You’re Not Behind. You’re Early.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.