Scientists have mapped the human genome, detected gravitational waves, and landed rovers on Mars, but some of the most fundamental questions about reality remain wide open. From the composition of 95% of the universe to the biological roots of consciousness, here are the biggest problems science has yet to solve.
What 95% of the Universe Is Made Of
Everything you can see, touch, or detect with a telescope makes up roughly 5% of the universe. The remaining 95% is split between two mysterious components: dark matter (about 27%) and dark energy (about 68%). Dark matter doesn’t emit, absorb, or reflect light. We know it exists because its gravitational pull bends light from distant galaxies, creating distorted images called gravitational lensing. But when scientists point instruments at the spots where dark matter should be, nothing visible is there.
Dark energy is even stranger. It appears to be driving the universe’s expansion to accelerate, pushing galaxies apart at an increasing rate. Scientists have built enormous instruments to study these phenomena, including the James Webb Space Telescope, the South Pole Telescope, and the Dark Energy Spectroscopic Instrument in Arizona. Despite decades of searching, no one has directly detected a dark matter particle or identified what dark energy actually is. Understanding that hidden 95% remains one of the defining challenges of modern physics.
Uniting Gravity With Quantum Physics
Modern physics rests on two pillars that flatly contradict each other. Quantum theory governs the behavior of the smallest particles in the universe with extraordinary precision. General relativity, Einstein’s theory of gravity, explains how massive objects bend the fabric of space and time. Both theories work spectacularly well in their own domains, yet they are mathematically incompatible. This conflict has persisted for over a century.
The core tension is simple to state but fiendishly hard to resolve: gravity describes a smooth, curved fabric of spacetime, while quantum mechanics describes a world that is fundamentally grainy and probabilistic at the smallest scales. Should spacetime itself be broken into tiny quantum chunks? Should quantum theory be modified instead? Or is the answer something no one has imagined yet? The prevailing assumption has been that gravity must be “quantized” to fit within quantum theory, but no attempt to do so has produced a complete, testable framework. A successful unification would be the closest thing to a “Theory of Everything,” and it remains out of reach.
How Life Began From Non-Living Matter
Scientists can explain evolution in remarkable detail once life exists, but the first step, how non-living chemicals assembled into a living cell, is largely unknown. Two leading hypotheses compete for the site where this happened: hydrothermal vents on the deep ocean floor and warm shallow ponds on early Earth’s surface.
The warm-pond hypothesis has a specific advantage. These environments go through cycles of wetting and drying, which are known to promote the linking of small molecular building blocks into longer chains like RNA. Deep-sea vents lack those cycles. But both hypotheses face a deeper problem: where did the building blocks come from in the first place? Early Earth’s atmosphere was dominated by carbon dioxide, nitrogen, sulfur dioxide, and water vapor. Under those conditions, the famous spark-discharge experiments that simulate lightning striking a primordial atmosphere aren’t very efficient at producing the complex organic molecules needed for life. One proposed solution is that the raw ingredients were delivered from space, riding in on meteorites and interplanetary dust particles. Although scientists have produced simple RNA strands in highly controlled lab settings, replicating the messy, uncontrolled conditions of early Earth and getting the same result has not been achieved.
What Creates Conscious Experience
Neuroscience can map which brain regions activate during specific tasks, but explaining why any of that electrical and chemical activity produces the subjective feeling of “being you” is a different matter entirely. Scientists don’t even agree on a definition of consciousness, which makes building a unified theory extraordinarily difficult.
Several competing models exist. One proposes that consciousness emerges in a distributed way: each sensory region of the brain generates its own “micro-consciousness” for sight, sound, or touch, and these get assembled into a unified experience by higher-level brain areas. A second model argues the opposite, that consciousness only arises when information reaches a “global workspace” in the brain’s frontal and parietal regions, where it gets broadcast widely. A third focuses on the synchronized firing of neural networks across the brain’s sensory and association areas, suggesting that the timing of brain activity is the key ingredient. Each model can explain some experimental results but struggles with others, and none has emerged as a consensus explanation. The question of how physical matter gives rise to subjective experience, sometimes called the “hard problem” of consciousness, may be the deepest unsolved puzzle in all of science.
Why Cancer Metastasis Is So Hard to Stop
Cancer isn’t a single disease. It’s hundreds of different diseases sharing a common trait: uncontrolled cell growth. But the reason cancer kills is usually not the original tumor. It’s metastasis, the process by which cancer cells spread to new organs. This process remains one of the most confounding questions in oncology.
The challenge is biological complexity on a staggering scale. Metastatic cells don’t just float to a new organ and start growing. They actively adapt to their new environment, reprogramming their gene activity to match the tissue they’ve invaded. Liver metastases, for instance, shift their genetic instructions to mimic liver-specific patterns. Roughly 65% of metastases in colorectal cancer originate from independent subgroups of cells within the original tumor, meaning the cancer is essentially running multiple invasion campaigns simultaneously. The timing and destination of these seeding events are highly unpredictable, varying from patient to patient. This spatial and temporal chaos is why a treatment that works against one metastatic site can fail at another, even within the same person.
What Actually Causes Alzheimer’s Disease
For 25 years, the dominant theory of Alzheimer’s disease was straightforward: a sticky protein fragment called amyloid-beta accumulates in the brain, which triggers the buildup of a second protein called tau, which then kills neurons. This “amyloid cascade hypothesis” drove nearly all drug development. The problem is that drugs designed around it have largely failed.
Clinical trials tell a discouraging story. Active vaccines against amyloid-beta proved ineffective and caused serious side effects. Monoclonal antibodies designed to clear amyloid plaques succeeded in reducing plaque levels but did not clearly improve thinking or memory. Drugs intended to prevent amyloid from forming in the first place either failed or actually worsened cognition. Tau-targeting drugs have fared no better: trials of compounds designed to break up tau clumps produced negative results on cognitive improvement. This pattern of failure has forced scientists to question whether amyloid is truly the root cause or merely a bystander. The disease may be driven by something upstream of both proteins, but what that something is remains unclear.
Making Fusion Energy Practical
Nuclear fusion, the process that powers the sun, promises nearly limitless clean energy. In December 2022, scientists at the National Ignition Facility made history by achieving fusion ignition for the first time: they delivered 2.05 megajoules of laser energy to a tiny fuel target and got 3.15 megajoules of fusion energy back. That’s a genuine scientific milestone.
But scientific breakeven and practical power generation are very different things. The lasers themselves consumed far more energy than they delivered to the target, so the overall system was still a net energy consumer. Building a fusion power plant requires sustaining reactions continuously (not in brief pulses), capturing the energy efficiently, and doing it all at a cost competitive with other power sources. No facility on Earth can currently do this. The international ITER reactor in France aims to demonstrate sustained fusion at scale, but it has faced repeated delays and cost overruns. Fusion has been “30 years away” for decades, and while the physics is now proven, the engineering remains unsolved.
Reversing Biological Aging
Scientists have identified twelve distinct biological processes that drive aging: accumulating DNA damage, shortening of chromosome caps (telomeres), chemical changes that alter how genes are read, the buildup of misfolded proteins, declining cellular cleanup systems, faulty nutrient sensing, deteriorating energy-producing structures in cells, the accumulation of “zombie” cells that refuse to die, exhaustion of stem cell reserves, breakdowns in cell-to-cell communication, chronic low-grade inflammation, and imbalances in gut microbes.
In lab animals, intervening in some of these processes has extended lifespan or reversed specific markers of aging. But translating those results to humans has proven far harder. These twelve hallmarks don’t operate independently. They interact in complex feedback loops, so fixing one can destabilize another. The theoretical framework suggests it should be possible to decelerate, stop, or even reverse aging by targeting these hallmarks therapeutically, but no intervention has yet demonstrated comprehensive reversal in humans. The biology of aging is now well-catalogued. Controlling it is another matter.
Six Million-Dollar Math Problems
In 2000, the Clay Mathematics Institute designated seven problems as the Millennium Prize Problems, each carrying a $1 million reward. Only one has been solved: the Poincaré Conjecture, proven by Grigori Perelman in 2003 (he famously declined the prize money). The remaining six are still open:
- The Riemann Hypothesis: proposed in 1859, it predicts that all the “non-obvious” zeros of a particular mathematical function fall on a specific line. It governs the distribution of prime numbers, and its proof or disproof would reshape number theory.
- P vs NP: asks whether every problem whose solution can be quickly verified can also be quickly solved. A resolution would have enormous implications for cryptography and computer science.
- Navier-Stokes Equation: the equations governing fluid flow are used constantly in engineering, but no one has proven whether smooth solutions always exist in three dimensions.
- Yang-Mills and the Mass Gap: concerns the mathematical foundations of the theory describing subatomic forces.
- Hodge Conjecture: a deep question about the relationship between geometry and algebra.
- Birch and Swinnerton-Dyer Conjecture: relates to the behavior of equations that define elliptic curves.
These aren’t abstract puzzles. Solutions would ripple through physics, computer science, engineering, and cryptography in ways that are hard to predict.
Pulling Carbon Back Out of the Air
Even if the world stopped all emissions tomorrow, the carbon dioxide already in the atmosphere would continue warming the planet for decades. Direct air capture technology, which chemically filters CO2 from ambient air, is one proposed solution. The problem is cost and scale.
Current projections from MIT suggest that by 2050, the two leading approaches could reach costs between $100 and $400 per ton of CO2 removed, depending on the method. Today’s costs are significantly higher. The fundamental barrier is thermodynamic: CO2 makes up only about 0.04% of the atmosphere, so extracting it requires processing enormous volumes of air and breaking strong chemical bonds, both of which demand substantial energy. Scaling the technology to remove billions of tons per year, the level climate models say is needed, would require a buildout of capture facilities, clean energy sources, and CO2 storage infrastructure that doesn’t yet exist. The chemistry works. Making it cheap enough and big enough to matter is the unsolved problem.
Predicting Proteins in Motion
AI tools like AlphaFold have revolutionized biology by predicting the three-dimensional shapes of proteins with remarkable accuracy. But proteins aren’t frozen sculptures. They flex, twist, and shift between different shapes depending on what molecules are nearby, and current prediction tools largely miss this dynamic behavior.
AlphaFold 3 predicts static snapshots, the kind of structures stored in databases, but it cannot reliably capture how a protein moves in the fluid environment of a living cell. In some cases, it predicts the wrong shape entirely. Certain enzymes naturally adopt an open configuration when no partner molecule is present and only close when one binds, but AlphaFold 3 predicts the closed shape regardless. It also struggles with “orphan” proteins that lack close relatives in existing databases, since it relies on comparing related sequences to make predictions. And when modeling complexes with multiple identical protein chains, it occasionally predicts physically impossible structures where entire chains overlap. These limitations matter because a protein’s function depends not just on its shape but on how that shape changes over time.