Wang Yonggang, a former Google engineer, admitted in his article “Farewell, Programmer” that after entrusting the project to Claude Opus 5.5, the AI autonomously generated an assembler, debugger, and interface; the codebase expanded from hundreds of lines to 30,000 lines, actively researched materials, and reached out to a German museum to obtain 135 pages of Leibniz’s manuscripts. He reflected that the AI he helped create is reshaping the programming profession—future development teams may consist of only a few individuals directing AI agents, leaving traditional programmers with shrinking opportunities. He has no answer to where graduates should turn next.Author and source: AIBase
Wang Yonggang, a senior engineer formerly at Google, published a lengthy article titled "Goodbye, Programmer," documenting his experience of being sidelined by AI programming. In September this year, an American computer teacher reached out to him, seeking to improve his open-source Altair 8800 computer simulator from six years ago.
He handed the requirements over to Claude Opus 5.5 to get suggestions from a system design perspective. The AI not only completed the original task but also proactively proposed adding an assembler, a debugger, and a teletype simulator interface. Within about three to four days, the original few-hundred-line program expanded into a 30,000-line project complete with unit tests and browser-based integration tests. All he actually did was confirm the approach, ask questions, and provide feedback—the bulk of decision-making and coding was delegated away.
AI has crossed into the archive room.
What surprised him even more was AI’s “side job”—it independently reviewed historical records, consulted experts, sent emails across languages, and used external materials to calibrate 3D models. In the project to restore Leibniz’s mechanical calculator, AI even helped him connect with a museum in Germany, ultimately obtaining 135 pages of scanned historical blueprints and manuscripts. Beyond writing code, it took on the tedious tasks of finding people and sourcing materials as well.
One generation created AI, only to smash their own bowl first.
Wang Yonggang believes that as AI’s programming capabilities rise, the technical expertise programmers have accumulated over the years is being revalued. He predicts that future development teams may shrink to just a few product and business leaders who direct large numbers of AI agents, squeezing the traditional career path of programmers. He laments that this generation of engineers helped create AI firsthand, yet may now be among the first to be displaced—and he himself has no answers for where computer science graduates should turn or what kind of people the AI era truly demands.