The Low-Hanging Fruit Theory of Innovation Is Only Half Right
A familiar theory of innovation says that progress becomes harder because the low-hanging fruit is picked first. Early discoveries lie close at hand; later advances require larger machines, larger teams, deeper specialisation and more money.
Nicholas Bloom, Charles I. Jones, John Van Reenen and Michael Webb developed an influential empirical version of this argument in “Are Ideas Getting Harder to Find?”. Across semiconductors, agriculture, medicine and firm-level innovation, they found research effort rising substantially while measured research productivity declined. Their most striking example is Moore’s Law: maintaining the historical rate of improvement in chip density requires far more researchers than it did in the early 1970s.
There is considerable truth in this. Mature fields accumulate technical difficulty. Once obvious routes have been explored, the remaining work often requires greater precision, more infrastructure and command of a much larger body of knowledge. The mistake is not taking this burden seriously. It is treating one pattern of research productivity as a complete description of innovation.
A 2025 US Census working paper complicates the picture. Using a 46-year panel of American firms and patents, Teresa Fort and her co-authors find increasing patent output per research input and little evidence of a common secular decline in high-quality patenting. Their results suggest that part of the weakness may lie between invention and growth—in diffusion, adoption, organisational change or other forces not captured by patent production.
The two studies do not measure exactly the same thing, and patents are an incomplete proxy for ideas. The newer paper does not establish that research has become easy, just as the earlier one does not establish that every source of technological renewal is being exhausted. Together, they suggest that “ideas are getting harder to find” may combine several different problems under one phrase.
The low-hanging-fruit metaphor imagines a single tree: a stable domain of opportunity where the easiest gains disappear first and every later discovery sits higher among the branches. That can be true locally. Innovation as a system is less static. It changes the ladder, alters the search process, embeds earlier difficulty in reusable tools and sometimes makes another tree possible.
Innovation Changes the Search Space
Electricity did not merely provide a better way to perform tasks that were already practical. It made entire classes of activity possible at useful scale. Computing, the internet and molecular biology produced similar changes. They created new ways to observe, represent, coordinate and intervene, allowing ambitions that had long existed to become technically and economically actionable.
The internet offers a clear example. Search engines, online marketplaces, streaming services, cloud software and digital publishing did not appear because earlier generations lacked any desire to find information, sell goods at a distance or distribute media more widely. What changed was the environment in which those aims could be pursued.
Packet switching, common protocols, personal computers, browsers, commercial networks and broadband were not one invention. They formed an infrastructure assembled through many linked technical and institutional advances. DARPA’s history of ARPANET describes the need to transmit information without depending on a single node, allow incompatible computers to communicate and create protocols through which separate networks could operate together.
Once that infrastructure existed, later opportunities could look obvious. Of course people wanted better search, convenient remote shopping and software available through a browser. Yet inevitability is often what possibility looks like after the platform has arrived. The opportunity was not sitting untouched at eye level; decades of engineering and institutional work changed what could be reached.
This is the limitation of treating innovation as movement along one fixed frontier. New instruments and systems do not guarantee an endless supply of easy discoveries, but they can redraw the boundary between a difficult question, a practical research programme and an ordinary application.
Difficulty Moves Into the Platform
Opening a new field does not mean discovering an untouched orchard where valuable ideas can be collected without effort. What appears to be a sudden reset in difficulty may represent a transfer of difficulty into the platform on which later work depends.
A sequencing machine, telescope, semiconductor fabrication process, programming language or network protocol allows thousands of users to begin closer to the frontier. They no longer need to reconstruct the instrument before asking the next question. The field feels more accessible because earlier scientific and engineering labour has been compressed into equipment, software, standards, supply chains and trained practice.
The difficulty has not vanished. Someone must design the instrument, manufacture it reliably, maintain it, establish compatible standards and create institutions capable of supporting its use. Users experience an opened field; the platform contains the accumulated cost of opening it.
This distinction explains why new technologies can generate bursts of apparently easy progress without refuting the difficulty of the work that preceded them. When enough pieces align, experiments become cheaper, feedback accelerates and participation broadens. Applications that once required a specialised laboratory may become available to a startup, a hospital or an individual researcher.
New fields also mature. Their earliest opportunities may be exhausted quickly, their infrastructure may remain expensive and their discoveries may diffuse slowly. Scientific output can increase without producing proportionate economic growth if firms cannot adopt the technology, complementary infrastructure is absent or institutions resist reorganisation. Creating another tree is therefore not proof that aggregate research productivity cannot decline. It identifies one mechanism through which renewal can occur.
Opening, Exploring and Diffusing Are Different Jobs
Debates about innovation often become confused because they treat progress as one continuous activity. At least three kinds of work need to be distinguished: opening a field, exploring it and diffusing the results.
Opening a field means producing the concepts, instruments, datasets or infrastructure that make a previously inaccessible domain available. Exploring it means testing hypotheses, applications, designs and business models within that space. Diffusion means turning successful results into reliable, affordable and widely usable practices or products.
These stages overlap and feed back into one another. A company can build an instrument that opens a scientific question. Academic research can solve a technical problem exposed by commercial use. Public procurement can create a market that supports private experimentation. A corporate laboratory can open a domain later explored by startups, while startup tools can make university research cheaper.
Innovation is therefore not a clean pipeline beginning with public science and ending with private enterprise. It is a recursive system in which institutions carry different combinations of scientific, technical, commercial and organisational uncertainty.
The distinction also helps explain why ideas and growth can move apart. A society may continue producing patents and discoveries while becoming worse at translating them into productivity. The reverse can happen as well: a wave of profitable products may exploit a domain opened by one earlier breakthrough without creating another field behind it.
Visible activity is consequently an unreliable measure of frontier health. New companies, product launches, adoption curves and valuations are easy to count. Instruments, protocols, negative results, technical standards, trained communities and institutional capabilities are slower and less visible, even though they determine what can be attempted next.
Different Institutions Carry Different Uncertainties
Venture capital is often effective once a frontier has become legible enough to organise a company around it. Investors can finance product discovery, test commercial applications and push successful models towards scale. A capability that never becomes reliable, manufacturable, distributable or affordable has limited reach, so this exploratory and translational work is not a minor contribution.
Venture capital is not equally suited to every uncertainty. Its model generally requires the possibility of ownership, rapid growth, a sufficiently large market and an eventual route through which investors can realise a substantial return. Open-ended research with uncertain applications, long timelines or benefits that spill broadly across society may fit poorly even when its potential value is considerable.
Research on venture funding for science-based startups illustrates the friction. Such firms often take longer to secure investment than startups less rooted in frontier science. The authors identify a tension between scientists focused on technological advancement and investors seeking evidence of market demand, particularly in early funding rounds. The finding does not mean venture investors reject science; it shows that scientific promise and venture readiness are different thresholds.
This is the distinction developed further in the Journal’s essay on which kinds of uncertainty venture capital is structured to carry. Venture finance can explore a newly legible field aggressively. It is less consistently suited to financing the decades of uncertain work required before anyone knows that a commercially navigable field exists.
Public research, universities, corporate laboratories, procurement programmes, standards bodies and patient private capital tolerate other combinations of risk. Their strengths and failures differ because they optimise for different outcomes. A public funder may value scientific knowledge, public health or national capability even when no single organisation can capture most of the eventual return. A corporate laboratory may combine specialised infrastructure with a long horizon but remain constrained by its parent company’s strategy.
Biomedical research provides evidence of interaction rather than simple substitution. A study of NIH funding and private-sector patenting found that additional public research funding stimulated subsequent private patents. Public investment did not merely replace commercial work; it altered the knowledge base from which companies developed patentable technologies.
No institution owns a permanent position in the sequence. Public science can open a field, companies can create research tools, startups can expose new theoretical problems and universities can preserve questions whose applications are not yet visible. A healthy innovation system depends less on selecting one heroic institution than on maintaining connections among organisations capable of surviving different kinds of failure.
Tools Can Change the Productivity of Discovery
The strongest challenge to a fixed low-hanging-fruit account is not the claim that human creativity is limitless. It is that the technology of research can itself improve.
Better instruments reveal phenomena that could not previously be observed. Automation increases the number of experiments that can be performed. Simulation reduces the need for physical prototypes. Shared datasets, standardised software and reusable laboratory methods allow researchers to begin from capabilities earlier generations had to construct themselves.
Artificial intelligence may become important here, although the direction is not guaranteed. It could simply produce more papers, candidates and simulations inside fields already generating too much weak or repetitive work. It may also alter the search process by proposing molecules or materials, identifying patterns, prioritising experiments and reducing the cost of exploring very large possibility spaces.
The relevant distinction, examined in the Journal’s essay on whether AI creates more paths or changes the map of discovery, is not merely whether scientists work faster. A tool can accelerate movement through an existing representation without changing which questions can be formulated or how answers are validated. A faster ladder may still lean against the same tree.
More capable research tools can nevertheless change the economics of attention. Questions ignored because each experiment was too slow or expensive may become practical. Unexpected results may expose weaknesses in the representation and help create a new field. The line between exploring a map and redrawing it is not permanent.
This means that research difficulty is partly endogenous. The frontier becomes harder as knowledge accumulates, but the methods used to approach it also evolve. Aggregate productivity can decline while particular tools create large local gains, just as patent output can rise while the connection between invention and economic growth weakens.
The Next Tree
The low-hanging-fruit metaphor captures a real tendency. Fields mature, obvious opportunities disappear and maintaining the same visible rate of advance can require much more effort. What it cannot provide by itself is an account of renewal.
The history of technology is also a history of domains becoming actionable: electricity, antibiotics, aviation, computing, the internet and molecular biology. Each emerged from difficult, uncertain and often expensive work. Each later created environments in which further experimentation became dramatically easier for people who inherited the resulting tools.
That history does not guarantee endless acceleration. New fields can disappoint, mature quickly or produce gains that remain inaccessible to much of society. Some discoveries spread slowly because complementary institutions are missing. Others generate commercial activity without opening a durable scientific frontier.
The real slowdown risk is therefore not simply that humanity runs out of ideas. It is that the systems needed to create, test, share and implement difficult ideas weaken while visible commercial activity remains lively. A society can become exceptionally good at recombining inherited technologies, improving interfaces and optimising distribution while investing less in the work that changes what can be attempted.
Confusing exploration with field creation encourages the wrong expectations. Venture capital is asked to behave like basic research, universities are judged like startups, public institutions are required to promise immediate commercial returns and markets are expected to finance value that cannot yet be priced. The system becomes better at refinement than renewal and mistakes that refinement for proof that the frontier is healthy.
Harvesting and planting are both necessary, but they are not interchangeable. One extracts value from an accessible landscape. The other creates the instruments, knowledge and institutions through which a new landscape becomes visible.
The danger is not that we suddenly discover there is no fruit left. It is that a crowded marketplace persuades us the frontier is alive while the horizon has quietly stopped moving—and that we remember how to recognise the next tree only after we have forgotten how to plant one.
Comments
Post a Comment