We have been here before.
In 1950, Claude Shannon published “Programming a Computer for Playing Chess,” the paper that first laid out how a machine could be made to play chess at all. It was not a curiosity. It was the opening move in a decades-long contest between human chess skill and mechanical calculation. Shannon’s paper did not stop people from studying chess. It did not empty the tournament halls. If anything, it sharpened the game. Players studied harder, calculated deeper, and treated the machine’s eventual arrival as a horizon to train toward, not a wall to stop at.
By 1997, Deep Blue beat Garry Kasparov. The machine had, by any measure, won. And yet chess did not die. Grandmasters did not retire in bulk. Tournament chess is more popular today, decades after machines definitively surpassed human play, than it was before Deep Blue existed. Players now train with engines, study engine lines, and use machine analysis to sharpen human intuition. The competition did not end when the machine became superior. It changed shape.
Coding is having its Deep Blue moment now. Language models write functions, debug logic, and scaffold entire systems faster than most engineers can type the problem statement. The instinct to feel obsolete is reasonable. But it mistakes the nature of the contest.
The chess world did not keep playing because humans could still beat the engine. They kept playing because the game itself, the discipline of calculation, pattern recognition, and judgment under constraint, remained valuable to practice regardless of who or what held the top rating. The engine raised the ceiling. It did not remove the floor.
The same logic applies to engineering skill. Writing code by hand, understanding why a system fails, reasoning about tradeoffs an AI has not been asked to consider, these are not activities that lose their value the moment a faster tool exists. They are the equivalent of a grandmaster studying an endgame that a chess engine solved thirty years ago. The solved status of the problem does not make the study pointless. It makes the study a discipline, not a race.
There is also a longer argument worth making plainly: human intelligence is not a fixed ceiling that AI simply walked past. It is a capacity that has always expanded under pressure. Every tool that has threatened to make a human skill obsolete, the calculator, the compiler, the search engine, has instead pushed the humans who kept practicing to operate at a higher level of abstraction. The people who stopped practicing because a machine was faster did in fact become obsolete. The people who treated the machine as a new opponent to train against did not.
This is not a claim that humans will always out-code AI at raw output speed. It is a claim about what structured practice does. A chess player who trains against an engine every day for ten years does not become slower than the engine. They become a different, sharper kind of player, one who understands the game at a depth the engine’s raw calculation cannot articulate. The same applies to a developer who keeps building, keeps debugging, keeps reasoning through architecture by hand, even when a model could do it faster. The practice is not about winning the speed contest. It is about building a kind of judgment that only comes from doing the work yourself, repeatedly, deliberately, over time.
Shannon did not write his 1950 paper to end chess. He wrote it to define what a machine would need to compete. Seventy-five years later, the game is still being played, by humans, against machines, and increasingly with them. The coding era will follow the same arc. The useless feeling is real. It is also temporary, and it is optional, if the response to it is structured practice rather than surrender.
We didn’t stop then. There’s no reason to stop now.