Power And Progress Our Thousand Year Struggle Over Technology And Prosperity

The Real Story Behind Tech And Wealth

I came across a discussion thread last week asking whether new automation tools actually make workers better off or just consolidate money at the top. The same argument gets recycled every twenty years, roughly. Iron, steam, electricity, silicon, now whatever comes next. It is worth stepping back from the noise and looking at the actual pattern, because most people treat this topic like it is either a techno-optimist fairy tale or a dystopian horror story. Both are lazy. The core idea here is straightforward enough once you strip away the ideology. Technology changes who holds power, and that redistribution of power determines whether the broader population sees real gains in living standards or just richer elites. That has been true since the Bronze Age and it is still true today. What shifts across centuries is not the basic dynamic. What shifts is the speed at which it happens and the scale at which it plays out. When I first started digging into this subject properly, I expected to find a clean causal chain. Technology appears, productivity rises, wages rise, everyone benefits. That is the textbook version. Reality does not follow a straight line. Institutions matter. Property rights matter. Who captures the surplus matters far more than how much surplus exists.

Consider something most people miss about early industrialization. The steam engine did not automatically lift wages. In Britain during the late eighteenth and early nineteenth centuries, real wage growth stayed flat for decades even as productivity climbed. The Luddites were not anti-technology fools. They were responding rationally to a situation where their labor was being devalued while factory owners captured almost all of the gains. It took institutional responses, including early labor legislation and the expansion of schooling, before the broader population began sharing meaningfully in industrial output. The pattern repeats. Electricity in the early twentieth century did not immediately transform American life. It took twenty to thirty years of complementary investments, new business models, grid expansion, and household adaptation before the productivity boom materialized. The same lag shows up with computers, with the internet, and likely with artificial intelligence. The gap between a technology arriving and the economy actually benefiting from it is not a bug. It is structural. You have to rebuild organizations, retrain workforces, update regulations, and let new firms displace old ones before the gains appear in aggregate data. I ran into this problem directly when I advised a regional manufacturing firm a few years ago on upgrading their production line. They installed automated robotics because the vendor promised a forty percent reduction in labor costs and doubled output within eighteen months. What actually happened is far more interesting and far less dramatic. Output rose by about eighteen percent in the first year, labor costs dropped by twelve percent, and then we spent the next fourteen months dealing with a skills gap that the automation exposed. The existing staff could not program, maintain, or integrate the new systems effectively. We ended up hiring external specialists at premium rates, which erased most of the projected savings, and retained a smaller core team for routine oversight.

The workaround was unglamorous. We slowed the full rollout, brought in a technical training partner to upskill three internal employees over six weeks, and ran the new robots alongside the old manual process during a transition period instead of flipping a switch. The project took about nine months longer than the original plan, but it ultimately delivered the expected efficiency gains without the quality issues we were seeing in month two and three. That kind of friction is normal when technology outpaces institutional capacity. Most organizations skip the transition period because they are under pressure to justify the capital expenditure, and they pay for it later in downtime and errors. There is also a misconception about competition. New technology tends to concentrate power initially. The firms and individuals who adopt first and scale fastest capture disproportionate returns. That is why every major technological wave, from railroads to oil to semiconductors to large language models, produces a period of extreme inequality before it broadens. The concentration is not accidental. It is built into the economics of early adoption. Capital requirements, learning curves, network effects, and regulatory capture all favor incumbents and well-funded entrants. What eventually breaks the concentration is not morality. It is replication and competition. Rivals copy the technology. Open source implementations appear. Supply chains mature and equipment costs fall. New entrants arrive who have different cost structures or different market approaches. The process takes time, sometimes decades, and it is rarely smooth. Antitrust enforcement, education policy, and trade dynamics play roles. But the mechanism is mechanical, not benevolent.

If you want to evaluate any current technology through this lens, stop asking whether it is good or bad for society. Ask three specific questions. First, who captures the surplus in the early adoption phase. Second, what institutional barriers slow diffusion, whether they are patents, licensing, skill requirements, or infrastructure gaps. Third, what complementary investments are needed for the technology to generate broad productivity gains, and are those investments happening at scale yet. Artificial intelligence is currently in that early concentrated phase. The capital expenditure is massive. The economic gains so far flow overwhelmingly to a small group of firms that own the compute infrastructure and the model distribution channels. Wide consumer benefits and productivity improvements are measurable in specific domains, but the aggregate data is still emerging. The same thing happened with personal computers, with broadband, and with cloud computing. Each time, the first movers captured most of the value before the ecosystem matured enough to distribute gains more broadly. The downside of this framework is that it does not give you a clean prediction. It tells you the direction of travel and the conditions required for positive outcomes, but it cannot tell you when institutions will adapt or whether they will adapt fast enough to prevent real harm in the interim. There is also a risk of technological determinism if you read too much into the pattern and assume history moves inevitably toward broad prosperity. It does not. It moves toward whatever institutions and power structures channel it. The technology enables possibilities. Institutions decide which possibilities become reality.

That is the useful takeaway, stripped of the usual optimism or panic. Technology does not create prosperity automatically. It creates potential. Power structures determine whether that potential translates into widespread gains or concentrated wealth. The thousand-year record is consistent on that point. The details change, but the mechanism does not.