The Loom and the Algorithm: What History Tells Us About AI and Jobs

There is a display at Colonial Williamsburg that explains how weaving worked before the Industrial Revolution. Cloth was expensive — genuinely, significantly expensive in ways that are hard to imagine today. A single garment represented hours or days of skilled hand labor. Ordinary people owned few clothes, mended them carefully, and passed them down. Cloth was wealth. Then the power loom arrived. The economics changed entirely. Cloth that had taken a skilled weaver a full day to produce could now be produced in a fraction of the time, at a fraction of the cost. Hand weavers lost work. There was real displacement, real hardship, and real anger — the Luddites who smashed the looms were not reacting to a phantom threat. The specific jobs they knew were genuinely being eliminated. But something else happened alongside the displacement. Because cloth was suddenly cheap, demand exploded. Not the same demand for the same amount of cloth at a lower price. Entirely new demand — demand from people who had never been able to afford certain goods, demand for new kinds of garments, demand for new applications of fabric in upholstery, furnishings, and goods that had previously been out of reach for ordinary households. The overall industry grew dramatically. Different jobs replaced the lost ones. More people, eventually, were employed making cloth than had been employed before the loom arrived. This is the pattern of automation. It has repeated throughout the history of technology. Understanding it is one of the most useful things you can do if you are trying to think clearly about what AI will do to the job market. **The Economics of Efficiency: Why Cheaper Usually Means More** The counterintuitive truth about efficiency improvements is that they usually produce more consumption of a thing, not less. When something becomes significantly cheaper and better, people do not simply buy the same amount for less money. They buy more. They find new uses for it. New markets emerge that did not exist at the previous price point. This concept is formalized in economics as the Jevons Paradox, named after the 19th century economist William Stanley Jevons, who observed that improvements in steam engine efficiency led to greater coal consumption, not less. The engines needed less coal per unit of work. The lower operating costs made steam power economical for more applications. Total coal consumption went up. The same dynamic appears throughout technology history. The printing press made books dramatically cheaper to produce. Book consumption did not stay flat — it exploded. Literacy rates rose, new markets for printed material emerged, and the number of people employed in the production of written content grew enormously over the following centuries. The scribes who hand-copied manuscripts were displaced. The overall industry expanded far beyond what hand-copying could ever have served. Electricity made manufacturing dramatically cheaper. Factory output did not stay flat — it grew to serve demand that had previously been impossible to meet at viable prices. New products became affordable. New industries emerged. Employment in manufacturing grew for generations after electrification, even though electricity made each unit of production cheaper. The internet made distribution of information essentially free. Media consumption did not stay flat — it grew to a scale that physical distribution could never have served. New forms of content emerged. New industries built on top of the distribution layer. The number of people employed in content creation, software development, and digital services grew to dwarf the industries that digital distribution disrupted. **Where Software Is Right Now** Software development today is expensive in the same way that hand-woven cloth was expensive before the power loom. Building quality software requires significant skilled labor hours. The cost of this labor means there is a large population of businesses, ideas, and problems that cannot economically justify the investment required to build software solutions for them. They are priced out. There is an enormous amount of potential software that does not get built because the economics do not work. The small business that would benefit from a custom inventory system but cannot justify the development cost. The research application that would be valuable but falls below the threshold of what a development team can prioritize. The internal tool that everyone knows would save time but never gets built because the backlog is too long. The startup idea that is abandoned because the cost to validate it with a real product is too high. AI tools are beginning to change those economics in exactly the way the power loom changed the economics of cloth. Platforms like GitHub Copilot, Cursor, and Replit are reducing the labor required to produce software. That reduction in cost is going to unlock demand that was previously priced out — demand that currently does not exist because the product to serve it was too expensive to build. **The Stackzilla Connection** Every wave of automation in software development has followed this pattern. AWS made infrastructure so cheap that countless businesses were able to run servers that could never have afforded their own hardware. The result was not fewer servers — it was an explosion in the number of applications running in the world, and more people employed in cloud computing than were ever employed in on-premise infrastructure management. GitHub made software collaboration so accessible that open-source development scaled to a level that was previously impossible. The tools that Stackzilla catalogs represent exactly this history: React made rich web applications economically buildable for teams without years of browser compatibility expertise. Stripe made payment processing accessible to developers who would previously have needed specialized integration knowledge. PostgreSQL made sophisticated database infrastructure available to projects that could never have afforded enterprise database licenses. Each of these made something cheaper. Each created demand that was previously unreachable. Each produced more employment in adjacent areas than it displaced in the specific tasks it automated. AI tools are the next step in this sequence. The specific tasks they automate will see displacement. The overall demand for software — and for the human judgment required to build it well — will expand. The question is not whether jobs will exist. The question is which jobs, and what they will look like. Parts 2 and 3 of this series address exactly that.