The future of science is not just about finding answers to questions we already know to ask—questions about life and the universe that current theories may raise. The history of science shows that it commonly depends even more on uncovering entirely new, fundamental questions and answers that largely new tools reveal—and we did not yet even imagine asking. New tools discover new questions: the microscope made it possible to discover cells and bacteria and ask fundamental questions about what they and life are actually made of. The telescope raised transformative questions about the structure of the universe, the motion of planets and the existence of galaxies beyond our own. X-ray imaging allowed us to look inside the body without surgery and led to entirely new questions about internal anatomy and disease diagnosis. The cathode ray tube (particle detector) enabled uncovering the electron and triggered new questions about what matter and reality are made of at the smallest scales, paving the way for particle physics. Chromatography enabled us to ask how complex mixtures can be separated and analysed, revolutionising chemistry, pharmacology and environmental science. These inventions and their discoveries provided new pillars for modern biology, astronomy, physics and chemistry. These major tools and methods reflect a large methodological leap and an entirely new type of invention—not just an incremental upgrade or combination of tools. Such major leaps range widely beyond those tools: from the first thermometer and Geiger counter, to the first electrophoresis method and the PCR method. These groundbreaking inventions—each sparking multiple discoveries—emerged primarily from practical experimentation rather than a guiding theory. They were born from iterative trial and error rather than theoretical prediction—as we explored in Chapter 6. They demonstrate how practical innovation can precede theoretical understanding, with explanations often developing after the tools reveal what we did not even know we needed to explain. This brings us back to the four main forces driving tool innovation—and so driving discovery itself—that we uncovered examining science's over 750 major discoveries: • We invent entirely new tools—from the first telescope to the PCR method. • We upgrade tools in new ways—from higher-resolution telescopes to faster DNA sequencing tools. • We combine tools in new ways—from MRI scanners coupled with advanced statistical techniques, to CRISPR gene editing with machine learning methods. • We adopt new tools from other fields —using the RCT method from medicine in economics, or machine learning methods from computer science in everything from chemistry to climate modelling (Chapters 1 and 6). We can think of these four innovation pathways like this: first come these largeleap tools (first-wave tools)—like chromatography itself. Once invented, they can be vastly improved into upgraded tools —like gas chromatography providing faster, higher-resolution separations. They can be combined into hybrid tools —like liquid chromatography-mass spectrometry that merges separation with precise molecular identification. And they can also be adopted into other fields as transfer tools —like chromatography moving from chemistry into biology and environmental science, tracking everything from pollutants to DNA fragments. Large-leap and transfer tools are first-generation innovations; upgraded and hybrid tools are secondand later-generation innovations. This pattern—a big methodological leap followed by layering invention on invention—is largely how science digs deeper, layer by layer, into reality through entirely new ways of seeing, measuring and understanding that are impossible without them. Today's lower-hanging-fruit discoveries are easier to reach by borrowing methods from other fields and recombining existing methods, while upgrading and especially inventing methods can often demand more time and resources. Yet until now, we have not had a general understanding for how we systematically extend the frontiers of science. We push back the goalposts and bounds of science by upgrading our toolbox through these innovative pathways. Other strategies—from scaling to automating tools—fall into one of these four categories. Ultimately, it is crucial to identify the blind spots, biases and assumptions built into our methods—and design ways to address them (as we mapped out in Chapter6). Taking a step back, a crucial lesson emerges here. Most scientists generally focus on experimental and theoretical research—often without thinking about how they could improve and refine their own methods to reveal new kinds of findings. But that is exactly the spark that pushes science into new terrain: finding ways to explore and experiment in completely new ways. Yet at first glance, it can seem overwhelming to imagine methods or tools better than the ones we rely on every day. But the first step is simply to notice when our own tools fall short—or hit a wall. It is when our x-ray experiments yield blurry images, our spectroscopes generate results too noisy to interpret clearly, our statistical methods cannot decipher well patterns in vast datasets, our computational power is not strong enough to simulate complex systems. Instead of seeing limitations, here we see where new innovations are needed. When all researchers begin deliberately looking for gaps and limits in the very tools they P themselves use, science will begin spotting vast opportunities to resolve them. It opens opportunities to spark transformative innovations across science. What we need is a shift in mindset about tools and discovery: the key is recognising that the gaps and limits we hit with our tools are far from failures or dead ends. They are the critical signposts showing us exactly where to look, test and push forward. The imperfections are commonly the best inspiration and roadmap we have for where to innovate next—where to repurpose, scale up and optimise what we use to study the world. After all, our tools hit all kinds of bottlenecks: from low precision in measurement and rare signals to slow processing and limited resolution. The powerful principle here: each shows how close we are to breaking new ground. We looked at how science's top ten most influential toolmakers worked exactly this way (in Chapter 6). They put science's fixation on chasing answers aside, they instead pursued what existing tools could not handle and filled the gaps—and ended up sparking an enormous cascade of discoveries across fields, arguably more than any other researchers. We already unpacked the question of how we innovate new tools; but here we come back to this core question from a different angle: laying out concrete examples of where science's most used methods and tools hit walls that hold back progress. We first map out the method-driven process of how we expand the boundaries of scientific fields. To expand our scientific boundaries we often need to address a fundamental problem or question – set the challenge Identify the constraint of a method or tool; and at times also a cognitive or social constraint to tackle the problem Develop a new method or tool, or improve or combine existing ones to tackle the constraint, or adopt a new tool from another field Discovery (sparked by the new tool that enables a previously inaccessible finding) So exploring science's top ten most used discovery tools, what do these instruments exactly let us do—and what barriers have they broken through? And crucially, what limits currently hold them back that we need to tackle to spur future discoveries? Each major tool extends what we can know, but also faces limits that define the edge of what we can know. Take the spectrograph—first built in 1859 by Gustav Kirchhoff and Robert Bunsen at the University of Heidelberg. Suddenly, we could see the hidden structures of atoms and molecules and the chemical makeup of distant planets and stars. This remarkable new instrument detected wavelengths of electromagnetic radiation we had no way to access before. It redefined the edge of astronomy and chemistry. Kirchhoff, a young professor from Königsberg Prussia, was even described 'as Bunsen's greatest discovery' because the tool transformed their entire field. But today, our best spectrometers again hit limits— from sensitivity that cannot detect faint, distant signals, to computational limits and software that struggle to process large data streams from space telescopes. We need to design more sensitive detectors to capture those weaker signals, better algorithms to analyse the flood of data and faster ways to process them without missing details. This same pattern shows up everywhere; take the electron microscope. Ernst Ruska, the inventor, said: 'The light microscope opened the first gate to the microcosm. The electron microscope opened the second gate to the microcosm. What will we find opening the third gate?' Can quantum microscopy be the next key to opening up another layer of reality through even higher resolution? To get there, we have to overcome its current challenges—from electron damage to samples, to slow imaging times. From supercomputers that struggle with processing power limits when modelling earth's complex climate, to MRI scanners that struggle with blurred motion in living patients, our tools all hit walls. Everywhere we look we find these bottlenecks waiting to be tackled, in each method we use—unlocking the next unexpected findings. Often the most important question we can ask ourselves to break new ground is: how can we better spot and tackle the bottlenecks facing our own tools? And the bottlenecks facing science's most influential discovery tools? That is the heart of much of how we push the current edges of science: not just running experiments and coming up with theories to explain what our existing tools reveal, but building the next generation of tools that let us see and imagine in entirely new ways. So while we have zoomed out and uncovered the broad patterns linking science's powerful tools to its biggest discoveries, here we zoom in further on science's top ten breakthrough tools. Spotting the bottlenecks in how we do science is crucial because we cannot design a better tool that aims to fix a bottleneck we are not yet fully aware of. That is why it is so important to continually scan for blind spots in how we observe, measure or understand the world—and then identify what is exactly blocking them. Redrawing the edges of science depends largely on breaking those bottlenecks. Let us return here to randomised controlled trials. Until we created the RCT method in the mid-1900s, we were limited in our ability to prove—and at times even think rigorously about—what caused many diseases and how to treat them. Many lives were lost. Doctors often relied on observational comparisons that were more prone to bias. But with RCTs, we were finally able to systematically test how effective treatments actually are and better explain diseases. Like many methods, the RCT method is the outcome of continual reworking and many cumulative method refinements over time. It was built layer by layer—addingrandomisation techniques to control for hidden variables, blinding techniques to reduce observer bias, placebos to handle patients' expectation effects. The RCT's experimental design is a masterpiece, able to reduce uncertainty to provide life-saving evidence. It is an extraordinary example of how we create and upgrade methods to push past our limits. To understand phenomena from how our brains work to the origins of life and the universe, we rely on this kind of method scaffolding: experimenting with and reimagining our tools so they can take us further. Yet it is not just about inventing new tools that can see and measure better—it is also about how reliable our evidence and theories are using our tools. That means we have to constantly test, validate and fine-tune our tools. They must be calibrated and their errors corrected. Medical imaging scanners for example need to be regularly calibrated to avoid misdiagnosis. Gravitational wave detectors demand astonishing precision to filter out faint ground vibrations and local noise. AI models have to be checked for hidden biases or flaws in their training data. Ultimately, the power of our tools comes not just from cutting-edge design, but from meticulously ensuring that they deliver the most reliable results. Developing new methods and tools drives science by reducing our constraints: the ten central tools used most to trigger discoveries
Resolution limitations, challenges in observing living cells, sample size restrictions, and damage to samples from the electron beam
No universal detector in high-performance liquid chromatography, low detection with substances not volatile enough to limits at high resolution, difficulties separate Challenges in analysing highly complex mixtures; heating blurs the resolution, and limited capacity for large-scale applications
Constraints in sample size, scale and power; limited to studying phenomena we can represent statistically, constraints in the complexity of what we can model, and potential measurement error and sampling bias
Our inability to spin samples at ultrahigh speeds (greater than 20,000 rpm) to separate particles by density Our limited capacity to visualise the atomic structures of matter because of short wavelengths Our limited senses to detect and analyse electromagnetic radiation across wavelengths beyond the visible spectrum Our limitations in precisely perceiving and measuring variations in hot and cold temperature Our limited vision to magnify and perceive distant objects, detect different regions of the electromagnetic spectrum Separate small biological particles like cells, viruses and nucleic acids from fluids Analyse the atomic and molecular structure of matter like proteins and nucleic acids; and detect bone fractures, pneumonia and certain cancers Study structures of atoms and molecules, determine the chemical makeup of planets and stars, and electron arrangements of elements in different energy states Measure climate temperatures, surface temperatures in natural and industrial environments, and body temperature in health Observe exoplanets, stars, sunspots, galaxies, nebulae and other astronomical objects across multiple wavelengths Limits on the volume of material we can process, risks of cross-contamination between separated parts and damage to samples from high-speed rotation Low intensity of x-rays for light atoms, difficulties to analyse crystal growth in real-time; and overlapping signals in some diffraction patterns (x-ray powder diffraction) Low sensitivity such as in NMR; often require coupling with methods like chromatography to probe complex samples (as in mass spectrometers); and face computational limits in interpreting spectral data Constraints in accurately measuring extreme temperatures, especially below −100°C and above +300°C, and issues of calibration and sensitivity Blurred resolution from the Earth's surface caused by atmospheric disturbance; and space telescopes also face size and resolution limits
These are the ten methods and tools most often used to make science's 761 major discoveries, including all nobel-prize and major non-nobel discoveries. The year marks when the method or tool was first developed—though all have been vastly improved since. The final column highlights some of their current limitations that need to be tackled. P It can seem odd to ask: how can we optimise not just our methods but also our own mind and senses? But it is precisely what we need to answer to invent sharper lenses, better sensors and more powerful processors—to stretch what we can understand. Which bottlenecks are currently holding us back? Which ones have we not noticed yet? These questions point us toward entirely unexplored frontiers. By spotting gaps in what our tools can do and relentlessly testing and improving them, we open the door to discoveries we cannot yet imagine. The speed at which we develop ever more sophisticated computational, statistical and experimental methods will be a key catalyst in accelerating science and discovery—in a cumulative loop. Yet many puzzles exist where we have not yet beat the current limits of our tools. Think about how neuroscientists try to map the tangled connections of the brain, or epidemiologists attempt to track vast global health trends over time with precision. Or geneticists struggle with enormous genomic datasets using big data methods and network analysis—they find much noise and at times can introduce as much noise as they aim to tackle. Pushing past our current barriers especially in computational power, machine learning, big data methods, etc. is likely to spur new frontier research—from medicine to astrophysics. So what innovative, new methods can we design? Recent breakthroughs in quantum computing, machine learning, advanced sensors, neuroimaging, brain-computer interfaces, CRISPR, cryo-electron microscopes and so on offer a hint of what is possible. It is possible if the scientific community as a whole were to begin focussing on refining our tools to think and see in currently unimaginable ways—as those new tools enable us to. We could create ever more powerful biosensors that detect more diseases even earlier and more accurately—before symptoms appear. We could design more advanced microscopes that reveal the inner workings of living cells in real time at higher resolution. We could run massive online studies with thousands of researchers collaborating in real time, continually improving a study without a preset outcome (like an evolving online encyclopaedia). In such studies, replication and peer review would be built into the research design. Up to this point, we have been describing where science can take us and where it hits a wall—and we now outline how we break those boundaries: Pushing the current limits of science is a cycle—the closer we get to the boundaries of discovery, the fewer tools and methods we often have available, such as electron microscopes and statistical simulations that are just not powerful enough. Here we commonly have to spot the bottlenecks holding our tools back and come up with ways to break them. Once we invent, upgrade, combine or adopt new tools of discovery—often by building on available ones—we unlock new ways to observe, measure and understand the world that are impossible without them. That is how we redraw the edges of science. And with each better tool or method in our hands, the cycle begins again: we explore further, hit new barriers and develop new ways to see and think past them. It is this continual loop of bottleneck and invention that drives science forward—with the pace of progress more in our hands than we realise. P Ultimately, the engine of science is building better tools that stretch the limits of what our current tools, minds and senses can do. Every time we sharpen how we see, measure and test the world, we expand the universe of what we can know. About everyday knowledge, Ludwig Wittgenstein once famously said, 'the limits of my language are the limits of my world' . But when it comes to science and discovery, it is mainly the limits of our tools and methods that are the limits of our world. We hit our scientific boundaries when we run into the boundaries of what our best electron microscopes can resolve, when particle accelerators cannot reach higher energies, when machine learning tools cannot spot hidden relationships in noisy data. For our tools determine not just the experiments we have conducted so far but also largely the theories we have developed so far. Yet in fields like theoretical physics, some exceptional theoretical advances have involved larger conceptual leaps—leaps from existing experimental findings and tools to theory—than in most other fields (Chapter 6). So here is the key insight: our understanding of the world only reaches as far as our current toolbox allows. Beyond our toolbox, there is often not much more we can say about the world. If we want to better solve mysteries like how consciousness arises, what dark matter is made of, how to tackle Alzheimer's and many others we cannot yet imagine, we need new tools. To expand the universe of possible questions we can even ask, we need to build better tools: better quantum sensors that can detect unimaginably tiny gravitational shifts, sharper cryo-electron microscopes that reveal proteins at nearly atomic resolution, faster AI systems that can accurately analyse immense genomic data in hours instead of years.