Wang Yonggang, a senior engineer who once worked at Google, published a long article titled "Farewell, Programmers," documenting how AI programming had "hollowed out" his role. This September, an American computer teacher reached out to him, wanting to improve the Altair 8800 computer simulator he had open-sourced six years earlier.
He handed the requirements to Claude Opus 5.5 and asked it for suggestions at the system design level. The AI not only completed the original work, but also proactively proposed adding an assembler, a debugger, and a teletype emulator interface. In about three or four days, what had originally been a few hundred lines of code swelled into a project of more than 30,000 lines, complete with unit tests and browser integration tests. What he actually did was only confirm the approach, ask questions, and give feedback—the bulk of making judgments and writing code had been handed off.
AI Crossed the Line into the Archive Room What surprised him even more was the AI's "side job." It dug through historical materials on its own, found experts, sent emails across languages, and even used external assets to calibrate 3D models. In a project to reconstruct Leibniz's mechanical calculator, the AI even helped him contact a museum in Germany and eventually obtain 135 pages of scans of historical drawings and manuscripts. Beyond writing code, it also took on the drudgery of finding people and materials.
A Generation Built AI—and Broke Its Own Rice Bowl First Wang Yonggang judged that as AI programming capabilities advance, the technical experience programmers have built up over the years is being repriced. He predicts that future development teams may shrink to a few product and business leads directing a large number of AI agents, and that the career space for traditional programmers will be squeezed. He lamented that this generation of engineers personally helped create AI, yet may now become one of the earliest groups to be hit. As for where computer science graduates should go, and what kind of people the AI era actually needs, he himself has no answer.
