Physics' biggest trouble isn't that two theories contradict each other — it's that they never contradict each other in front of any instrument humans can build.
The theory of matter and the theory of gravity are each absurdly precise: one gets the electron's magnetism right to a dozen decimal places, the other keeps satellite navigation accurate to the metre. Put them together and they only truly clash in two places — the centre of a black hole, and the first instant of the universe — and neither is anywhere we can go. So the predicament is not that the theories were refuted; it is that no experiment can refute them. What a field runs on when the data dries up, and what mistakes it makes then, is worth seeing clearly in its own right.
Why Gravity Refuses to Be Quantised The Quantum Gravity Gap
quantum gravity · the edge of an effective theory
Intuition
Quantising the electromagnetic force threw up infinities too, but those infinities could be absorbed into a handful of constants — measure them once and you are done forever. Gravity won't cooperate: its coupling strength itself carries energy, so every additional layer of correction spawns another new constant that only experiment can supply. The name for this is non-renormalisable: to absorb the infinities you would need infinitely many constants.
Mechanism
Expand the result of a single gravitational scattering into a series:
A = c0 + c1(E/EP)2 + c2(E/EP)4 + ⋯
A is the amplitude of that scattering (roughly, how likely it is to happen); E is the energy involved; EP is the Planck energy, about 1.2×1019 GeV; each cn (the subscript n labels which term) is a constant the theory cannot compute and only experiment can fix. Everything hinges on the ratio in the brackets: in an accelerator E is a dozen orders of magnitude below EP, so squaring it leaves nothing and only the first term matters; but as E approaches EP, every term counts equally.
That dividing line is not chosen by hand. There is exactly one way to combine the quantum constant (ħ, read "h bar"), gravity (G) and the speed of light (c) into a length:
ℓP = √(ħG/c3) ≈ 1.6×10−35 metres
That −35 is an exponent, not a count of decimal places: twenty orders of magnitude smaller than a proton. At that scale the very assumption that "spacetime is a smooth sheet" ought to break down.
Both theories are tested over and over on the left; they only clash at the far right. The difficulty is that the stretch in between holds no experimental point at all.
The counterintuitive part
"We have no quantum gravity" is too blunt. The low-energy part has existed for decades; what is missing is the region near the Planck scale. This is the same situation as Newtonian mechanics — not wrong, just bounded — except that this time the boundary happens to sit where humans cannot reach.
Fluid engineering: the Navier–Stokes equations treat a fluid as a continuum and are superb on an aircraft wing; once a gas is rarefied enough that the molecular mean free path is comparable to the device, you must switch to kinetic theory. The equations aren't wrong — they were used past their range.
Machine learning: scaling laws are startlingly accurate across the compute range they were fitted on, and extrapolating them several orders of magnitude is never safe — precise inside the range, silent outside it.
In one line: gravity isn't un-quantisable; its quantum theory is simply right everywhere humans can measure.
Think it through: if the two theories only clash at the Planck scale, why not just ignore it?
Because two places refuse to be ignored: the centre of a black hole and the very early universe genuinely reach those energy densities, and both left consequences we can see. More pressingly, forcing the two theories together produces outright contradictions — and a contradiction is the theory itself throwing an error.
Two Routes, Each With Its Own Bill Strings vs. Loops
string theory · loop quantum gravity
Intuition
The trouble starts with the "point": a particle has no size, two points can approach arbitrarily closely, and the interaction grows arbitrarily strong. String theory swaps the point for a short vibrating string — it has length, so nothing can get closer than that, and different modes of vibration are simply different particles. Loop quantum gravity says the trouble is spacetime instead: chop space into pieces of finite size, and "arbitrarily close" never arises.
Mechanism
String theory's bill is written into its consistency conditions: for the quantum theory of a string to behave, spacetime must have 10 dimensions. The extra ones have to be curled up into tiny shapes, and the number of ways to curl them is astronomical (a widely quoted estimate is 10500), each giving a different low-energy physics — which makes "predict the particle list of our universe" desperately hard. What it buys in exchange is genuinely beautiful: for certain black holes the entropy can be counted state by state, and the count matches Hawking's formula exactly.
Loop quantum gravity pays at the other end. The areas it computes can only take a discrete set of values:
A = 8πγℓP2 Σ √(j(j+1))
A is the area of a small patch; ℓP is the Planck length; γ is the Greek letter gamma, a number the theory cannot currently pin down; j is the label carried by each line piercing that patch, restricted to values like ½, 1, 1½; Σ is sigma, meaning "add up every line that pierces it". Area therefore comes in a smallest portion. The bill arrives in the opposite direction: how to grow smooth spacetime back out of these pieces, and how to fit the Standard Model in, still has no good answer.
One route rebuilds the matter (points become strings), the other rebuilds the stage (continuous spacetime becomes a network). Both are self-consistent, and no experiment can call the winner.
The counterintuitive part
This gets described as "two candidate answers waiting for experiment to choose". The honest version is that neither is yet a complete answer. Nor does it resemble the old wave-versus-particle dispute about light: both of those camps could make quantitative predictions about the same experiments, whereas these two struggle to write down even one prediction that would tell them apart.
Genetics: in the early 1900s the Mendelians (traits set by discrete factors) and the biometricians (traits varying continuously) each had data on their side and deadlocked for nearly two decades. What ended it was not argument but Fisher's 1918 proof that many discrete factors superposed give exactly a continuous distribution.
Artificial intelligence: symbolic and connectionist programmes ran side by side for three decades, and what settled it was not whose argument was prettier but measured results once compute and data became available.
In one line: not two answers awaiting an experiment, but two half-finished ones awaiting an experiment nobody can build.
Think it through: string theory needs 10 dimensions and we only see 4 — isn't it already falsified?
No — but don't rush to praise it either. Dimensions curled up small enough really are invisible at low energy; Kaluza and Klein used the trick back in the 1920s. The problem is that this consistency is too cheap: there are so many ways to curl that almost any low-energy physics can be arranged, so "agrees with observation" stops being a constraint.
The Numbers That Don't Add Up Cosmological Constant & Friends
cosmological constant · hierarchy problem · a shortlist
Intuition
A second kind of trouble is more unsettling: the theory can compute, and the number comes out wildly wrong. Quantum field theory says the vacuum is not empty — every field is fluctuating; add those fluctuations up to the Planck scale and the energy density you get is about 10120 times the dark energy actually observed.
Mechanism
These problems share a shape: an observable is a sum and difference of enormous terms, yet the result is minuscule — meaning those terms must cancel digit by digit for over a hundred digits.
naive theoretical estimateobserved value ≈ 10120
The 120 here is an exponent: written out, that is a 1 followed by 120 digits 0 (zero). Physicists refuse such a coincidence not because it is "ugly" — everywhere else, exact cancellation of this kind has a symmetry behind it (the photon is strictly massless, guaranteed by gauge symmetry). Failing to find that symmetry amounts to admitting a mechanism is missing.
The second instance is the hierarchy problem: the Higgs boson weighs 125 GeV, while quantum corrections naturally drag it toward 1019 GeV. Supersymmetry was the prettiest explanation — every particle gets a partner whose corrections carry the opposite sign and cancel. The collider has looked for over a decade and not one partner has turned up.
The observed value hugs the far left, the theoretical estimate the far right, and no known mechanism pulls the two together.
Mechanism · the rest of the shortlist
Other holes stand on their own: what particle dark matter is, and whether dark energy is a constant (see "Dark Matter & Dark Energy"); why neutrinos have mass at all and whether they are their own antiparticles; why there is slightly more matter than antimatter, by just enough to build galaxies; the strong CP problem — the strong force is allowed a parameter that would violate time-reversal symmetry, yet experiment pins it below 10−10; and the unsettled account in "The Measurement Problem & Interpretations".
The counterintuitive part
None of these are problems nobody has got around to computing. Most of them have answers; the answers just disagree with observation. That is worth more than being stuck — a wildly wrong prediction carries information, pointing at which assumption must go. The difficulty is that the candidate replacements are many and the data is thin.
Numerical computing: subtracting two nearly equal large numbers in floating point eats your significant digits — catastrophic cancellation. The standard fix is not more precision but rewriting the formula so the small quantity appears directly. A physicist's instinct is identical: the arithmetic isn't imprecise, the expression is wrong.
Statistical inference: Wald studied the bullet holes on returning wartime aircraft and pointed out that the places to armour were the ones without holes — planes hit there never came back. The other kind of answer to "why is it just right" is not a mechanism but selection bias; its costs are laid out in "Fine-Tuning & the Multiverse".
In one line: the most valuable open problems aren't the incalculable ones — they're the ones we calculated and got neatly, enormously wrong.
Think it through: why won't physicists accept "the constant is just that big, no reason"?
That stance has lost repeatedly. The regularities of the periodic table were once "just how it is", and later came from electron shells; the hadron mass spectrum was once accident, and later came from quarks. Every time "no reason" is accepted, a line of questioning closes.
Doing Science When There Is No Experiment Doing Science Without Data
methodology · an honest audit
Intuition
When the decisive experiment cannot be done, a field has to pick theories by other criteria: consistency, simplicity, and a far vaguer word — beauty. Physics has a glorious record of beauty paying off (symmetry intuition led all the way to gauge theory), and so beauty was quietly promoted into a predictive tool. Hossenfelder's Lost in Math tallies the bill unflatteringly: the concrete predictions this taste produced over the past forty years have almost uniformly failed.
Mechanism · which principle actually broke
The hardest hit is naturalness: a dimensionless parameter shouldn't sit far from 1 unless a symmetry protects it. The rule once correctly bracketed the charm quark's mass; but the naturalness problem of the Higgs mass has waited decades and received only a collider narrowing the room for new particles, round after round. The conclusion can only be that naturalness is a rule of thumb, not a law, and deserves less weight. Conversely, swinging "unfalsifiable" as a club is too easy — one string-theory by-product, the gauge/gravity correspondence, is used to estimate the viscosity of the quark–gluon plasma, and that really does meet data. The risk isn't falsifiability; it is funding and attention staying locked on a few programmes.
Mechanism · the real progress of the same period
Reporting only the criticism would be its own dishonesty. Decisive results kept arriving: the 2017 neutron-star merger delivered gravitational waves and light together, pinning the difference between their speeds below 10−15 and eliminating a whole class of modified-gravity theories at a stroke; pulsar timing arrays found evidence in 2023 for a nanohertz gravitational-wave background; and the muon's anomalous magnetic moment, long the hottest hint of new physics, saw its gap to experiment shrink sharply once the theory side moved to lattice methods in 2025 — a vanishing clue is progress too.
The counterintuitive part
"Physics has stalled" is a lazy verdict. What stalled is only the deepest layer; over the same period condensed matter, quantum information and observational cosmology all raced ahead. The contrast is the thing worth noticing: progress is fastest exactly where the data is densest.
Cross-disciplinary reading · clinical research / experimental method
Clinical research: to stop stories being written after the fact, medicine uses pre-registration — hypotheses and analysis plans are registered publicly before the trial and cannot be changed afterwards; the ICMJE has required registration as a condition of publication since 2005.
Experimental method: particle physics and gravitational-wave experiments use blind analysis, forbidding anyone to look at the real result until the procedure is fixed. LIGO goes further: a tiny group secretly injects fake signals, and in 2010 the whole collaboration analysed one as usual and was told it was a drill only after the paper was written. Writing "I may not see what I want to see" into the procedure is precisely the antidote when taste runs loose.
In one line: with no data, the danger isn't being wrong — it's treating "this is beautiful" as evidence.
Think it through: if a theory yields no testable prediction for decades, should it keep getting funded?
Ask a better question: not "is it worth it" but "is the portfolio sane". A few people betting long on a high-risk programme is healthy; what is unhealthy is one programme holding the dominant share of posts, referee seats and grants, so alternatives cannot recruit. That is a resource-allocation question.
Going Deeper
If a Planck-scale accelerator is impossible, is quantum gravity forever untestable?
Not necessarily, though every window is narrow. One is using the universe as the accelerator: inflation may have run near 1016 GeV, so a specific pattern left by primordial gravitational waves in the polarisation of the cosmic microwave background would be an indirect glimpse. Two is accumulated effects: some models predict a minuscule dependence of light's speed on photon energy, and comparing arrival times of different-energy photons from gamma-ray bursts billions of light years away already tests this, pushing such models up near the Planck scale. Three is tabletop: put two milligram-scale masses each in superposition, let them interact only gravitationally, and if entanglement appears the gravitational field itself must be quantum — a classical mediator cannot create entanglement. What that needs is not energy but isolation from vibration and long coherence.
Why did the old rule of "naturalness" fail in this era?
Naturalness says a dimensionless parameter should be a number around 1, and anything extreme implies a symmetry or a deeper mechanism behind it. Three explanations for this failure are on the table and none can be decided yet: new physics does exist, just at energies out of reach; naturalness was only ever a heuristic that got mistaken for a law of nature; or a selection effect is at work — only values that happen to permit structure have anyone present to ask the question. That last one is expensive: it weakens the point of "explaining a constant" at all.
Which items on this list are most likely to be crossed off within a decade?
The best prospects are the ones with a dedicated experiment already running: the neutrino mass ordering, whether neutrinos are their own antiparticles (neutrinoless double beta decay experiments tighten the limit yearly), and the fate of dark matter in its classic parameter space (direct detection is closing on the neutrino floor, a natural limit). The hopeless ones are a complete quantum gravity and the measurement problem: the first lacks energy, the second lacks an experiment that distinguishes the interpretations. Whether a puzzle will fall soon depends less on how important it is than on whether an instrument is pointed at it right now.