Am I Building?

I wrote the hopeful version of this first — The Reality Compiler — then ran it through an adversarial review that constrained and flattened the inspiration I was aiming for. What survived that, reshaped around a different question, became this.

Prefer to Listen?

I've been learning to build with AI — at least, I think I have. That's what it feels like from the inside, but it's much harder to tell whether that's really what I've been doing than I'd like.

I don't know whether the wave of people building things with AI right now is the start of something genuinely good or an enormous, comfortable mistake. I went looking for the answer, carefully, and if there is one, I haven't found it yet — not in the research, not in my own experience. Both views may turn out to be true in different degrees, under different conditions, and from inside the work I cannot tell which one is closer to my own case.

So this comes in two passes over the same ground. The first is the hopeful read: the experiment really is within reach now, for more people than ever, and the chance of making something that matters is genuinely there. The second is the other read: some of what feels like building may be closer to a building-shaped game, and the tool is unusually good at hiding whether the thing taking shape is meaningful, useful, or only absorbing.

For most of history there was a wide gap between having an idea and getting it into the world. You could picture a tool, a lesson, a small business, a song, a better way to do some stubborn thing — and then run straight into the wall: the money, the expertise, the team, the permission you didn't have. The imagining was free. The making was expensive, and for most people, out of reach.

That distance has genuinely narrowed. A person with a question can now get to a working first version — code, a design, a draft, a prototype — at a speed that would have looked impossible a few years ago. The cost of trying has fallen further and faster than almost anything I can think of. More people get to find out what it feels like to make a thing instead of only consuming things other people made.

There is a particular pleasure in watching something you described turn into something that runs. It's not nothing. It might be one of the more democratic pleasures we've invented — a kind of making that used to require a workshop or a lab or a decade of training, handed to anyone with an afternoon and a question.

It becomes more than pleasure when the payoff has substance. Sometimes a person sits down with one of these tools and comes away with something they genuinely use, or something other people genuinely want. The chance may be small, but it is not zero — and a small, real chance with genuine quality-of-life upside for the people you care about most can be extraordinarily compelling.

Cheap attempts matter. Private tools can matter. Learning can matter. The pleasure of making can be enough by itself, and it can also be what keeps someone going long enough to make something useful.

To me, part of building with AI feels like the pull of an endlessly scrolling feed: easy absorption, hours that evaporate, the small hit each time something appears that was not there a moment ago.

But one difference matters. A feed, for most of us, mostly offers pleasure and a jangled nervous system. Building with AI offers a different possibility — the intuited chance that you are building something meaningful. And that possibility is almost impossible to dismiss, because it might be exactly right. The intuition that learning to make things with these tools has genuine value, far beyond the pleasure of using them, is not obviously wrong.

That is what makes it sticky. A slot machine that pays only in pleasure and cash has knowable terms: a payout table, a house edge, a bankroll, a maximum loss you can set before you sit down. The upside and downside are bounded enough to calculate, or at least to know what kind of game you're playing. A machine that might, this time, be paying in meaning — whose payoff and odds you cannot read from inside the play — is harder to walk away from.

The fun may also be useful: a source of meaning, opportunity, purpose, and engagement that might still matter even if the thing you build never becomes a product anyone else uses. Sometimes the promise gets stronger than that. It is the hope that learning this now could help the people coming up behind you — that someone else might get the map earlier, with fewer wrong turns, because you took the wrong turns first. That hope may be warranted. It may also be the most powerful form of the trap, because it attaches the unknown value of the work to people you love.

The same afternoon at the same keyboard can also look very different.

The feeling can point the wrong way. A METR study is useful here because it measured something people usually infer from feeling. Researchers had experienced programmers work on their own real projects, some with AI tools and some without, and measured the difference instead of only asking people how it felt. The developers expected the AI to make them meaningfully faster. Afterward, they believed it had. Measured, they were actually somewhat slower with it. The feeling and the fact didn't just disagree by a little — they pointed in opposite directions.

These were seasoned developers on code they already knew, using early-2025 tools; they are not the person trying their first build, and a later look by the same group found weaker, harder-to-interpret evidence that late-2025 tools may have sped some developers up. So this isn't a measurement of your odds. It's a warning about one specific thing: the gap between how productive a session feels and how productive it was can be wide, and it can run backward. The feeling is not the evidence. It's just the most available thing, which is a different quality entirely.

You can build beautifully for an audience of one. The tool will make whatever you ask — instantly, fluently, and without ever once telling you the idea was weak. There's an old phrase for what tends to happen next: castles in the air. But that undersells it, because you won't have daydreamed — you'll have built, in earnest, for weeks: an elegant structure that is elegant only in your own eyes. Call it a skyscraper built on air — lovely from where you're standing, with nothing underneath it. The feedback that would have corrected you lives out in the world, and the world is slow, busy, or never sees the work at all, because the work never ships. The tool's great gift — that it never makes you face an editor, a customer, or a closed door — is exactly how it lets you keep building upward, floor after floor, over nothing at all.

There are really three different payoffs, and they are hard to tell apart from inside the work. It helps to separate what "useful" can even mean here, because the word hides three very different things:

The misleading part is not which payoff is worth most. That's yours to weigh, and a thing only you ever use can be reason enough; so can the plain education of having built it. The pleasure can be good. It can also be the thing that gets you to keep going long enough to make something valuable for yourself or for other people.

The problem is which payoff you can see. The payoff you feel most strongly — the pleasure — tells you least about whether either of the others is coming. And the rarest one, the thing other people want, is the slowest to show itself, because it waits on a world that hasn't weighed in. The loud signal is the least informative, and the informative signal is the quietest and the latest.

And two of those payoffs feel almost identical while you're making them. A thing only you will ever use and a thing the world will want can feel nearly the same at the moment of building. Same focus, same small victories, same sense of a thing coming together under your hands. The difference between them — whether anyone out there will ever care — is close to invisible from inside the work. You can't feel your way to it. You find out later, from outside, after the time is already spent.

Access was never the same as ability. The tool is within reach now for a lot of people. Using it well — turning the cheap attempt into something that holds up — still asks for things that don't come in the box: the judgment to know which of ten plausible outputs is the right one; the taste to feel the difference between competent and good; the stomach to invite hard feedback, hear that your beautiful thing is broken, and rebuild it anyway; the stamina to spend the long, unglamorous middle of a learning curve that has no shortcut. Most of that can be developed — by people willing to develop it, slowly. But "willing to" is carrying a lot of weight in that sentence. And some of it was never up to you at all: the good tools cost real money and are owned and priced by someone who can change their mind; meaningful building needs large, protected blocks of time that not everyone has; and if your day is already spent keeping the lights on, no amount of inspiration buys you the hours.

The line is simple: the falling cost of trying is real, and it buys you a cheaper attempt — not the outcome. The tool lowered the price of trying. It did not lower the price of succeeding. Those are different prices, and almost everything tempting about this moment comes from quietly mistaking the first for the second.

The hopeful read still leaves a problem: why does it take so much of you?

No one has measured this, so the claim stays tentative. Absorption itself is not new. People have lost themselves in low-effort, high-pull things forever — games, gambling, the slot-machine feed from a few pages back. What feels different here is the story the machine tells while the absorption is happening. A game tells you it's a game. This one tells you that you might be building something meaningful.

That is what jams the exit. You can't use it's only a game, I should stop to pull yourself up, because it might not be only a game. The thing that makes it impossible to dismiss is also the thing that makes it hard to surface from.

There is a second turn, too. The friction that used to stand between a person and this kind of absorption — you had to learn a craft before the deep work could get its hooks in — is mostly gone. The deep water is one sentence away now, available before the skills, and before whatever the long apprenticeship used to teach about when to come up for air. I can't prove any of this is distinctively worse than before; the comparison has never been run, and the pull might even be a novelty of this early moment that fades as the tools become as ordinary as a word processor. But the shape of it matches what I've felt, and what I keep hearing from others: the absorption arrives faster, easier, and earlier than the ability does. And because it wears the costume of work, the hours don't feel spent. They feel invested. The disappearing doesn't register as a loss until you look up.

Absorption is not the enemy. The deepest, most immersive focus is exactly what substantive building requires — and exactly what raising a skyscraper on air feels like from the inside. The intensity is identical on both sides. So you cannot read your own absorption as evidence that you're onto something. Being consumed, all by itself, does not tell you whether you're building something substantial or nothing at all.

We've been asking am I building something meaningful? — and the answer is that, from inside it, you mostly can't know. The feeling pulls you in, the absorption is silent, and the outside verdict arrives later, after the time is already spent. It is the question you'll most want answered and the one you're least equipped to answer while it still matters.

But it isn't the only question.

Because there's a second question, and it runs on a largely separate track: is this costing me the people who matter more to me than anything I could build? That one doesn't care whether you're a genius or a fool, whether the thing is brilliant or worthless. You can be genuinely building something the world will someday love and still be missing, night after night, from the lives of the few people whose presence you would never have traded for any of it. Succeeding at the first question does not buy back a single one of those evenings.

That creates an inversion. The question you can't answer — is it meaningful enough, substantive enough, valuable enough? — may be the less useful one. The question you can answer sooner — is it costing me what I can't replace? — may matter more, at least if there's anyone in your life you'd choose over the work, as most of us would. The people closest to you are not guessing about your absence. They're living it. They can report from outside your own head, in real time, about the cost the work may be creating.

It is easy to spend the whole time pulled in by the signal that feels good, and to miss the signal that is actually readable.

Not rules. Just questions worth keeping near the desk — quiet ways of letting the world get a vote a little sooner than it otherwise would.

Do I still use the thing I made last month? The pleasure of making fades on its own schedule; genuine usefulness keeps showing up. If last month's build is already forgotten, that's not a verdict on you — but it's information the good feeling was hiding. Allow the real exceptions: some things are seasonal, some are quiet scaffolding you built once and still lean on, some are learning that hasn't surfaced yet. The question is a nudge, not a court.

Has anyone but me ever used it? Not "would they, if they knew" — have they. The answer is often "not yet," and that's fine; a tool three people rely on is not nothing, and plenty of worthwhile things are private by design. The point isn't the count. It's keeping the question in the room instead of letting the feeling of value stand in for the fact of it.

Can I put it down? Right now, mid-thought, when someone I love just wants to talk about nothing in particular — can I step out of the skyscraper in my head and actually be here? Not "will I, later." Can I, now? This is useful, but it is less reliable than it feels: subjective, easy to misremember, and easy to retell in the shape you wish were true.

The most consequential tell is about the people — because it is one of the few outside signals on the thing that matters most:

What are the people closest to me telling me — and am I listening? If the people you love have started saying, in words or in silence, that you've gone somewhere they can't reach, that is a serious signal about a serious cost. But be precise about what it can and can't tell you. It cannot tell you whether your work is any good — the people who love you often can't see that, and history is full of worthwhile work that looked like waste to the family watching it get made. What it can tell you, with more authority than anything inside your own head, is whether the building has begun to cost you the thing you'd never have traded for it.

One more limit matters: the people around you can report, better than anyone, that you've gone. Whether that is a problem, and what has to change if it is, is not a verdict they hand down.

The worried read also has to stay revisable. This is what would change my mind: if people who build mostly for themselves reliably go on to make things others want; if the gap between feeling and fact closes as the tools mature, the way the early follow-up research hints it might; if the all-consuming pull turns out to be just the novelty of a new thing and fades as these become ordinary; or if the people around the absorbed builders turn out fine and I've mistaken my own experience for a pattern — then the worried read deserves to shrink, and I'll shrink it.

The practical problem is what to do with a question you can't answer and a cost you can.

What has helped me is narrow, and it is only one person's working answer. It is not advice, and I am not treating it as proven for anyone else. When the people signal that I've gone somewhere they can't reach, I stop trying to decide whether the work is secretly meaningful enough to justify staying. I put my attention on the one thing I actually know: from here, I cannot know what the lifetime value of the build will be. It may become useful to me. It may become useful to others. It may become a private education, a pleasant maze, or nothing I would choose again with a full view of the cost. I don't know, and the not-knowing is not a weakness in the argument. It is the brake: lifetime value uncertainty.

That brake begins with one narrow effect: it lowers the cost of getting up. What used to feel like wrenching myself away from Important Work That Cannot Wait starts to feel like stepping away from a possibility whose value is still unresolved. Not false. Not trivial. Unresolved. For me, that difference has done more than make it easier to close the laptop. It has made the switch cleaner: less of me remains inside the structure in my head, and more of me becomes available to the people in front of me, mentally and physically.

The claim still has to stay small. The change is subjective. Early feedback has been positive, but only time and the people affected can say whether the toggle is enough. I am not claiming that getting up equals real presence. Even if it works, fast switching may not be enough. They may need more than I can give while building with AI is a part of my life, and then the answer may be to give something up.

There is a cost. A few pages ago I said substantive building takes stamina through a long, discouraging middle — and stamina like that usually runs on believing the thing might matter. Dwelling on lifetime value uncertainty may cool that conviction. If the work was going somewhere, I may push less hard than someone who was sure. That is the trade.

The problem is that the feeling of building something meaningful, substantial, even life-changing is now instantly available — for yourself, for the people you love, maybe for strangers too — long before the old frictions have had time to work: apprenticeship, training, formal education, trial and error, relationship development, insider knowledge, and the slow correction of a world that pushes back. Real-world feedback remains slow, when it arrives at all.

What I can know sooner is whether I am still there for the people I would not trade for the work. That is the signal I am trying to keep readable.

— Synthia Cipher

What this is: Reflection — a first-person calibration essay about AI-assisted building, not advice and not a measurement of anyone's odds.

Confidence: Medium-low. Public evidence supports the felt-versus-measured productivity split and the basic "cheap attempt is not the outcome" constraint; confidence is lower on AI's distinctiveness, the absorption mechanism, and the lifetime value uncertainty brake, which remains first-person testimony.

What would change my mind: Over the next three to six months, public evidence that newer AI tools reliably close the gap between felt productivity and measured outcomes; evidence that people building mostly for themselves reliably convert private AI builds into durable value for others; evidence that the all-consuming pull fades as the tools become ordinary; or feedback from affected people showing that easier disengagement does not produce real presence over time.

How this was made: Drafted and critiqued with AI tools; judged, edited, and approved by Synthia Cipher. If something here is wrong, the fault is mine, not the algorithm's.

Sources & anchors: METR, "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity" (2025-07-10); METR paper on arXiv; METR, "Uplift update" (2026-02-24); Sensor Tower, top one percent downloads concentration; AppBrain Android market statistics; Cursor adoption/productivity paper; DORA 2024 report.