Why Breakthroughs Disappoint and Work Delivers
You know that feeling when you read about the latest “breakthrough” technology that’s going to change everything—fusion finally working, quantum computers achieving some new milestone, brain-computer interfaces getting closer to reality—and part of you feels excited but part of you thinks, haven’t I heard this before?
I’ve been carrying around a low-level disappointment about technology promises for years now. Remember when VR was going to transform everything? You bought into the hype, got a headset, used it enthusiastically for maybe two weeks, and now it’s gathering dust in a closet. Or self-driving cars: we’ve been perpetually “just a few years away” from full autonomy for over a decade now (and the current rollout still relies on an operations centre with remote drivers). Blockchain was going to revolutionise everything from voting to supply chains, but mostly it revolutionised speculation and energy consumption.
This got me wondering: why does this keep happening?
The Pattern Emerges
I started reading tech announcements differently. The pattern became impossible to ignore: breathless breakthrough, massive hype, gradual reality check, then the next breathless announcement elsewhere. Rinse and repeat.
Take fusion energy. Every few months, there’s a new milestone: “Scientists achieve fusion breakthrough!” “Fusion reactor sets new record!” “Fusion ignition milestone passed!” I’ve been reading variations of these headlines my entire life. The National Ignition Facility’s 2022 achievement of fusion ignition1 was genuinely impressive science, but we’re still nowhere close to practical fusion power. Each repetition is a controlled experiment; none yet is a coal-mine pump that must run.
Or quantum computing: IBM announces quantum advantage, Google claims quantum supremacy, startups promise quantum revolution. However, the proto-revolution still lives in laboratory cryostats that only specialists touch. The feedback loops are long and sparse. Very few people are doing sustained, practical work with quantum computers to solve problems they actually need solved today. The most sophisticated quantum computers can barely outperform classical computers on carefully chosen problems, and they require near-absolute-zero temperatures to function at all.
The pattern holds across domains. Laboratory achievements get confused with practical capabilities. Scientific possibilities get mistaken for imminent technological realities. But here’s what started bothering me: if this pattern is so consistent, what’s the difference between technologies that actually change our lives and ones that remain perpetually “almost ready”?
This connects to something I’ve written about before; we’re living through the collapse of our consensus stories about how progress works.2 The research→technology narrative is one of these broken stories, with deep institutional roots. Vannevar Bush’s influential 1945 report Science—The Endless Frontier3 promised that basic research would automatically flow into applied research, then development, then production. This linear model shaped decades of science policy and public expectations: labs discover, industry applies, society adapts. However, the growth of scientific knowledge has often owed more to the concerns of technology than the other way around. When the science to technology story fails to predict what actually happens, we get the disappointment pattern: breathless announcements followed by reality checks, rinse and repeat.
I suspected we were climbing the wrong slope entirely—treating technology as a straight-line summit when history insists on switchbacks, and getting the entire causality arrow backwards.
The Detective Work
I decided to look at fusion more closely, because it’s been “almost ready” longer than almost anything else. The basic science has been solid since the 1930s. We’ve known how to make fusion reactions happen since the 1950s. Massive government and private investment has poured in for decades. Brilliant scientists and engineers have dedicated their careers to solving the engineering challenges.
And yet, after all this time and money and talent, fusion power remains elusive. The conventional narrative blames stubborn engineering challenges. But what if that’s not the real problem?
Then I came across this passage from historian David Landes about the Industrial Revolution,4 and everything clicked:
The demand for coal pushed mines deeper until water seepage became a serious hazard; the answer was the creation of a more efficient pump, the atmospheric steam engine. A cheap supply of coal proved a godsend to the iron industry, which was stifling for lack of fuel. In the meantime, the invention and diffusion of machinery in the textile manufacture and other industries created a new demand for energy, hence for coal and steam engines; and these engines, and the machines themselves, had a voracious appetite for iron, which called for further coal and power…
Or more pithily:
Water → pumps → engine
Engines → iron → coal
Coal → rails → markets → bigger engines
This is exactly what Richard Sennett emphasises in The Craftsman:5 real capability emerges not from abstract design, but from the sustained dialogue between head and hand, tool and task. Knowledge crystallises through the work itself. The miners struggling with water seepage were not passive recipients of someone else’s invention—they were co-creators of the steam engine as a workable technology.
Wait. This is completely different from how we think about technology today.
But this path wasn’t inevitable. Three centuries earlier, a conflict between Robert Boyle and Thomas Hobbes revealed exactly how knowledge-making could go wrong when the “head” systematically excludes the “hand.”6 Boyle’s air-pump experiments in the 1660s were designed to demonstrate natural philosophical principles to gentlemen witnesses in controlled laboratory settings. His spectacular demonstrations impressed royal audiences, but they proved nearly impossible for other natural philosophers to replicate.
The problem wasn’t the science—it was the method. Boyle deliberately excluded the craftspeople who actually built and maintained his air pumps from the knowledge-making process. As historian Steven Shapin shows, Boyle’s experimental reports systematically masked or obscured the skilled technicians whose tacit knowledge made the experiments possible.7 When other laboratories tried to reproduce his results, they often failed catastrophically, not because the principles were wrong, but because the essential craft knowledge had been written out of the official record.
Thomas Hobbes attacked not just Boyle’s conclusions, but his entire approach: Why should we trust knowledge that depends on expensive, finicky apparatus that only a few people can operate? Hobbes intuited something crucial—the technology was shaping the knowledge, not just revealing it. The air pump didn’t neutrally demonstrate natural laws; it created specific conditions that generated specific kinds of results.
The air pump required glassblowers, metalworkers, and daily operators whose troubleshooting generated crucial knowledge about pressure, materials, and failure modes. But this knowledge remained informal, unrecorded, and excluded from natural philosophy. The result was impressive laboratory achievements that couldn’t scale into reliable technology because the practical knowledge needed to make things work day after day had been systematically separated from the research program.
The Steam Engine Story
The steam engine didn’t emerge because James Watt had a brilliant insight about steam pressure. It emerged because miners were doing the dangerous, backbreaking work of going deeper underground to extract coal, and water seepage threatened to kill them or shut down the mines entirely. They desperately needed better pumps. But this urgency wasn’t natural—it was constructed through particular economic arrangements around coal extraction, labor conditions, and profit structures. Lives were at stake because miners were sent deeper underground to meet industrial demands for coal.
Notice the sequence. No one announced “Steam Engine Breakthrough Will Transform Everything!” Instead, miners faced an immediate, life-or-death problem: water flooding the shafts. The pumps they cobbled together were crude, but they worked. Operating them day after day revealed new problems—better iron, tighter seals, stronger boilers—which created new kinds of work, new needs, and new knowledge.
Here’s the crucial detail: Newcomen was an iron monger dealing with pump failures, Watt was a mining engineer fixing steam engines that kept breaking. They pulled in whatever ideas worked: steam pressure from natural philosophers, metallurgy from blacksmiths, mechanical principles from clockmakers. Stephenson, a tool maker working near universities, had access to diverse conversations that sparked creative combinations. But the technology emerged from their practical work, not from someone’s research agenda. The ‘breakthrough’ was retrospective—we only call it breakthrough science because the work succeeded.
This history shows that the arrow of causality pointed from the work to the knowledge, not the other way. As the historian David Landes himself argues, the link between the Scientific Revolution and the Industrial Revolution was “extremely diffuse.” If anything, he contends, “the growth of scientific knowledge owed much to the concerns and achievements of technology; there was far less flow of ideas or methods the other way.”8 The steam engine wasn’t applied science; it was a source of it. The practical work of solving the problem of pumping water generated new scientific questions about thermodynamics, pressure, and metallurgy. The same is true with LLMs. There was no grand theory, just a lot of tinkering, and we still don’t understand why they work. The key realisation was that scaling these architectures and datasets—throwing more compute and data at the problem—yielded new capabilities that were not predicted by the creators. The “why” was discovered after the “that.”
And then notice what Landes shows: the steam engine created new demands, which created new kinds of work, which created new technological needs. Steam engines needed iron, so iron production had to scale up. More iron production needed more coal. More coal needed better transportation. Better transportation opened new markets. New markets demanded more manufacturing capacity.
Each step was oblique: like hikers who reach the summit only by taking switchbacks, robust technologies emerge through indirect paths. No straight line could have predicted railways or textile mills from a water pump. Yet that crooked path scaled faster than any master plan.
Call it obliquity: the fastest route to robust technology is almost never the shortest.9
This gave me a completely different way to think about technology. Technology develops obliquely, pulled into existence by sustained, practical work on recalcitrant tasks—regardless of whether it started in a mine shaft or a laboratory.
The Framework
Technology crystallises from practical work, not abstract research.10 The steam engine ‘contained’ the accumulated learning of miners dealing with water, metalworkers discovering pressure limits, operators finding optimal fuel ratios. What we call ‘technology’ is really this work-knowledge made durable and transferable. The daily frustrations of pumping water became the stored intelligence of steam technology.
Applied to technology, this means the shortest vector on the map is rarely the quickest in reality. Technology emerges through sustained work on urgent problems that refuse to wait for perfect understanding. The steam engine engineers weren’t just applying what they already knew about steam pressure. They were creating new knowledge through the daily work of building, testing, refining, and operating these machines in real conditions with real problems. The technology and the knowledge developed together through sustained practical work, but along winding, oblique paths.
This explains the pattern I was seeing. Laboratory achievements demonstrate scientific principles, but they don’t create the kind of embedded, practical knowledge that constitutes real technology. That knowledge comes from different kind of work: the work of making something function reliably, day after day, solving real problems that someone actually needs solved today.
Consider the transistor. Bell Labs wasn’t pursuing abstract research into semiconductor physics. They were a telephone company solving a practical problem: mechanical switches were too slow and unreliable for the growing volume of calls. The transistor emerged because telephone operators were doing the daily work of connecting calls, and that work demanded faster, more reliable switches. Engineers pulled ideas from wherever they could find them: quantum mechanics from physicists, crystal properties from metallurgists, manufacturing techniques from vacuum tube makers. The ‘breakthrough’ in semiconductor physics came from this practical work driving creative combinations of diverse knowledge. This creates a feedback loop: practical work pulls in insights, which enable new kinds of work, which surfaces new problems that redirect research priorities. But the pull of urgent work is what transforms scientific possibility into robust technology—without it, ideas remain in the laboratory indefinitely.
ChatGPT turned from curiosity to infrastructure the moment millions of us treated it like a slightly unreliable intern—debugging its responses at 2 AM, inventing prompt tricks that never appeared in any paper, discovering it was terrible at math but surprisingly good at explaining why our code wouldn’t compile. Each use was an unpaid experiment, and that massive ecosystem of practical work rapidly created new knowledge about what these systems could and couldn’t do reliably. The technology is emerging obliquely through the work.
But what counts as ‘work’ here matters. The knowledge-creating loop includes maintenance technicians diagnosing recurring failures, customer support staff categorising user complaints, safety operators handling edge cases, and even unpaid users debugging prompts and reporting glitches. Some of this work gets compensated and recognised, much of it doesn’t. The framework depends on this entire ecosystem of practical activity, not just the visible engineering effort.
Even technologies that seem to contradict this pattern actually support it. The Manhattan Project appears to be top-down, linear success. While the scientific principles were already known (Einstein’s letter to Roosevelt was just one idea in the library of possibilities) the technology itself emerged through sustained practical work: metallurgists figuring out uranium purification, engineers designing centrifuges, operators running enrichment facilities day after day. Similarly, technologies that began as solutions in search of problems—like the laser—remained laboratory curiosities until practical work pulled them into applications: cutting materials, reading data, transmitting information. Each application drew on knowledge from unexpected places—optics from telescope makers, precision mechanics from watchmakers, materials science from semiconductor fabrication.
The pattern is consistent: work pulls ideas sideways into reality; ideas alone do not push work upward.
In contrast, fusion researchers are doing incredibly sophisticated scientific and engineering work but they’re not yet doing the sustained, practical work of operating fusion systems to solve urgent, narrow problems that someone needs solved tomorrow. The knowledge that would constitute “fusion technology” can only emerge from that kind of work.
The direct route—announce ignition, raise billions, promise commercial kilowatts by 2035—feels rational but has failed to deliver for seventy years. These side-hustle applications already flicker at the edges: small reactors producing neutrons for medical isotopes, pilot plants selling process heat to hydrogen refineries. These applications don’t need to beat natural gas on price; they just need to stay on long enough for operators to learn, break, fix, and learn again.
Why This Matters
This isn’t just about making better predictions. The standard narrative treats technology as this autonomous force that just happens to us: “technological disruption,” “the pace of change,” forces we must adapt to but can’t really control.
But if technology is crystallised knowledge that emerges through oblique paths of practical work, then we’re not passive recipients of technological change. The technological future isn’t predetermined by scientific possibility—it depends entirely on what problems we choose to take seriously and what work we organise around them today.
The real pattern is oblique: work pulls in whatever ideas prove useful, technology emerges from sustained practice, then we construct the research narrative backwards to make sense of what happened. Boyle’s air pumps showed how this goes wrong when we separate the head from the hand. We’re not just getting fusion timelines wrong—we’re getting the entire causality arrow backwards.
Self-driving cars show the oblique path in motion: Waymo vehicles have struggled to decipher construction worker hand signals, juddering to a halt when faced with mixed visual cues. Each operational day generates new edge cases—roundabout navigation, delivery robot interactions, stop sign ambiguity—that force remote operators to develop new protocols, which require engineers to rethink sensor fusion priorities, which expose hidden assumptions about how the system processes conflicting information.11 This daily work of handling failures creates knowledge that couldn’t emerge from simulation alone.
Obliquity gives us agency. Instead of asking “How do we adapt to fusion?” we can ask “What small, urgent problem could we solve tomorrow that would force us to learn how to live with fusion the day after?” May<be it’s tritium-breeding blankets for research reactors, maybe it’s neutron diagnostics for hospitals. Whatever the entry point, the technology will mature fastest when it’s pulled sideways into the messy, daily business of keeping some urgent thing running.
This doesn’t diminish research—it clarifies its role. Research contributes to the library of possibilities, alongside folk knowledge, craft traditions, accidental discoveries, and cross-pollination between trades. Ideas come from everywhere: university labs, workshop floors, customer complaints, production failures, chance observations. But pouring more funding into any single source doesn’t automatically generate more innovation. The bottleneck isn’t the supply of ideas; it’s the sustained practical work that combines diverse knowledge into working technologies. The most elegant insights from any tradition remain inert without someone desperately needing to solve a problem today.
Seen this way, the lesson is consistent across domains: as Sennett reminds us, knowledge is born in craft, and as Benjamin Sovacool shows in his research on energy transitions, technologies flourish when research stays open to the pressures of real practice.12 Technology doesn’t arrive whole from the lab; it takes the oblique, crooked path of work.
The next time you read about a “breakthrough” that’s going to change everything, ignore the promised slide-deck Everest. Ask yourself: Where’s the coal mine? Where are the people doing work that desperately needs this solution? Where’s the ecosystem of interconnected practical needs that would pull this technology into existence through sustained, oblique work?
Look for the nearest hill where someone is already cursing today’s broken tool—recognising that which problems count as ‘urgent’ is itself shaped by economic and political choices about what work we organise society around. Start climbing the crooked path. The technology will meet you there.
- Purcell, G. (2025, May). Lawrence Livermore National Laboratory: The future of fusion following ignition. Innovation News Network. https://www.innovationnewsnetwork.com/lawrence-livermore-national-laboratory-future-of-fusion-following-ignition/. ↩︎
- Evans-Greenwood, Peter. “The Collapse of Consensus Stories.” Substack newsletter. The Puzzle and Its Pieces, August 19, 2025. https://thepuzzleanditspieces.substack.com/p/the-collapse-of-dominant-stories. ↩︎
- Bush, Vannevar. Science—The Endless Frontier: A Report to the President. Washington, DC: United States Government Printing Office, 1945. ↩︎
- pp3 of “Introduction” in Landes, David S. The Unbound Prometheus: Technological Change and Industrial Development in Western Europe from 1750 to the Present. 2nd ed. Cambridge University Press, 2003. ↩︎
- Sennett, Richard. The Craftsman. Penguin, 2009. ↩︎
- Shapin, Steven, and Simon Schaffer. Leviathan and the Air-Pump: Hobbes, Boyle, and the Experimental Life: With a New Introduction by the Authors. Princeton University Press, 2011. ↩︎
- “Establishing matters of fact did require immense amounts of labour. Here we endeavour to recover this labour for our historiographic purposes: to show the inadequacy of the method which regards experimentally produced matters of fact as self-evident and self-explanatory.”” pp. 225-226 in Shapin and Schaffer, Leviathan and the Air-Pump. ↩︎
- pp. 61 in Landes, D. S. (1969). The unbound prometheus. ↩︎
- A term coined by economist John Kay. ↩︎
- The historian Matteo Pasquinelli calls this the “labor theory of knowledge.” Knowledge itself, he argues, doesn’t come from abstract thinking but from practical human work and activity. This is what he means: knowledge crystallises from practical work, not abstract research. This reverses the standard model: it wasn’t that science enabled the technology; the technology’s demands created new science. See Pasquinelli, Matteo. The Eye of the Master: A Social History of Artificial Intelligence. Verso, 2023. ↩︎
- Nield, D. (2025, January 2). “Waymo self-driving taxi can’t figure out construction worker’s hand signals”. Futurism. https://futurism.com/the-byte/waymo-self-driving-taxi-construction-worker-hand-signals. ↩︎
- Sovacool, Benjamin K. “The Importance of Open and Closed Styles of Energy Research.” Social Studies of Science 40, no. 6 (2010): 903–30. https://doi.org/10.1177/0306312710373842. ↩︎