Learn to Code

When the coal industry began its terminal decline in the mid-2010s, a consensus emerged about what displaced workers should do. The advice was consistent enough to become a cultural shorthand: learn to code. Vice President Joe Biden said it directly at a New Hampshire rally in 2019. A startup called Bit Source in Pikeville, Kentucky, took the advice seriously and built a program training displaced Appalachian coal miners in web and software development. “Appalachia has been exporting coal for a long time,” said the company’s president. “Now we want to export code. We’ve got blue-collar coders.”

The program worked, to a degree. Some miners learned to code. The story was told as a parable of adaptation — proof that the American worker, given the right retraining, could ride the wave of technological change. The assumption embedded in the parable was that the wave had a destination, that there was a stable shore called the knowledge economy where digitally skilled workers would be safe.

That assumption is now demonstrably wrong, and the speed of its refutation is remarkable.

By 2025, AI systems were planning, building, testing, and deploying full software applications with minimal human input. Junior developer employment had fallen roughly 20 percent from its peak. Big-technology company hiring of new computer science graduates was down 55 percent from 2019. Employers surveyed in 2026 said they would rather allocate budget to AI tools than hire a new graduate. Kevin Park, who spent fifteen years at a major cloud provider, described the situation plainly: “We taught AI everything we knew. We documented our processes, recorded our meetings, shared our code repositories. We basically created the perfect training dataset for our own obsolescence.”

One analyst noted that programmers were “the canary in the coal mine.” The metaphor was not accidental. The creature sent ahead to detect the gas that would kill the miners. The programmers built the AI. The AI learned to program. The canary delivered its warning and then the gas arrived.

The Babylon Bee ran the headline in 2026: “Coder Displaced By AI Told He Should Just Learn To Mine Coal.” The joke lands because the cycle closed in under a decade. The coal miner was told to learn to code. The coder built the AI. Now the displaced coder is pointed back toward the mine, because the data centers running the AI that displaced him require the energy of the coming microreactor buildout, and someone has to work the supply chain.

This is not an anecdote about job disruption. It is a structural observation about who pays the cost of technological transitions and who collects the benefit. In every transition, the cost is paid by workers in the sector being automated. In every transition, the benefit accrues to the owners of the automating technology. In every transition, the workers are told that the next transition will be different, that this time the new jobs will be better, that the retraining programs will work, that the shore is just ahead.

The coal miner who learned to code in 2017 and watched his job evaporate in 2025 spent eight years on a shore that turned out to be a sandbar. The question for the programmer being displaced in 2026 is not whether there is another shore. The question is whether the wave has a destination at all, or whether the wave is simply the permanent condition of a world in which the owners of the technology determine what it does next, and the people who do not own it adapt or do not adapt, and the outcome in either case is the same.

The utopian promise is real. The question is who the utopia is for.