Fireside Chat: Early Adopters of Quantum Technology — Annotated Lecture Notes & Key Takeaways

Here's the setup you missed. It's a post-lunch fireside chat at what appears to be a quantum conference at MIT, hosted by a moderator named Danna, with one guest: Eric Lander, the geneticist who led the Human Genome Project and founded the Broad Institute. The joke that opens the whole thing is that Lander knows essentially nothing about quantum. When that gets thrown at him, his comeback is "I can spell quantum," and the room loses it. There's a bit about how no political jokes are allowed, and then they get down to it.

The session is built on a deliberate mismatch. The moderator's premise is that the quantum field right now runs on early adopters​ — the technology is only just starting to be good enough to matter, and the people who figure out what to do with it are going to shape the whole thing. So instead of asking a quantum person about quantum, she asks someone who ran a completely different big-science-and-technology story whether any of it transfers. The honest answer, delivered repeatedly across the half hour, is: maybe, and I genuinely don't know.

That hedge is worth sitting with, because it's the most intellectually honest thing in the room.
 

The Genome Project Was Organized Around the Question, Not the Machine

Lander's first substantive move is to push back on the framing. Both quantum and genomics qualify as early-adopter situations, he says, but he's not convinced they line up neatly. Then he tells the story.

Sequencing in the late 1970s was, by his account, miserable work — plastic shields, radioactive reagents, gels, X-ray film laid over the results. What changed wasn't just automation. What changed is that the field stopped treating the machine as the point and started treating a piece of information as the point. The goal was the sequence of a human genome, full stop, and the technology got dragged along behind that goal.

The numbers he uses to make this concrete are the part I'd remember if I were taking an exam. Contemporaries estimated the human genome would cost around three billion dollars. Lander's view is that if anyone had actually done the arithmetic honestly at the time, the real figure was more like thirty billion. Today a genome runs around three hundred dollars, and the record for a complete sequence is under four hours. He offers that compression as evidence that a lot can happen when a field is organized the right way — though note he's describing correlation, not proving causation, and he doesn't try to.

The mechanism he keeps returning to is a loop: people pushing the technology forward, and people inventing new things to do with it, in roughly equal numbers and roughly equal status, each side feeding the other. His institution worked, on his telling, because neither side got to dominate.

Then comes the detail I found most useful. MIT didn't build the automated sequencers. Companies did. But Lander's group became, in his words, far more sophisticated users of those machines than the companies that made them. They knew how to break them. They knew how to find the failure modes. Vendors kept coming back precisely because this group would stress-test hardware harder than the vendor's own staff would, and would then invent applications nobody had thought of. When a company showed up with a new instrument, the conversation started with: what is this actually good for?

The Solexa story is the concrete illustration, and it's the best anecdote in the session. A small company in England — later absorbed into Illumina — came to show off their read technology. Lander's side told them the reads had to work from both ends. The company said that wasn't possible. Lander's side said it was necessary. He describes walking around Cambridge repeating himself for about two months, until the company called back and conceded that yes, they could read from both ends after all.

The generalizable claim buried in that story: you need a community where the people who understand what a technology could enable have enough standing to make demands of the people who build it. In the genome case that meant physicians, engineers, mathematicians, computer scientists, chemists, the occasional physicist, and — his aside — some astrophysicists who had effectively taken refuge in the project.
 

How You Get People Who Can See What Isn't There Yet

The moderator asks the obvious follow-up: fine, but how do you find those people and make sure anyone listens to them? Lander's one-word answer is young people, and he tells a story from 1989 to back it up.

Early-morning meetings on the seventh floor of the Whitehead Institute, pre-Genome-Center. Wally Gilbert, who had a Nobel Prize for inventing DNA sequencing, showed up daily. So did a junior technician from Lander's lab named Alex Weaver. At some point Gilbert held forth on something, and Weaver told him flatly that he was wrong. Gilbert thought about it, said he was right, and sat down. Lander says that's the moment he knew the thing was going to work.

His interpretation is sharper than the anecdote. The reason a junior technician could see what a Nobel laureate couldn't is that she hadn't yet absorbed the field's consensus about what was and wasn't possible. The established molecular biologists and human geneticists of the late 1980s had, in his phrasing, too clear a vision of the limits. Expertise was the obstacle. So the strategy was to recruit a generation that hadn't learned the limits, protect them from learning the limits, give them resources, and let them do things that looked crazy.

It wasn't pure anarchy. Sequence production was tightly run and accountable. But even there, the people grinding out output on the production side rotated into the development group, so that for a month at a time their job was to ask how the process could be made better. Structural rotation, not just vibes.

This is where he locates MIT's specific contribution, and it's a genuinely interesting division of labor. Google and IBM, he says, are going to build the enormous machines; a university won't compete on that. What a university can do is generate the unreasonable ideas that make those machines worth having. He's careful to frame it as mutual respect rather than academia lording it over industry, and he explicitly thanks the people in the room who are building hardware.

The story that lands the point is about a defense contractor — EG&G, Doc Edgerton's company — approaching him at the very start of the Genome Project and offering to manage the grant. Lander asked what they knew about genomics. The answer was essentially: nothing, but we're very good at managing things. He declined, on the grounds that the project needed to find its way through unmapped territory, which is not the same skill as management. He notes that this turned out to be the right call, though of course it would have been the right call in his telling either way.
 

Taste, Arithmetic, and Why You Let Bad Ideas Die on Their Own

About sixteen minutes in, the conversation turns to a question about technological maturity, and it's framed through an unexpectedly funny bit of crowd work. The moderator points out that "thermometer" covers everything from the device you use on a sick child to the instrument a colleague named Paola would use to measure temperature in a spin system — at which point someone in the audience offers nanokelvins, and there's a round of appreciative noises. The actual question: given that range of maturity, how do you decide whether a technology is ready for co-design?

Lander gives two criteria. First, is there a genuinely good problem — he's problem-driven, not capability-driven. Second, how many orders of magnitude separate you from solving it. His group's rule of thumb was that if you couldn't see a path to producing even one percent of the thing you ultimately wanted, the project didn't make sense to start.

He pairs that with a cautionary tale about hype. Right after the human genome was sequenced, the press announced that the human proteome was next. Lander says nobody in the field knew where that came from; the tools simply weren't there. The collective decision was to ignore the idea for roughly a decade, and he thinks that was correct. His warning is that we live in an especially hype-prone moment, and the discipline is being able to say: here's an extraordinarily valuable use, we're two or three orders of magnitude away, and therefore here are the interesting sub-problems available inside that gap.

Then he says something that's either profound or a dodge, and I'm still not sure which: it comes down to taste, and you need people with taste. He elaborates just enough to make it non-empty. Most projects you can imagine are bad ideas. They sound good, and when you actually examine them they turn out not to sit at the intersection of practical and interesting. So the job is killing most of them.

But then he immediately complicates it, and this was my favorite turn in the whole talk. You don't kill projects. You nurture the good ones and you let the ones that deserve to die die on their own. When someone brings you an idea you think is weak, you ask a couple of probing questions and send them off. When someone brings you a good one, you ask what it would take to pilot it, or ask them to write it up. You water it, and you get them money fast.

His reasoning for the gentleness is not sentimentality. If you tell someone their idea is terrible and drive a stake through its heart, you might be wrong — and you've now taught the room that the place is closed to ideas. Encouragement is cheap insurance against your own fallibility.

Which leads, fairly inevitably, to money. For whatever MIT Quantum becomes, he argues for a pool of unrestricted discretionary money that can be spent on bets. Grant funding arrives earmarked; you can't pivot with it. He compares it to venture capital, where you're betting on ideas and on teams, usually on some combination of both, and he's blunt that you will lose some of those bets — the point is that you win more than you lose. There's a running gag through this section about not calling it a slush fund, and the moderator eventually settles on "discretionary, unrestricted." He extends the point to institutions generally: donor money, industrial support, and federal support combined is a more powerful configuration than any single stream.
 

Platforms, Walls, and Turning Junior Faculty Into Gorillas

The back third is about institutional design, and it's where the talk gets most concrete about what "building an institute" actually means.
The moderator asks what justified creating a Broad-type organization rather than just coordinating across existing departments. Lander's answer is a jab at the fashionable idea of centers without walls: when it rains, you get wet. You sometimes need anchors and magnets — real things that pull people into the same room.

The specific mechanism he names is the platform. At the time, calling it that was unusual; he thinks the term has since spread. A platform means industrial-grade technology run by serious, experienced people, because there are tasks where you don't want a graduate student learning on the job — you want someone who understands process. Critical mass, he argues, does not emerge from scattered individual labs. It emerges when a shared project gives people a reason to converge.

The example: human geneticists around Boston all wanted to map genes for common diseases, and every one of them would have had to independently invent the same underlying platform, which no single lab could do. Pooling it meant they could build that platform, and then the next one, and the next.

Then the Brad Bernstein story. A young Harvard professor in the Broad orbit was applying for a grant to set up an epigenomics center and assumed he had no chance against the established heavyweights. He wrote the application with the shared platform standing behind him. He got the top score. He came into Lander's office and admitted, sheepishly, that he supposed he was one of the heavyweights now. Lander's framing: junior faculty can punch far above their weight when there's serious infrastructure behind them, and universities are among the few institutions that can arrange that. He calls it not just a superpower of MIT's but a responsibility.

The moderator then presses on the obvious tension. If young people are the source of the good ideas, how do you square that with a tenure clock that lets someone fail for five years but not seven? Lander's concession is that older people still have jobs, and a reminder that he counted as young when he started the Genome Project. His answer isn't "leave them alone." It's an example: three junior faculty — an immunologist, a cancer biologist, and a metabolism researcher — all wanted to use a then-new technique that required producing an enormous library of reagents. Individually, impossible. They pooled roughly forty percent of their time into building the shared resource, and each then published pathbreaking work in their own field that they could not have produced alone.

The lesson he draws is that encouragement without structure is insufficient. You want bold ideas from young people, and then you help them figure out how the idea could actually get done in a way that produces results on short, medium, and long timescales.

That gets him to the last big argument, which he makes twice from different angles. Wally Gilbert's original proposal for the Genome Project was to build one enormous center that would sequence the whole genome in one push. Lander calls it a terrible idea, and notes Gilbert agrees: the only output was a very long wait followed by a single result. What replaced it was a committee that chopped the project into staged deliverables — a genetic map usable for mapping disease genes within a few years, then more, then more. Each intermediate output was transformative on its own, even though the full sequence was still a decade out.

And the comparison he can't resist, addressed to the physicists in the room: the Genome Project and the Superconducting Super Collider started in roughly the same year. His line is that half a superconducting super collider neither superconducts nor super collides, and what the country got for the money was half a very large hole in Waxahachie, Texas. The Genome Project survived because there were usable deliverables the whole way through — people could say the organizers were out of their minds and still be using their output. He dislikes the phrase "quick wins," preferring important wins spread across different timescales: a layered portfolio.
 

Where I Came Away Unconvinced

The weakest link is the one the speaker himself keeps pointing at: nobody in the session actually demonstrates that the genomics playbook transfers to quantum. Lander is candid that he doesn't know how far you can take a current quantum device apart and tinker with it the way his people took sequencers apart and told manufacturers what to fix. He says he won't underestimate young people on this, which is gracious but is a hope, not an argument. If you came for quantum-specific guidance, you leave with heuristics about organizing research, not answers about quantum.

I'd also flag the claim about why governments fund quantum. Lander states that the ability to perform computations that break accumulated encrypted traffic is the real driver of massive public investment, and he attributes this to knowledge from inside and outside government. It may well be right, but no evidence was offered, and it's a strong claim delivered in a room full of people whose funding depends on the answer. Related: he says Shor's algorithm unlocked tens of billions of dollars and doubts it will be the last interesting thing to run on a quantum machine, which is the most quantum-specific substantive point anyone makes, and it gets about ninety seconds.
The "taste" criterion bothered me more than any of it. It's doing enormous work in his framework — it's what separates the projects worth funding from the ones that should quietly die — and it's defined almost entirely by outcomes. Someone with taste is someone who was right. That's not operational; there's no way to identify taste in advance, which is precisely when you'd need it. The one-percent rule and the two-to-three-orders-of-magnitude range are better, but they're also stated as intuitions without a method behind them, and I'd have liked a worked example of a project that failed the one-percent test.

Two smaller gripes. The Superconducting Super Collider comparison is hindsight-friendly: big projects get cancelled for budgetary and political reasons that have nothing to do with deliverable structure, and the Genome Project faced its own near-death moments that went unmentioned. And the whole narrative is survivorship in its purest form — this is a winner explaining why winning was wise. That doesn't make it wrong, but it does mean the confidence of the prescriptions is running ahead of the evidence.

Still, the reason to have watched it is the middle: the paired-end story, the technician telling the Nobel laureate he was full of it, the junior faculty who spent forty percent of their time on shared infrastructure and came out ahead. Those are concrete descriptions of how a field organized itself around uses instead of machines, and they're the reason the session is worth your time even though the quantum mapping never quite lands. The moderator closes by saying this isn't the last conversation, and Lander, noting he works right across the way, agrees — he isn't sure which of what he said will turn out to be useful or parallel, but he thinks the early adopters​ in that room are at an incredibly exciting moment. For a talk whose recurring theme is that nobody knows yet, ending on "let's keep talking about it" felt exactly right.