The better mousetrap fallacy

The better mousetrap fallacy, as Lubershane defines it, is the assumption that some new invention is the thing that will unlock a market, applied to a market where invention is not what is missing. He is careful about the boundary, and the care is the point. He refuses to re-run the old innovation-versus-deployment argument, says this is plainly a both-and, and insists there are many problems on the route to net zero where we do not have anything close to the right technology and need not just better mousetraps but entirely different kinds of trap. His narrower worry has two parts, and he says explicitly that he is not much troubled by the first: venture investors putting money into new technology can look after themselves. What bothers him is the second, that a proliferation of options is close to paralyzing for the people who would otherwise deploy. Working with utilities and large energy infrastructure operators, he sees analysis paralysis among buyers weighing novel storage, unable to pick a lane and get comfortable with it because there are so many on offer.

The mechanism behind that judgment comes from solar. By the 2000s and 2010s the winner was already the incumbent, crystalline silicon, and the enormous cost decline that followed came from scaling it over and over rather than from any of the hundred or so startups then pursuing thin film, cadmium telluride, amorphous silicon and other cells. Lithium-ion repeated the pattern. Innovation did not stop, but it moved inside the scaling: Lubershane estimates that 95% of the time the improvement is a tweak or iteration on the existing formula rather than a new chemistry, and every new factory teaches the industry something. China is extremely good at exactly this kind of incrementalism, which he offers as part of the reason the world became dependent on it, while the American taste for novelty, boldness and restlessness is a genuine strength that becomes a blind spot when the task is scaling. That produces the threshold. Kann puts it first, saying that 20% better than the incumbent is not better enough because the industry will pass you within five or ten years, so a challenger has to be five or ten times better; Lubershane agrees and adds the harder part, which is that when the incumbent’s improvement is happening mostly inside Chinese factories you cannot see clearly how fast its cost is falling even two or three years out. To bet against a curve you cannot observe, the bet has to be enormous.

Storage is where he draws the line, and where he says the fallacy has taken hold: for more than a decade a near-continuous stream of approaches in the 20% to 50% better-than-lithium-ion range has been crushed, repeatedly. The test he applies is not whether a technology is ahead today but whether it sits outside anything lithium-ion could plausibly reach. Form Energy’s multi-day system qualifies in his view because its target cost is beyond any reasonable vision of lithium-ion’s future, which makes it a different product rather than a competitor, and Rondo’s thermal storage qualifies by doing something electrochemical storage cannot do at all. Both are his own firm’s investments, and both cost claims are targets for products still commercializing. Sodium-ion is the case he explicitly cannot place: a real deviation from lithium-ion, but a close cousin in manufacturing and principle, and already being scaled by the same large battery manufacturers, so it is a better mousetrap of a kind startups are poorly positioned to build. Lithium iron phosphate he treats as plainly incremental. Nuclear is his canonical example, and there the argument is not about atoms at all. Reactors built from the late 1950s through the 1970s, before computer modeling and modern engineering tools, are running well today, and nuclear is still deployed cost-effectively in Korea and China. On that reading the problem was always public opinion, regulation and policy, so incremental Gen 3+ designs are all that is needed, while dozens of developers pursue clever new coolants and passive safety features. Kann sharpens it into the learning curve: the first plants will be expensive whatever the technology, so if a suddenly permissive regulator licensed twenty designs and one of each got built, nobody descends a curve at all. Lubershane’s hypothetical is to pick one to three designs and build ten of each, spaced a year to eighteen months apart, which he says would beat the vast majority of next-generation designs on cost. He catches himself mid-claim there, having first said it would beat any of them, and softens it. He is also sympathetic to the safety pitch in principle, since inherent safety means less redundant protective equipment and therefore a cheaper plant, but suspects it is a red herring, because arguing that today’s fleet is insufficiently safe is both untrue and bad for the industry. What he does not say is that Gen 4 research should stop. His claim is about where commercialization capital and government policy should go.

Direct air capture is where his own confidence drops and where Kann breaks with him. Lubershane raises it saying he does not know whether they will agree or how much he believes it himself. There are over 150 direct air capture startups on a list he considers credible, so there is no shortage of traps; the shortage is mice, meaning buyers willing to pay what the technology costs. Essentially one large buyer, Microsoft, plus a long tail. So the race may be ten years early, though he allows that a decade spent proving out a few is not necessarily wasted. Kann concedes two of the three points: there are probably more carbon removal startups than buyers at meaningful scale, and since almost none have built anything, buyers are forward-contracting on promises and have to guess what is real. But he disagrees on the conclusion, because nobody yet knows whether any approach gets cheap enough, and he wants a lower benchmark than the customary hundred dollars a ton. So he wants more mousetraps in direct air capture, not fewer, even expecting a painful shakeout, and Lubershane accepts the point. That leaves the question of who does the sorting, which neither answers cleanly, and a market force that may do it for them: both think the end of a fifteen-year era of near-zero interest rates will filter ideas earlier. Lubershane calls that a net positive at this stage, conditional on the surviving capital going to areas where no incumbent exists to be incrementally improved. Kann notes it cuts both ways, because earlier filtering also means fewer shots on goal at the biggest problems.