to. As mentioned above, ultimately a customer makes the decision to not renew, and the timing is critical. Without knowing customer contract length, it’s very hard to predict renewals. You really need to know which customers are up for renewal, and establish a pattern of renewals in order to drive a projection.

Subscribe now

Enter the Cohort Analysis

Up until now, we’ve just been doing calculations with this data:

There is nothing wrong with this data, but it leaves so much to be desired. If you want to really understand customer behavior and the value of a company’s revenue, you need to go into the cohort data, which involves digging into these sort of eyecharts:

And this still isn’t even enough! If a company has multiple contract length options, then you could mark each customer as “monthly” or “annual,” and based on the contract start date (i.e., which cohort they belong to), you could assume a renewal date. Of course, this ignores the details of free extensions given or upgraded contracts that extend renewal dates. That level of data lives either in a CRM, billing tool, or subscription management tool. If every contract is annual, this is simpler to understand, but you still need to look at contracts to figure out if an upsell/expansion resets the clock on renewals or not.

It is important to note that there are several ways to put customers into cohorts:

1. When they became customers

2. When they are up for renewal

3. Size of contract

4. Billing cycle

5. Contract length

For the purposes of understanding Net Revenue Retention, we’ll just focus on cohorting customers by when they became customers. The reason for doing this is that you can map it to sales cycle and therefore calculate an average CAC for each cohort. This allows you to do powerful yet simple analysis to answer important questions like: How long do we need to retain these customers for this customer cohort to be unit profitable? And then look historically to see how long it actually took to reach cohort profitability. Trying to understand CAC payback without this analysis is a gross oversimplification which may give you a rough relative efficiency analysis, but not an absolute efficiency metric.

Before talking about Net Revenue Retention, let’s first start with Customer Retention. Just to clarify, a “non-renewal” or “churn” only occurs when you officially lose a customer. If their MRR/ARR changes up or down, it is called upsell/expansion or downsell/contraction. Why does losing customers matter so much? Because acquiring them is expensive and is the core operation of the entire GTM team for a fast-growing company. Once a company is not expected to grow as fast (i.e., it has reached scale), then customer success, retention and upsell become the priority.

When cohorting using customer acquisition date, it’s critical to remember that cohort size can only go down. You start with a cohort of 10 customers that were acquired in a given month, and it only goes down from there. You may ask “what about Reactivations?” Ok, so sure, a customer can leave and come back, but a) that almost certainly costs the GTM team meaningful time and money, so maybe it should be looked at as a new customer from a CAC perspective, and b) cohort size still can never be more than the original 10: it can go from 10 to 9 to 10, but only if that same customer leaves and returns (within a reasonable amount of time).

As a result, the customer retention by cohort chart looks like this:

This is not so easy to read and really doesn’t tell us much, but it’s a good check that the analysis is working properly: all of the lines go down and to the right or stay flat.

On a percentage of original cohort customers retention basis, the chart looks like this:

As you can see, it never goes above that 100%. But this still doesn’t tell us much besides how fast customers are dropping off from each cohort. Naturally, the flatter the better.

Now let’s dig into the revenue numbers. First, we’ll just look at a single cohort:

It’s interesting! But it still doesn’t tell us much about the customer behavior en-masse to the point where we can use it to predict future ARR fluctuations. We see here that the cohort started at $4,080 MRR, expanded up $5,200, and then contracted back down to $4,755.

The next step is to layer on all the cohorts…

Now we’re looking at modern art spaghetti. Also not so helpful.

The next step is to normalize this data. Instead of starting at the initial MRR, each line will start at 100%.

Maybe even less helpful, now it looks like a bad hair day.

Next step: take an average. The only problem with an average here is that the newer cohorts are likely bigger and more indicative of future behavior, so we’ll also take a weighted average by cohort size to try to correct for this a bit.

Now we’re talking! We’re looking at a nice set of retention curves here which give us a sense of the customer behavior over time. Notice that in month 12, we see a drop off. This is an indication of annual contract non-renewal, which makes sense.

But now what? What is our Net Revenue Retention number?

Often NRR gets presented simply as a single number, and it’s really hard to know what that number represents. Is it at the one year mark? Is it before or after the first opportunity for annual contracts to renew? Is it an average or a weighted average? Were any other sorts of cohorts eliminated from the analysis (i.e., small customers)? What percentage of revenue do those represent?

Net Revenue Retention really needs to be communicated as a curve.

Though it can also be communicated clearly as “weighted average net one-year revenue retention is 173% across all cohorts.”

There are also several other questions that need to be answered to better understand the quality and pattern of behavior of the customer base, which is the reason we’re quantifying net retention to begin with.

• Where is upsell/expansion coming from? More seats? New products? Overages or Add-ons that are re-occurring?

  • These all have big implications on the quality and value of revenue in the future.

• Where is downsell/contraction coming from? Did our salespeople oversell the product? Are our customers doing layoffs? Are our premium features/products sticking? Or is it just because we converted some $100/month customers to $1,000/year customers?

  • Sometimes downsell is good and sometimes it is reason for alarm even more so than churn.

This is all to say that it’s critical to remember WHY we are performing any analysis:

I am including a link to the spreadsheet with the above analysis and charts here: https://docs.google.com/spreadsheets/d/1qpL162TA0eZcYNXUXQrbi4wzQwNuSlWWpKNrTg-FhC4/edit?usp=sharing

Note that this is only a historical analysis. For investors, this may be enough to diagnose a business, but for operators, the key is to take these conclusions and use them to drive assumptions in a forward-looking model. That is where the real art comes in and it’s very situation-specific.

One hack method I have seen people use in the absence of cohort data is:

Numerator: full year of upsells + downsells + churns

Denominator: Beginning ARR + full year of New ARR

Calculated another way (with even less data):

Numerator: Ending ARR

Denominator: Beginning ARR + full year of New ARR (minus 1)

While I understand why people use these calculation in the absence of cohort data, they don’t capture the aging of cohorts, and instead treat MRR added to early cohorts the same as MRR added to larger, more recent cohorts which will grossly underestimate expected cohort growth. As cohorts get larger and the company’s growth rate changes, this number will be decreasingly indicative of future revenue retention behavior. Equal parts imprecise and inaccurate.

Just know that if you are not using the cohort data, then it isn’t Net Revenue Retention, and it may be misleading.

This matters for both investors doing analysis to assess valuation and future prospects and even more so for founders who are trying to optimize growth and make useful projections.

One final note… in some cases, high NRR isn’t a good thing. Better to just start at full contract size and keep it forever than to start with a tiny contract and need to constantly upsell it. So if you’re trying to measure and optimize your NRR, maybe first ask if it’s a good indicator for your specific business case.

And always get back to the core reason for any of this analysis: to understand and gain confidence in projections.

Subscribe now

"Annualized Recurring Run-Rate."

Huh? Run rate of what? Which part is recurring?

While there isn't any GAAP definition of ARR, I think the generally accepted idea matters: it has something to do with revenue, annualizing and recurring.

The ARR definition most people agree on in the software world is: Annualized Recurring Revenue.

The idea is that you take earned revenue that is recurring (i.e. subscription contracts) and then annualize it to better understand the current velocity of the business. 

Toast uses a very different definition of ARR as one of their three core KPIs. They explain it on page 72 of their 10-K. The other two are: Gross Payment Volume and Net Retention Rate. These two KPIs both make a lot of sense for this business and as a result, the explanations for both of them are concise and clear, unlike their definition of “ARR,” which I assume you didn’t read and know that you didn’t understand (because it doesn’t make sense).

The biggest issue I have with their definition is what they are doing with the Payments Revenue Component. They explain how they calculate it as "four times the trailing three month cumulative payments component." They then add this to the subscription component to get to the total "ARR."

Here's where we diverge:

What's the difference?

The subscription contract.

If the payments revenue was recurring, then it would be adding apples-to-apples, but it does not recur, it re-occurs. These few letters make the world of a difference.

The reason this matters is that the VALUE of contractually recurring revenue ("subscription services") is greater than the value of re-occurring revenue ("financial technology solutions").

In this case, it is for 2 reasons:

1. The subscription contract means that it's more reliable and therefore more valuable in the future (it's locked in, higher NPV).

2. The associated gross margin: 67% vs 22% (subscription revenue hits the bottom line much stronger). 

Contrast this with Shopify, who does not talk about ARR at all and instead simply refers to: "subscription solutions," which is 81% gross margin and "merchant solutions," which is 39% gross margin. And adds them together to get Total Revenue. From there, you can annualize as you please. 

The major takeaway here is that it is actually OK to annualize anything. But it's not OK to call that anything ARR unless it's recurring because that is what it implies. 

Toast’s legal caveats in the 10-K are accurate, but saying that "ARR is not a forecast of future revenue" defeats the purpose of ARR as a metric altogether. The reason we use ARR is that it implies that it WILL be there reliably over the next 12 months and therefore is a forecast. (See reference to velocity above, which by definition is a forecast for distance to be covered.)

This underpins the point of the subscription model: You will be able to keep what you have, in other words: it recurs. 

So to the IR team at Toast: just do what Shopify does and call it like it is...
Simply show quarterly revenue split out by these two types, let people multiply it by 4 if they want to annualize it. Playing games with metrics usually means you’re trying to hide something or you don’t know what you’re talking about. 

OR... call it ARRR and make the whole 10-K pirate-themed.

Subscribe now

Note: Shopify trades at 9.0x 2024 revenue and Toast trades at 2.8x 2024 revenue.

Toast 10-K: https://www.sec.gov/ixviewer/ix.html?doc=/Archives/edgar/data/1650164/000165016424000084/tost-20231231.htm 

Full definition of Toast “ARR” here:

We monitor Annualized Recurring Run-Rate as a key operational measure of the scale of our subscription and payment processing services for both new and existing customers. To calculate this metric, we first calculate recurring run-rate on a monthly basis. Monthly Recurring Run-Rate, or MRR, is measured on the final day of each month as the sum of (i) our monthly billings of subscription services fees, which we refer to as the subscription component of MRR, and (ii) our in-month adjusted payments services fees, exclusive of estimated transaction-based costs, which we refer to as the payments component of MRR. MRR does not include fees derived from Toast Capital or related costs. MRR is also not burdened by the impact of SaaS credits offered. The MRR calculation includes all locations on the Toast platform and locations on legacy solutions, which have a negligible impact on ARR.

ARR is determined by taking the sum of (i) twelve times the subscription component of MRR and (ii) four times the trailing-three-month cumulative payments component of MRR. We believe this approach provides an indication of our scale, while also controlling for short-term fluctuations in payments volume. Our ARR may decline or fluctuate as a result of a number of factors, including customers’ satisfaction with our platform, pricing, competitive offerings, economic conditions, or overall changes in our customers’ and their guests’ spending levels. ARR is an operational measure, does not reflect our revenue or gross profit determined in accordance with U.S. Generally Accepted Accounting Principles, or GAAP, and should be viewed independently of, and not combined with or substituted for, our revenue, gross profit, and other financial information determined in accordance with GAAP. Further, ARR is not a forecast of future revenue and investors should not place undue reliance on ARR as an indicator of our future or expected results.

The typical LTV calculation usually looks something like this:

This method is based on the annuity formula for financial instruments. In very few circumstances does this formula yield anything remotely useful - both for investors trying to evaluate a business and more importantly for the operators themselves. Companies and their customers typically do not behave anything like engineered financial instruments. As a result, there is so much lacking in this LTV formula that I don’t even know where to start, but the key problems are 1) at the early stages (and even into the mid stages) the data simply is not good enough to reliably establish these metrics, and 2) most people have no idea where it comes from or how it works and simply apply it blindly.

Which leads to another, more informative method:

This method is actually the same… but in the previous over-simplified case you are assuming that n = infinity. (Assuming anything equals infinity in a practical business setting is obviously questionable.) Using this sum formula with the requisite understanding of customer size and behavior can yield much more accurate results, but what it takes to really understand customer behavior is up for debate. In addition, you can make a simple assumption like 5 years for n, but why choose 5 over 3 or 10 years? If the company has only been around for 2 years, it’s very difficult to predict and quantify the customer lifetime.

That is why I came up with a third option… the customer half life formula:

Here again I use the same formula, but instead of making an assumption for n, I use the half life formula to understand the probability of maintaining a customer based on historical customer behavior. The theory behind this is what once the probability of a customer churn goes above 50%, we can no longer count on retaining that customer, so we end the customer lifetime there.

This results are finite numbers for n based on measurable annual churn numbers:

• 5% churn = 13.5 year half life

• 10% churn = 6.6 year half life

• 20% churn = 3.1 year half life

But we run into the same problem: in the early days of a company without much data on customer behavior over time, it is very difficult to nail this churn number which now has even more impact on the LTV calculation. If you sell monthly contracts and have a large number of customers with industry-standard behavior, this can work (i.e., consumer cybersecurity), but if you sell annual or multi-year enterprise contracts (i.e., accounting software), then churn means something entirely different and it takes years to understand this dynamic. In either case, customers are retained rather than churned - this is the essence of a subscription. The impact of upsells, usage-based billing and multi-year annual contracts throw another wrench into the calculation. Whereas a company like Spotify or Netflix with a simple monthly subscription mechanism and tons of data can probably use these simplified calculations, as soon as the math gets a bit more complicated, accuracy suffers immensely.

So we’re back to square one. And this is just the top half of the famous LTV/CAC ratio that people discuss and calculate constantly. The good news is that CAC is easier to measure, but it can be harder to calculate and predict as you move through the phases of product adoption.

For later stage companies with more data, it benefits both evaluating investors and operators to try to improve precision and thereby improve accuracy of these metrics. The result: powerful diagnostics which can help assess the viability of the business longer term as well as improve operations and resource allocation in the short term. Defining the right metrics and capturing them from an early stage make this exercise much easier down the road.

But for early stage, we are still at a loss. Trying to do calculations with just a handful of customers is not useful, but understanding the concepts behind logarithmic decay and net present value is helpful in defining metrics to drive business success in the future.

So how do we apply these very mathematical concepts to nuanced and nebulous situations to understand business model opportunities at the early stage?

The answer is to think about what we can know at the early stage, then see if any of those data points map back to the fundamentals we were trying to calculate above.

I have identified three key product-market characteristics that drive lifetime value even before launching a product:

Average Contract Value (ACV)

• Average contract values need to be high enough to create interesting opportunities

  • High ACVs mean more margin to play with when it comes to fixed and variable costs. Mistakes and experiments are more easily forgiven.

  • With a lot of outstanding questions, knowing that there is pricing leverage due to market demand and value de-risks the entire situation.

• Pricing is also an engagement method: for most enterprise and consumer buyers, the more they’re willing to spend, the more they care, the more they’re willing to invest in making the product work for them, resulting in longer customer lifetimes and higher LTVs.

Adoption Pattern / GTM Strategies

• With the proliferation of digital marketing and modern sales tools, there are very few industries where a new company can sell efficiently at scale. The market for eyeballs and attention is priced to perfection, so any sector that is not yet efficient will quickly head in that direction once the word gets out that there is a new opportunity.

  • Figuring out how to create advantages in customer acquisition is critical.

  • For larger companies, this usually means branding and partnerships.

• In certain parts of the enterprise software stack, there are well-trodden sales motions, but it’s almost always a race to the bottom before long and one competitive player with a loose budget can drown everyone in the market.

  • The brute force method can be an effective way to get started, (i.e., hiring well-trained sales reps who can get the ball rolling right away), but it is likely going to become a challenge without access to zero-interest-rate levels of venture dollars.

• The solution:

  • Markets with group-think buying behaviors

    • Certain verticals tend to hire the same consultants, go to the same conferences, buy the same tools, etc. If you’re the first software player in that space, then there’s a good chance you could be the great beneficiary of this buying behavior if you can snag an influencing customer early.

  • Virality

    • Certain product-market combinations encourage initial adopters to bring along their friends. The product and the market must be cleverly engineered to make this happen, and it is possible for both enterprise and consumer.

  • Network Effects

    • This may be the single most powerful adoption dynamic in venture capital history. This is how Facebook got where it is, but it’s also how Bill.com has developed so impressively. Finding a market where you can build a product that can unlock network effects is not easy, but when it happens it is pure magic.

Retention Dynamics

• The longer a company retains its customers, the more money it makes. Even though cash now is worth more than cash later, it still makes a big difference to maintain customers for 10 years instead of 1 year.

  • Finding markets and products where multi-year contract and enterprise buying behavior are evident is great.

  • Pairing that with variable customer adoption behavior is a super power.

• There are two distinct approaches to retention, which are often intermingled: economic (i.e., revenue) and cash.

  • Economic retention is what is in the contract: is it monthly, annual, multi-year? Is there auto-renew?

  • Cash is king. An annual contract that gets billed monthly is effectively a monthly contract. Getting the cash upfront makes a big difference when it comes to retention.

  • Cost of capital is incredibly high for early stage companies, getting free net working capital loans from your customers in the form of upfront cash payments is a huge unlock for growth.

• Upside alignment with customers is huge opportunity.

  • No company has done this better than Shopify:

    • Lock in a subscription tool that enables more sales for your customers

    • Take a piece of the transaction value

  • This results in a highly predictable revenue stream with a more volatile, but customer-aligned upside revenue stream.

• Executing the land-and-expand well is incredible, but it’s often much harder than it seems.

This exercise takes me back to engineering classes in college which allowed us to have an “equation sheet” for exams. They usually looked something like this, as we tried to cram as much as possible onto a notecard to help us during the exam. The trick was always that making the equation sheet was more useful than having it during the exam. Understanding the meaning behind the variables and how to apply the equations is more important than the equations themselves. Our professors knew this and preferred to test our ability to apply our learnings over our ability to memorize formulas.

This is all to say it is important to get behind the simplified formulas and understand the variables that drive them. Despite not being able to calculate them with precision, we can still understand an opportunity pretty accurately by looking at what we do know. The key variables in the LTV formula above are:

• ARR = size of a customer

• n = lifetime of customer

• churn = retention dynamics

Pairing that with Customer Acquisition dynamics, which are also characteristics of a market, we can get a fairly accurate picture of the unit economics of an opportunity without a single operating metric in hand. The math at this stage is simple and it’s worth everyone’s time to run through it to better understand which variables matter, how they interact, which are the most sensitive, and ultimate what to be on the lookout for assessing risk and upside along the way towards product-market fit. We can also figure out which operating metrics to track so that founders and their investors will understand if what they are doing is actually working and how well.

An example: we have a certain portfolio company that many would call “consumer fintech,” which as a category has a bad reputation for retention, distribution and transaction sizes - overall bad unit economics. But this company operates a viral GTM, has SMB-sized ACVs and retention on par with government contracts. The lesson here is that oversimplifying and labeling can cause you to miss the best opportunities, but understanding the fundamentals and abstracting key business characteristics is a secret weapon.

Subscribe now

If you’re wondering where I have been the last few months, I have started also publishing less quantitative ideas via the Verissimo Ventures Substack, which you can find here:

Verissimo Ventures

Verissimo Ventures

By Alex Oppenheimer

Preferred stock investments work differently from common stock investments. This is absolutely critical to understand when calculating holding values.

When a company gets "written up," it is common practice to take the latest (and highest) preferred share price and apply it to all holdings (common, options, seed, Series A, etc.). This is actually not accurate. Because of seniority in liquidity stacks and the wide range of possible outcomes, Series D shares are worth more than Series A shares (for example).

Exactly how to apply that discount is an entire discussion on its own and is in the realm of secondary investing experts.

When it comes to writing investments *down*, it's important to remember that it only makes sense if either:

A) you previously wrote it up and the company raised subsequent financing at a lower share price than you had written it up to.
or
B) the company is worth less than the size of its liquidation preference as it relates to your investment.

If you invested in a company at $1 per share, then it raised at $3 per share, you're probably going to write that up 3x. If after that, it raised at $2 share ... then we get some complexity. Most likely given how liquidation stacks work, your true holding value will be slightly below that $2 per share. This all depends on the details of the financings themselves.

If you invested at $1 per share and the company has not raised any additional capital, there is no reason to write down that investment unless you think the value of the company is less than the size of the financing you participated in. That means if multiples have come down in the market or the company has not hit growth targets, but it is NOT in danger of becoming insolvent, there is no reason to write down this investment.

Subscribe now

A practical example: you participated in a $25M financing at $150M valuation. The company has $20M left in the bank, but is only at $4M of revenue. The question is not whether the company is still worth $150M (I have made many arguments that it never really was), rather it is whether the company is worth more than $25M: if it is, then you still hold your investment at cost.

You will make your principal investment back at any outcome between $25M and $150M. Above that you have a positive return, and below that negative. This is where misaligned incentives occur:

If a founder wants to keep going for another 2 years to try to get to $15M of revenue, the result for the investors in that $25M on $150M round will likely be the same ($150M exit, they get 1x their money back - excluding any additional financing needs). But for the founder, it makes a huge difference. Let's say they own 20% of the company, that is the difference between the founder earning $30M and $0. But the investors may prefer to just get their money back now because the chance they make meaningfully more is low enough and will take long enough that they think it's just not worth it to risk their principal investment. This is another reason why high valuations are more likely to put companies in a bad position than a good one. If the round was done at $75M post, those same investors would be looking at a 2x, which is a big difference inside a fund.

Subscribe now

Additional thoughts on this subject:

There are also amazing founders who are not great business leaders.

This may sound obvious, but I have observed so many examples in the last few years where the two have been conflated or even completely confused, especially at the founding stage.

Every once in a while, you find someone who has both the operational / execution-oriented chops to build a business AND the visionary / borderline insane mentality of a standout founder. I am lucky enough to work with some of these folks and it’s absolutely inspiring and quite intense. But I don't think it's necessary to have both to be a member of a great founding team.

The most important thing is to maintain the awareness of which special talents are present and make sure they are appropriately complemented by other team members for each stage of the business.

Over the last few years, it seems there was a trend towards viewing starting a company as a step on a career trajectory - by both investors and potential founders. It's not. It's on an entirely different path. You can definitely jump to that path, but it's a nonlinear growth trajectory and requires a major mindset and attitude shift. Going from working for someone to being that person is a major change. I have gone through it over the last few years and it has required a major mindset shift. It’s also exhausting both physically and emotionally.

The CEO/CPO/CTO founder of the company needs to have that itch, that insane drive, that visionary attitude to make things really interesting. It is someone who doesn't need to do 100 customer validation calls to know what product to build, and instead has the killer instinct to bring their vision of the world into reality. Getting excited about something is fundamentally different from being obsessed with that thing.

There are a handful of entrepreneurs in the world who can get excited about something because of the monetary opportunity that it represents and stay excited for 5-10 years. These people are very rare - they have the ability to deal with the ups and downs of a business while keeping their eyes on the prize for an unreasonably long period of time, then once they’re done, they walk away (usually very wealthy) and never think about that market or industry again. Most people who think they have this ability learn that they actually do not, especially when the going gets tough.

The broader truth is that functionally every business needs a bit of everything. You see this as businesses scale: initially they only need a CEO and CTO, then comes Sales and Marketing, then Operations, then they will bring in someone for Product management and Engineering management, then HR, then Finance, then Admin, etc. Each of these roles is actually critical from day one, but may not require the attention of a full time team member until the business gets larger. There is a broader question about how to fill these gaps in the interim, but the focus of this post is on the founding stage.

These various roles build out in a fairly predictable way as businesses scale up. This process of company building is critical to the success of any business. But there is a step before that - it’s the big bang step: the one that starts with something coming from nothing, absolute creativity or invention. Regardless of the skills or experience of an operator, this step may just be beyond their reach: not quite creative enough, not quite aggressive enough, not quite disruptive enough, not quite obsessed enough. There are those who are meant to make that entrepreneurial jump and those who are meant to join the journey somewhere along the way. Sometimes it’s as employee number one, sometimes it’s as the CFO who joins 6 months before the IPO. And usually that changes for each person in each situation and as they develop in their career.

The paradox is that everyone knows that in order to be an entrepreneur, you need to have stamina and the “won’t take no for an answer” mentality, so it’s often hard for people who have those two things to recognize that their vision is just not quite enough. But it’s very hard to know if your vision is actually good enough - you have to believe that it is in order to step into the entrepreneurial ring, but you can’t know for sure. In theory, early investors are the “gatekeepers,” sifting through and selecting for the true visionaries and the teams they need. But investors have actually become generally less sophisticated over the last 10 years in this respect. Not because any investors got worse, just because the number of investors has increased so dramatically in such a short period of time, and their perspective is inherently limited by time.

Subscribe now

So what's the opportunity?

In the last few years, there are more than a few great business leaders who have started companies (and got meaningful funding because they hit certain signals for investors), but they know that they don't have that insane drive towards their vision to take them through the hard times as the fearless leaders of the businesses they started. Most will either involuntarily or voluntarily call it quits, which is totally fine. Starting a company and swinging for the fences is a crazy thing to do, and it’s not for everyone - even the right people usually fail. It’s hard to know that before you give it a real shot. (I am in no way discouraging anyone from making a run at being an entrepreneur.)

These former founders are the next amazing group of Series A/B/C COOs that visionary CEO/CPO/CTOs need to make their visions a reality. To be clear, these operators are absolutely leaders; they are not followers. Followers are not what most great companies need - they need people who will take charge, have strong opinions and make big things happen, but people who benefit greatly from sitting beside a true visionary.

I hope some of the visionaries can see the opportunity to bring in someone complementary to help them realize their vision more aggressively, and that the business leaders have the humility, awareness and understanding to find the right visionary to work with.

It is the overlap between these two talents and the shared experience that will ultimately make them a successful team.

There’s a massive opportunity here for synergistic collaboration, and VCs may play a role in making some of these matches despite the awkward conversations involved.

https://alexoppenheimer.substack.com/p/expensive-options-and-losing-big

And I wrote about investing in turbulent markets here:

https://alexoppenheimer.substack.com/p/capitalization-and-financing-for

But there is one thing I wanted to call out specifically, which is probably the most painful way to be wrong: When you invest early in a company and it ends up being very successful, but you don’t see any real returns.

I think the first thought for many people is: ah, you must have sold secondary too early. The reality is that that is usually not a mistake at all. Exiting positions gradually at specific, pre-defined thresholds is just savvy early stage investing.

The real issues come with mispricing and over financing.

Let’s go through an example:

• Seed: invest $250k into a $2M round at $10M post-money

• Series A: the company raises a $10M at $50M

• Series B: the company raises a $50M Series B at $300M (at $1M of revenue)

• Series C: the company raises $200M at $1.5Bn (at $15M of revenue)

• You can walk through the spreadsheet here

This looks like a 85x+ return on paper - you turned a $250k investment into upwards of $20M and you’re excited to see this thing exit at at least double that (that is the target return of the newest investor and they certainly know what they’re doing). But then… disaster strikes: multiples shrink back down from 100x revenue to a measly 10x revenue, and top line growth also gets hit. Even though the company grew from $500k to $10M revenue (20x!!) in just 3 years, growth has slowed to 80% and the company is only going to do $18M of revenue next year on higher than expected burn.
While there are a handful of founding teams who could make this work and properly manage burn while effectively growing into the valuation, statistically this is not one of those teams. And now you’re looking at a business that, on a good day, is actually worth $200M.

The only problem is that you invested first… so you sit on the bottom of the capitalization stack and your liquidation preferences are below hundreds of millions of others. If the company gets acquired now, then you’ll get $0. Only the latest round investors will get their money back and hopefully the founders will get some sort of carveout or earnout. All of a sudden your 85x+ gets written down to 0x - you had some great MOIC but the conversion to DPI is looking dismal. More likely what happens is that the company is forced into a downround scenario.

It’s important to recognize that pretty much every downround is unique: there are a lot of ways to skin a cat (a fitting metaphor here), so in this theoretical example we’ll stick to the general themes in a downround and use illustrative numbers. (Note for the sticklers: for simplicity we’re assuming no option pool increases in any of the rounds.)

What most likely happens in this downround is some sort of pay-to-play, meaning that in order to maintain your ownership level, you’ll need to write another check. Let’s assume the company needs $25M. The post money will collapse down to $150M, the options will get wiped out and your $250k investment became a $350k investment (an additional $100k required to invest to protect any ownership). This is assuming you have the $100k available (which, if it was an SPV or a rolling fund, then you probably don’t), and even if you raise another SPV, you can run into legal issues because this new group of SPV investors effectively wipe out the first group of SPV investors by using their investment rights. It’s possible to navigate this, but will require waivers from every single initial SPV investor. At the end of this transaction recap, you’re sitting at ~5x.
This is not the worst place to be, but the company still has a long, tough road ahead to exit and you now understand that writeups aren’t everything, so have lost confidence in the meaning of this 5x on paper.

This is your investment in the context of the cap table and how it changes through each financing:

It’s not the worst situation, but it’s a realistic one and it is painful.

Let’s rewind for a minute and go through a couple different versions of history… what if you had sold secondary along the way? (assuming there was an opportunity) Let’s say you programmatically sell 25% of your initial stake at each of A, B and C rounds, and hold the remainder for the long term exit.

That 86x+ sure felt good on paper, but that 25x+ cash in the bank feels a whole lot better. If you effectively executed this strategy, most likely you realized a better return than anyone else who invested in the company. Note that I didn’t make the unrealistic assumption of timing the secondary sales at the top of the market and instead assumed a linear sell-down strategy that one could implement and stick to.

Ultimately venture building and investing is about maintaining an up-and-to-the-right trajectory for as long as possible (not about getting out of great companies with impeccable market timing). Operational and market hiccups happen along the way, so the key is to make sure that those bumps do not cause too much damage to the company’s capitalization, culture or long-term prospects. This is not a business for short term optimization.

Let’s run through another scenario: a more conservative financing trajectory with a reasonable but still impressive exit at $250M.

Scenario 3 might be a lot less exciting, but it looks pretty attractive to me.
If you’re thinking:

• What about the founders?

  • In scenarios 1 and 2, I would, as an early investor, strongly encourage them to sell meaningful secondary along the way. Their new investors are swinging for the fences only, which can create real misalignment with founders for whom a $25M windfall would be life-changing.

• But you *need* 50x-100x returns to really make money in this business.

  • If that is the case then either A) you’re delusional about how much money people make in this business, or B) your fund model needs a major overhaul.

Subscribe now

The most important takeaway:

While investing in the top 1-2 companies of each generation justifies any entry price, you cannot effectively drive returns while operating under that assumption. Let those outliers be an upside, not a necessity. Most *great* companies go through ups and downs, market cycles and operational challenges on their way to successful outcomes. In addition, building and exiting a $250M business is EXTREMELY difficult - ask anyone who has done it. You need to make sure that the financing of a company maintains upside optionality and doesn’t push founders (or early investors) into a corner where the success of the business does not result in the success of the early shareholders.

And of course, the main point of the exercise is the exercise itself - understanding and gaining intuition around these sequential preferred stock transactions makes us all better investors, operators and stewards of capital.

Detailed calculations can be found here: https://docs.google.com/spreadsheets/d/18Z5DNoOrzde4kwJSWu5SSXu86RqIzh0_/edit?usp=sharing&ouid=114625196600448749152&rtpof=true&sd=true

(Forgive any mistakes - this scratchwork was done at 1am)

The headline or implied valuation of a financing is a hurdle: if you clear it, that’s great. If you clear it by a lot, then it might have been good to make it a bit higher (i.e., lower dilution). But if you run into it, it’s very difficult to recover and still win the race (i.e., future down round).

It's critical to remember that almost every company (and every market) experiences ups and downs over the 5-10 year period that early stage companies plan to operate in to find success. As a result, focusing on the terms and scenario modeling is super important for long-term and shared success. In a slight variation on the saying “hope for the best, plan for the worst,” I think early stage companies should “hope for the best, assume it may not work out at all, but make sure that you can still do really well with a positive thought modest outcome.”