I wrote a piece a while back about how much I hate flying, which I composed in a foul mood in Frankfurt after United lost my luggage, and which ended on a line I still believe: flying didn't get worse by accident, it got optimized, and the optimization function doesn't include you.

I blamed the airlines. That was the easy half.

The harder question showed up later. Airlines don't build airplanes. They buy them, from essentially two companies, and those two companies decide what a passenger aircraft is going to be for the next thirty years. So I went looking for who actually makes that call, expecting to find somebody to be angry at.

I didn't find one. What I found was more interesting, and more uncomfortable, than a villain.

The Plateau

In 1958 a passenger could board a Boeing 707 and cross the Atlantic. Today that passenger boards a 787 Dreamliner, and almost everything about the experience of the machine has improved beyond recognition. The engines are dramatically more efficient and vastly quieter. The airframe is composite rather than aluminum. The avionics would have looked like science fiction in 1958. The cabin is pressurized to a lower altitude and humidified, so you arrive less wrecked. And the safety record of commercial aviation over that period is one of the genuine triumphs of modern engineering, a curve so good it has almost stopped being newsworthy.

The cruise speed went from roughly Mach 0.8 to roughly Mach 0.85.

I want to be careful with that number, because the sloppy version of this argument says the speed never changed at all, and an aviation-literate reader will stop trusting you immediately. It did change. Modern widebodies cruise a few percent faster than the early jets, around Mach 0.85 by manufacturer specification, while narrowbodies like the 737 and A320 families sit lower, nearer Mach 0.78 to 0.80. The 747 cruised a touch faster than many of the single-aisle jets flying today.

So the honest framing isn't that aviation stood still. It's that across seventy years of furious engineering progress, the single dimension a passenger experiences most directly moved by a few percent, while nearly every other dimension moved by multiples.

And then there's the part that genuinely surprised me. Some published flight times have gotten longer. When fuel prices climbed hard through the 2000s, airlines discovered they could save real money by flying a bit slower, since the fuel saved exceeded the value of the minutes lost. Carriers tune this deliberately through what's called a cost index, which trades fuel against time according to what each is currently worth. Airlines also pad schedules to protect their on-time statistics against congestion and delay, so the number printed on your boarding pass drifted further from the aircraft's capability.

Read that carefully and the picture inverts. We're not flying as fast as our airplanes can go. We're flying as fast as it's worth flying them, and somebody did that arithmetic on purpose.

The Wall at Mach One, and What It Actually Costs

The physics here is real, and it's where any honest version of this essay has to start.

As an aircraft approaches the speed of sound, air accelerating over the curved upper surface of the wing hits Mach 1 locally well before the aircraft itself does. That's the critical Mach number. Push past it and shock waves form on the wing, and drag doesn't rise gently, it diverges. Engineers call the point where that happens the drag-divergence Mach number, and NASA's own aerodynamics literature treats raising it as one of the central achievements of transonic wing design, through swept wings and the supercritical airfoils developed at Langley in the 1960s and 70s.

This is the part people mean when they say the speed is physics. And they're partly right.

But notice what the physics actually says. It doesn't say Mach 0.85 is a wall. It says drag climbs steeply somewhere past it, so each additional increment of speed costs disproportionately more fuel. That's not a prohibition. That's a price.

Which means the real question was never "can we fly faster." We demonstrably can. The question is whether anyone is willing to pay what the next increment costs, and that turns out to be a question about economics and institutions wearing a physics costume.

I'll be straight about a limit in my own research here. To tell you precisely what going from Mach 0.85 to Mach 0.90 costs a specific airframe per flight, you need manufacturer performance data that isn't public, and fuel burn doesn't scale uniformly with speed, so a back-of-envelope guess would be worse than useless. I'm not going to invent a number. What I can tell you is the shape: the curve bends upward sharply in that region, and every operator buying airplanes knows exactly where it bends.

Who Actually Decides

Here's the thing I had wrong when I was angry at airlines.

Boeing and Airbus don't independently decide what gets built. They build what airlines will buy, and airlines buy against a spreadsheet that has almost nothing to do with how fast you get there.

What airlines actually optimize is cost per available seat-mile. Fuel burn per seat. Maintenance cost. Dispatch reliability, because a cancelled flight is catastrophically expensive. Fleet commonality, because a pilot type-rated on one variant can fly another with minimal retraining, and because maintenance infrastructure and spare parts inventories are enormous fixed investments. Residual value, because a huge share of the world's fleet is leased and somebody has to care what the asset is worth in year twelve.

Speed is on that list. It's just far down it, and it trades directly against the item at the top.

Show me the incentive and I'll show you the outcome. An airline buying a jet that burns twenty percent more fuel to save passengers forty minutes is an airline that loses to the competitor who didn't. Not because its executives lack ambition, but because the fare a typical passenger will pay for those forty minutes is smaller than the fuel bill. Travelers say they value time. At the checkout page they sort by price, and they do it so reliably that the entire industry has reorganized around it.

So the manufacturers optimized for exactly what their customers rewarded. They got extraordinarily good at it. And an entire axis of progress quietly fell off the roadmap, with nobody ever deciding to drop it.

The Thirty-Year Bet

The time horizon makes this far stickier than ordinary market preference.

A commercial airliner is not a product in the sense a phone is a product. It's a capital asset with a service life measured in decades, bought with debt or leased through structures that assume predictable economics across multiple business cycles. An airline ordering aircraft today is making a claim about fuel prices, route demand, labor costs, and regulation deep into the 2050s.

Under those conditions, "radically different" is a synonym for "unquantifiable risk." An incremental improvement with well-understood operating costs beats a revolutionary design with uncertain ones, not because the buyer is timid but because the buyer is financing it over twenty years and the lender wants a residual value estimate.

Scott Kirby made a version of this point on the Cheeky Pint podcast recently, talking through United's fleet decisions with John Collison and Ben Thompson. His framing of aircraft as very long-duration bets, and his argument that air traffic control rather than aircraft performance is his real operational constraint, are worth hearing. I'd treat it as an airline CEO's testimony about his own incentives rather than as established industry economics, which is also how he presents it.

But the structural point stands on its own. Long asset lives create a preference for continuity. That preference is rational for every individual actor, and collectively it produces a fifty-year plateau.

Two Companies and No Villain

Then there's the market structure, which is where people usually expect the indictment.

Large commercial aircraft is effectively a duopoly. Development costs for a clean-sheet airliner run into the billions before the first revenue flight. Certification with the FAA and EASA is a multi-year undertaking whose cost scales with novelty, which creates a powerful incentive to derive new aircraft from existing type certificates rather than start over. Supplier bases are concentrated and deep. Production learning curves mean unit costs fall substantially over hundreds of airframes, so a new entrant is competing against decades of accumulated manufacturing experience it cannot shortcut.

The barriers are real, and they're mostly not conspiratorial. They're the natural consequence of building something that must not fail, at scale, profitably.

What I'd resist is the conclusion that the duopoly simply stopped innovating. The 737 versus A320 and the 787 versus A350 competitions were ferocious, and they produced genuine advances in composites, engine efficiency, and operating economics. Competition there is intense.

The duopoly doesn't appear to determine how much innovation happens. It shapes which kinds. Both firms compete hard along the axes airlines score them on, and neither has an incentive to open a new axis unilaterally, because the first mover on speed would carry all the certification risk and development cost for a benefit its customers have shown they won't pay much for.

I'd also note, carefully, that the 737 MAX story is often conscripted into this argument and shouldn't be flattened into it. What happened there involved engineering decisions, certification process, competitive time pressure, and organizational culture interacting badly. It's a serious case study in how institutional incentives can corrupt safety judgment. It is not evidence about cruise speed, and borrowing its moral weight for an argument about Mach numbers would be dishonest.

The uncomfortable version is this. There's no villain. Every actor is responding sensibly to the incentives in front of them, and the aggregate outcome is a half-century freeze on one dimension of progress. That's a much harder problem than corruption, because there's nobody to fire.

What Kelly Johnson Removed

Which brings me to the comparison I actually wanted to write about, and the one where I think I have some standing.

In June 1943 the Army's Air Tactical Service Command told Lockheed it urgently needed a jet fighter. Kelly Johnson's hand-picked team designed and built the XP-80 Shooting Star in 143 days, seven fewer than the deadline required. Lockheed's own account of it includes a detail I find almost unbelievable: the formal contract for the aircraft didn't arrive until October 16, 1943, four months after work had already started. The company began building a fighter jet on a handshake. They worked out of a rented circus tent next to a plant that smelled bad enough to give the operation its name.

Johnson wrote down how he did it, as 14 Rules and Practices that Lockheed says are still in use. The themes are familiar to anyone who has worked inside a large engineering organization: a small number of very good people, a program manager with near-total authority, minimal external reporting, few people with access, and extremely tight feedback between designing a thing and testing it.

I work on cloud and distributed systems inside a very large company, which is not aerospace but is the same organizational physics. I've watched small teams with clear ownership move at speeds that look impossible from outside, and I've watched capable people spend quarters producing documents. The difference is almost never talent. It's how many people hold a veto, how far a decision has to travel, and whether the person who will live with the consequences is the person making the call.

So the Skunk Works lesson is real. But the comparison breaks in ways that matter, and the honest essay has to say so.

The XP-80 was a wartime prototype for a military customer with an existential deadline, built in a handful of units, with a risk tolerance no civil aviation authority would permit for a moment. Nobody's family boarded it to visit grandparents. Commercial aircraft carry four billion passenger journeys a year, and the certification process people complain about is a substantial part of why boarding one is among the safest things you will do all week. Much of the bureaucracy that slows commercial development down isn't waste. It's the accumulated scar tissue of accidents, written into process so they don't happen twice.

The useful question isn't "why can't Boeing move like 1943 Skunk Works." It's narrower and much more interesting: of everything Johnson removed from the engineering process, how much was genuinely protecting safety, and how much is organizational sediment we've since mistaken for safety? Those two look identical from the inside. Distinguishing them is the whole game, and it's the same problem every large institution has with its own defaults.

Concorde: Capability Was Never the Constraint

If you want proof that the engineering was never the binding limit, it flew for twenty-seven years.

Concorde cruised at 1,350 miles per hour, more than twice the speed of sound. British Airways' own account notes that a typical London to New York crossing took a little under three and a half hours, against about eight hours subsonic, and that the record run in February 1996 did it in 2 hours 52 minutes and 59 seconds. The airframe stretched between six and ten inches in flight from the heat. It was, by BA's reckoning, the most tested aircraft ever certified at the time, with 5,000 hours of testing before carrying a passenger.

It also carried about a hundred people at a time, burned fuel at a rate that only made sense with transatlantic business fares, was barred from supersonic flight over land because of the sonic boom, and in its entire service life flew just under 50,000 flights carrying 2.5 million passengers. For comparison, that's a rounding error against what the global fleet does in a single week. British Airways withdrew it in October 2003.

So was Concorde a commercial failure, or evidence that the industry wouldn't pay for speed?

I think it's genuinely both, and the distinction matters. Concorde proved supersonic passenger service was technically achievable. It also demonstrated that at that fuel burn, that capacity, and that route restriction, the economics only closed for a tiny premium market. The boom ban wasn't an economic constraint, it was a regulatory one responding to a physical phenomenon. Three different kinds of limit, stacked, and people routinely collapse them into one story about lost ambition.

The current attempts are interesting precisely because they're attacking the stack rather than just the speed. NASA's X-59 exists to demonstrate that a sonic boom can be reshaped into something closer to a distant thump, which would let regulators write noise-based standards instead of blanket speed prohibitions over land. It first flew in October 2025 and went supersonic for the first time on June 5, 2026, reaching Mach 1.1 at 43,400 feet on an 81-minute flight.

The best detail in NASA's account is what the pilot reported. Jim Less said he knew he was supersonic because the gauges said so, and that he didn't feel anything. Which is the entire thesis of the program in one sentence: the barrier was never the speed, it was the noise the speed makes on the ground. The next milestone is Mach 1.4 at 55,000 feet over actual American communities, to find out what people below think of it.

That's a real achievement and it is not a commercial aircraft. Boom Supersonic is working the economics side. Neither has yet shown that the full stack closes.

The Airport Is Part of the Aircraft

And now the part that I think dissolves the original question, which is why I saved it.

Nobody actually wants a faster airplane. People want to be somewhere sooner. Those are not the same product, and cruise speed is one term in a sum with many terms.

Take the trip I actually make. Getting to Dulles. Parking or being dropped. Checking a bag. Security, which on a bad day at a US airport is its own small ordeal. Walking to the gate, which at IAD may involve that standing-room bus I complained about for a full paragraph. Boarding, which is a ritualized scramble. Pushback, taxi, and a queue for the runway. Then, finally, the part the Mach number governs. Then descent, holding if the airspace is busy, taxi in, the jet bridge, bags, and getting from the arrival airport to wherever I'm actually going.

Use British Airways' own numbers for the flying. London to New York was about eight hours subsonic and about three and a half on Concorde, so supersonic cruise removed roughly four and a half hours of airborne time. That's enormous and I don't want to minimize it, because on a long transatlantic route speed genuinely dominates.

Now run the same arithmetic on a domestic hop. If the airborne portion is ninety minutes and the surrounding process consumes three hours, then doubling cruise speed, an achievement that would require supersonic flight over land, new engines, a new airframe, and a rewritten regulatory regime, takes a four and a half hour door-to-door trip down to about three and three quarters. That's my arithmetic rather than a published statistic, and the inputs vary enormously by airport and traveler. But the shape of the result is robust: on short routes, the flight is a minority of the journey, and optimizing a minority term has bounded returns.

Which reframes everything. Kirby's claim that air traffic control is his binding constraint, not aircraft performance, stops sounding like deflection. Airspace capacity, separation requirements, routing that isn't a great circle, runway throughput, gate availability, and turnaround time are all terms in that sum, and several of them are larger than cruise speed on most trips people actually take.

If the goal is genuinely to get people places sooner, the highest-leverage investments may not be in airplanes at all. That's an unsatisfying answer for anyone who grew up on the SR-71. It's probably the correct one.

Three Constraints, and How to Tell Them Apart

So here's what I actually came away with, and it generalizes well past aviation.

When any axis of progress stalls, there are three candidate explanations, and they are constantly mistaken for one another.

A physical constraint is binding regardless of money or will. Transonic drag divergence is real. The energy cost of pushing a body through air near Mach 1 is not negotiable by committee.

An economic constraint means it's possible but nobody will pay. Speed past Mach 0.85 is available for a fuel bill, and the market has consistently declined to buy it. This looks like a hard limit from inside the industry, because for any individual firm it functions as one.

An institutional constraint means it's possible, somebody might pay, and the organizations involved can't or won't produce it. Certification designed around derivatives rather than clean sheets. Procurement cycles longer than the technology cycle. Decision authority so distributed that nobody can commit. Regulation written as a speed prohibition when the actual concern was noise.

The reason this matters is that the three have completely different remedies, and confusing them wastes decades. You cannot innovate your way around physics. You can change an economic constraint by changing what gets rewarded, which usually means pricing something that was previously free. You can change an institutional constraint by restructuring authority, which is the actual lesson of the circus tent in Burbank.

Aviation's speed plateau is some mixture of all three, and the mixture differs by segment. Over the ocean it's mostly economic. Over land it was mostly regulatory, which is exactly why the X-59 is pointed at the noise standard rather than at going faster. Inside the development process it's substantially institutional. And the drag curve is sitting underneath all of it, perfectly real, setting the price everything else is negotiating against.

The failure mode is treating an economic or institutional limit as though it were a physical one, because that's the version that requires nothing of anybody. "It's physics" ends the conversation. It also happens to be the most comfortable thing a large organization can believe about its own plateau.

What Progress Was

I started out looking for someone to blame and I'll admit I'm a little disappointed not to have found one.

What I found instead is that we did not stop making progress in aviation. We made staggering progress, measured against a scorecard that was quietly rewritten sometime around the first oil shock. Safety, fuel burn per seat, emissions per passenger-mile, reliability, range, and above all accessibility. Flying used to be for the wealthy. I didn't get on an airplane until I was a teenager, and now I complain about it from an airport lounge in Frankfurt, which is its own kind of answer.

That rewrite was mostly correct. If you'd offered 1958 the choice between halving the crossing time and making the crossing dramatically safer, cheaper, and available to ordinary people, the right answer is obvious and it's the one we took.

But a scorecard you didn't notice adopting is still a scorecard. Nobody voted on it. It emerged from a few decades of airlines buying sensibly, manufacturers responding rationally, regulators writing rules in response to real accidents, and passengers clicking the cheapest fare. And once it set, it became invisible, which is the thing I keep running into in every direction lately: the most consequential decisions in a system are usually the ones nobody remembers making.

So the question I'd leave open isn't whether we hit the limits of aviation technology. It's narrower and harder.

Of the things we stopped pursuing, how many did we actually choose to stop pursuing?