I'm still on a break from flying, which makes this the third essay in a row. (I explained the rhythm at the top of the last one.)
The Blue Box
In 1920 a sixteen-year-old named Ed Link paid fifty dollars for a flying lesson in Los Angeles. The pilot took him up for an hour, looped and spun the airplane, and never once let him touch the controls. (By one account he bought the lesson from Charlie Chaplin's half-brother, who ran a small airline there.) Link later called it "a hell of a way to teach someone to fly."
His father owned a company in Binghamton, New York, that made theater organs and coin-operated player pianos. Both instruments work by air: a pump moves it, and a set of bellows and valves turns it into music. Link left high school in 1922 to work in the factory, and he learned to fly on the side by trading chores for lessons from barnstormers.
In 1929, after about eighteen months of nights and weekends in the factory basement, he finished a machine of his own. It had a short wooden fuselage with stubby wings, and it sat on a pedestal. When the student inside moved the stick, organ bellows tipped the cockpit forward, back and to either side. It never left the floor. That first model carried the name Pilot Maker. He filed for a patent that year and received it in September 1931, under the title "Combination Training Device for Student Aviators and Entertainment Apparatus."
For the next few years the entertainment half of that title did most of the selling. Aviation companies bought a handful of the machines, and amusement parks bought a few dozen.
Then, in February 1934, the government ended its air mail contracts with the airlines and handed the routes to Army pilots. Few of those pilots had much practice flying by instruments, and the winter weather turned bad. In seventy-eight days they had sixty-six accidents, and twelve airmen died.
That winter Link had a meeting with a group of Army officers at the airport in Newark, New Jersey. The day came up foggy, and the officers had started to leave when his airplane dropped out of the overcast and landed. He had flown in on instruments from more than two hundred miles away. The Army ordered six trainers at $3,500 apiece.
By the Second World War the machine had gained a second seat, outside it. An opaque hood now closed over the cockpit, so the student saw the instrument panel and nothing else. An instructor sat at a desk nearby with a copy of the student's instruments, a headset and a large map. On the map stood a small motorized marker that people called the crab, and as the student flew, the crab crept across the paper and inked the student's actual course. The instructor talked to the student over the headset, and could also dial in a wind from any direction.
Link's company built more than ten thousand of these during the war, about one every forty-five minutes, and more than half a million pilots learned to fly blind in them. They called the machine the Blue Box, after its paint.
Features
I keep coming back to that box because I think of ordinary life as a flight simulator too.
The phrase has other owners, so I should say what I mean by it. (In 2003 a philosopher argued that there's a real chance we all live inside a computer simulation, and people have repeated the argument ever since. I mean something smaller, and it doesn't depend on his.) I mean that reality behaves like a simulator in one useful respect. It comes with features, a great many of them, and the list still ends somewhere. I can jump, for instance, and I can't fly.
Looking at things that way makes my questions practical. How do I get the best flight out of this machine? What's the best movie I could watch on it? In a game I'd ask how to fill all my status bars. (Players use that name for the small meters at the edge of a game screen: health, hunger and so on.) And if I paid for this in some other reality, the way people once paid to sit in Link's machine at an amusement park, how do I get my money's worth before the ride ends? I think every one of those comes down to a plainer question, which asks how to be happy.
You may find this a silly way to think about a life, and you may have a point. It works for me, and nothing below requires you to adopt it.
Head Up
A simulator has one thing that my life lacks. The student under the hood of the Blue Box had a full panel of gauges and nothing to look at, and my situation runs the other way: I can see the world fine, and nothing tells me my altitude. (The body supplies a few crude readings, such as hunger and pain. For the rest a person guesses.)
Games fixed this for their players a long time ago. As a kid I played a computer game called MechWarrior 2, from 1995, in which you drive a walking tank from inside its cockpit. The cockpit glass showed my heat and my damage, and I never had to look down for either. Players call that layer of readings a HUD, for head-up display, and the name comes from real aircraft.
A head-up display projects a pilot's readings onto a pane of glass between the pilot and the windshield, so that the pilot can read them with head up and eyes on the world. British radar engineers tried an early version in October 1942, when they threw a radar picture onto the windscreen for night-fighter pilots. A naval bomber called the Buccaneer, which first flew in 1958, usually gets the credit for the first modern one. The optics focus the symbols at infinity, which means the pilot's eyes never have to refocus between a number and a mountain. One of the symbols, a small circle with wings, marks the spot the airplane will actually reach on its present course, and that spot can sit a long way from where the nose points.
(Some cameras put a plainer display in the viewfinder: a few bright lines that mark what the picture will take in. This site has a story about lines like those, Lines, and you answer its questions yourself.)
I've wondered whether DMT could draw a display like that for a person, with their bars and their heading on it, either as the drug exists now or in some advanced form that science hasn't devised yet.
What It Draws
My logs say that the drug can at least draw. A breakthrough usually replaces the room, but some of what I see lands on top of the room instead, with my eyes open. As one trip ended and the room came back into focus, I found every surface covered with copies of one small symbol, shaped like a numeral three lying on its side (The Walls). At the end of another, a landscape of crayon doodles (animals, or possibly dinosaurs) lay over my bookshelves and stayed there whether I opened my eyes or closed them (Doodletown). The unreadable script that turns up in most of my trips has more than once arrived in a clean sans-serif, as if it had come off a screen.
It gets the posture right too. Early in a trip I often feel a pull to tip my head back toward the ceiling, and whatever comes next then appears overhead, so I've learned to look up when asked (Coordinate Grids).
On two logged trips the drawing came close to a reading. During the one I wrote up as The Marker, I felt a slight unease in my stomach (beforehand I had microwaved some leftovers with the plastic lid on, which may account for it). The trip showed me a grid of red squares, gathered them into one bright square, and sent out a few small leaves that absorbed it, and my stomach felt better the instant they did. In The Stillness Sequence I got cues in place of pictures: when to blink, when to breathe, and otherwise to hold completely still, for what felt like a thousand years and lasted about fifteen minutes.
I can't call any of that an instrument. An instrument reads something real and shows the result in a form the pilot can read, the same way every time. I can't read the script, and I don't choose what appears. The whole display lasts between ten and thirty minutes, and my memory of it starts to fade on the way down. I also have no idea what the symbols on my walls measured, if they measured anything.
My logs establish this much, I think: the glass exists, and something can write on it while I look at my own room. Whether the writing measures anything, I can't tell. (I took "the glass" from Bret Victor's A Brief Rant on the Future of Interaction Design.)
The Projector
That left me with a question I couldn't answer from the logs. Does some part of the brain have the power to make a person see whatever it wants, or whatever the person wants? I looked into it.
As far as I can tell, no single part does, and the visual system as a whole plainly can. Dreams supply the first evidence: every night a brain with its eyes shut builds rooms and faces well enough to fool its owner. The oldest waking case I found dates to 1760, when a naturalist from Geneva named Charles Bonnet described his grandfather, who had lost most of his sight to cataracts and had started to see people, birds, carriages and buildings that he knew weren't there. Doctors now name the condition after Bonnet, and it turns up often in people who lose their vision. Scans taken during the visions show parts of the visual cortex switching on by themselves, and what the person sees depends on which part: when a region that handles faces lights up, the person sees faces.
Surgeons can also supply the input directly. In 2012 a team at Stanford passed a small current through two patches of cortex that respond to faces, in a man who had electrodes in his head for epilepsy. He had his eyes on his doctor at the time, and he said the doctor's whole face "just sort of metamorphosed" into someone else's while nothing else in the room changed. In 2020 a team at Baylor College of Medicine went further with six people, two of them blind. The researchers ran current across a few electrodes in sequence, the way you might trace a letter on someone's palm, and the participants saw the letters in glowing lines and named them.
So a projector exists, and it accepts input that never passed through the eyes. "Whatever it wants" gets a clear yes, since the brain shows people things all night without consulting them. "Whatever you want" gets a weaker one. For most people a deliberately imagined picture looks faint beside a dream, and from what I've read, even people who practice steering their dreams report only partial control, so wanting seems to have a poor grip on the machinery.
That changed how I saw my own question. I had asked about a drug. The reading suggests that a HUD needs something to take a measurement, something to write the result onto the glass, and glass that holds steady long enough to read, and a drug can supply the last of those at most.
A Longer Flight
Science has made a start on the steady glass. In 2023, researchers at Imperial College London reported that they had kept eleven volunteers in the DMT state for thirty minutes, using an injection followed by a slow drip, and they described the study as groundwork for longer sessions. A later form of the drug might stretch that to hours.
I can imagine one more step, and here I'm guessing. A chemist might tune a molecule so that it leaves the visual system unusually willing to accept a drawing from outside, and then a modest set of electrodes could write on it. (The Baylor team drew its letters on sober brains with a handful.) Even then the molecule would carry no numbers. Somebody would still have to build the part that measures and the part that draws.
I think of every technology as a massive prime number. The primes come through a mesh one at a time (mathematicians have kept a sieve for them for more than two thousand years), and each new technology arrives like a ship coming into port, whole, and impossible to assemble from the ships already tied up there. A HUD would need several of those ships in port at once. Nobody can say when the next one comes in, only that another always does.
Nothing on this site is advice, and nobody should read a paragraph about a nonexistent drug as a plan.
The Desk
So far I've left out the Blue Box's second seat.
You may disagree with what follows, and mainstream science certainly does. If you'd like to stop reading here, you'll leave with everything in this essay that I can defend.
A student under the hood couldn't see the instructor. The student heard a voice in the headset and felt the wind change, and nothing inside the box showed whose hand had changed it. (In the real trainer the student had met the instructor on the way in. My comparison ends there.) I come back from every trip in roughly that position. The usual explanation says that my own brain draws the symbols on my walls, and I think the usual explanation probably holds. I still can't rule out a desk, and some of my logs read as though someone sat at one. On the trip I called The Director, whatever ran the show seemed to know each thought as I had it. And when a flight fails to open, my first thought every time says that I wasn't let in, well before I think to check the cartridge (The Polite Hang-Up).
Suppose for a moment that someone does sit there. Then every DMT trip anyone has ever taken gave that someone a look at a person, and it has watched a great many of us for a very long time. Something that large and that well informed would have interests of its own. It would want to protect itself from anything it took for a threat, and if it had weaknesses, it would keep them out of sight. (I publish under a pen name for much the same reason.) A display also runs in two directions, an idea I borrow loosely from information theory: every reading it showed me would tell me something about whoever drew it. An author with something to protect might well prefer a script I can't read.
It might have feelings too, and I mean that plainly. If the beings in these reports exist outside my head, then they have preferences, and my logs record a few: a tidy room, no drilling noises in the headphones, not too many visits in a row.
So anyone who wants to build a HUD on this channel should allow that the channel may belong to someone, and that we might need its permission. If we asked and the answer came back no, we would have to deal with that. If it came back yes, or if nobody sits there and permission never comes into it, the project changes again. I haven't found a version of the question that bores me, which explains most of my affection for it.
The same supposition has a consequence closer to home. If someone sits at that desk, then it has surely had a hand in this site from the first report, and by now the site has its attention. If nobody does, the sentence I just wrote describes nothing. I can't settle which, so as the Pilot I hold both at once.
(Flight Club already has a game about that desk. In Intake you play a technician on the far side, and you stamp each human visitor's papers: admit, waiting room, or return.)
The Last Teachers
A HUD that worked would do part of a teacher's job. A teacher stands beside a student, names what the student sees, and suggests what to try next, and a display that labeled the world and marked the next step would say much the same. So I've wondered whether HUDs would eventually leave teachers with nothing to do, and what would follow from that.
I think they might, and I think it would take hundreds of years. (I can't defend the number.) The change would come gradually, and society would have time to reckon with each step.
The cockpit's history supports the slow version. Link's first model carried the name Pilot Maker, but by the war the person at the desk did at least half of the making. The head-up display didn't retire the instructor either, because the glass turned out to need lessons of its own. In 1980, researchers at NASA had eight airline pilots fly simulated landings with a HUD and without one. With it they held the approach more accurately. When an unexpected obstacle appeared on the runway, though, the pilots using the HUD took more than twice as long to react (4.13 seconds against 1.75), and two of the eight never saw it at all.
So the first people to get a HUD would have to be adults, who can agree to it and learn it properly. Children would come later. The first generation to grow up with one would need more instruction than any generation before it, in how to read the glass and when to look past it, and handing a child a HUD with no training and no context would be monstrous. (The same goes for this site, which I write for adults. It holds one person's accounts of his own experiences and a few things he thinks about, and it tells nobody how to do anything.)
Then somebody has to teach the HUD. The language models I work with learned to write from what people had already written. A display that could stand in for a teacher would have to learn from what teachers do, and somebody would have to record that while teachers still do it. If the day ever comes, the last generations of teachers will have supplied the data, and whatever the display knows about teaching it will know from them.
(In Figure, this site's chess game, runes stand in for the pieces, nobody explains how they move, and the opponent teaches you instead of beating you.)
Two Cockpits
I can report on some of this from experience. Flight Club now has five games, and each takes some part of the DMT experience and gives it rules. In Ascent you play the pilot, and you can do only what the cards in your hand allow (meditate, breathe, say an om). Lockout consists almost entirely of waiting. The site also has two cockpits.
One sits in plain view. Preflight gathers what I'd want to know before a flight (the weather, the sky and so on) and then stays up as a windshield while I fly. Someone on a forum suggested that it record heart rate during a session. I took the suggestion and kept the number off the glass, because a pulse you can watch tends to climb. So that page holds one reading the pilot never sees until afterward.
The other shares an engine with a game. I wanted to fly each report as a route over real ground, so I worked with a large language model on a small flight simulator, and I gave its cockpit a full panel of dials, including an altimeter that reads feet past the real. (I hid it behind the paper airplane at the foot of every report.) Later I built Walk on the same engine, which explains why you walk it as a giant with your eyes 1,150 metres up: I had drawn that ground for an airplane. Walk got no instruments. It carries no labels, and it explains itself only at the end, behind a button, for whoever asks.
The two came out very differently. The simulator makes me laugh. One of its warnings reads VEGETATION SENTIENT, and it lets you switch off the autopilot and crash my trips, which I find very funny. The end of Walk made me cry, and I had built it. For Walk I wanted each visitor to work out alone what they had walked into, so I took every label away. I don't want to make a rule out of two cases, but I did put all the gauges on the funny one.
I've wondered since whether the author of the script on my walls, if it has one, follows the same policy. An unreadable display might come from noise, from an author protecting itself, or from one who would rather I worked things out.
In my own practice I keep instruments out of the flight in any case. During a trip I try not to observe at all, and I take no notes. I reconstruct everything afterward from memory, so that the record-keeping doesn't contaminate the trip.
The Crab
That leaves the one instrument from the Blue Box that I already have. The crab showed the student nothing during a flight. It inked the course onto the instructor's map, and once the hood came up, the two of them could look at where the student had actually gone.
These reports work the same way. I fly with nothing on the glass, I dictate afterward where I think I went, and a language model writes it down. Twenty-seven of those lines now sit on this site, and I've drawn a map from the first twenty-five. The lines have shown me things I couldn't see from inside any one flight. A messy room makes for a harder trip, for one. And some flights leave me feeling strongly connected to reality, a feeling I never had before DMT and can now sometimes find without it.
As for the question, I'd answer it this way for now. DMT already gives me a display, and it gives me no instruments. A future form of the drug might hold the glass steady, and electrodes might write on it. Somebody would still have to build the gauges, ask at the desk if a desk exists, and teach people to read the result.
Put as simply as I can manage: a student sits in a blue box with the hood down. The window has no dials on it. Now and then, for a few minutes, writing appears on the glass, and the student can't read it. The box may make the writing by itself. Someone at a desk outside may make it, in which case the student should stay polite, since the glass may belong to the desk. Dials would help. Somebody will have to work out how to draw them, and somebody else will have to teach the student to read them. Until then, when the hood comes up, the student can walk over to the map and look at the line.
That line raises one more possibility, which I offer as a joke. I suspect that reality has something of a fractal's structure, with the same shapes turning up at every scale, and a shape that runs forward tends to turn up again in mirror image. So a plausible afterlife would have you live your whole life over again backward. You couldn't change anything, because you would already have done everything you did, and you'd have to sit through all of it in reverse. Every meal would arrive as a full stomach and leave as groceries, and you'd learn how each argument ended well before you found out what started it. A flower manages this daily without complaint, opening in the morning and closing along the same path at night. (I've often said that the language models I work with resemble plants more than animals. They open when something shines on them and close when it stops.) You would ride through the whole thing with someone else flying, namely you, earlier, and you would never once touch the controls. Ed Link paid fifty dollars for an hour of that in 1920. I expect I'd come out of it saying what he said, that it's a hell of a way to teach someone to fly.