The Altair emulator experiment
According to a report by IT之家, engineer Wang Yonggang, formerly at Google, published on November 6 a long essay titled “Goodbye, Programmer,” in which he recounts his experience of AI-assisted software development. The starting point was a request received in September from a U.S. computer science professor, who wanted to improve an emulator of the Altair 8800 computer that he had open-sourced six years earlier.
Wang entrusted the request to Claude Opus 5.5, also asking for suggestions on the system architecture. According to his account, the tool did not merely fulfill the initial request: it autonomously proposed adding an assembler, a debugger, and an interface that simulates a paper-tape reader. Over roughly three or four days, the project reportedly grew from a few hundred lines to more than 30,000 lines, accompanied by unit tests and browser-based integration tests.
At this stage, according to the author, his role focused on validating proposals, asking questions, and giving feedback, while most technical decisions and code writing were delegated to the AI. This is a declarative account, not independently verifiable on the basis of the available material: no metrics, repositories, or comparisons with development conducted without assistance have been published.
Beyond code: research and contacts
The aspect Wang calls most surprising concerns activities ancillary to development. Again according to his account, the tool autonomously consulted historical sources, sought out experts, facilitated email exchanges in multiple languages, and used external materials to calibrate 3D models. In the project to reconstruct Leibniz’s mechanical calculator, the AI reportedly helped contact a German museum, from which 135 sheets of historical drawings and manuscripts were obtained.
These episodes shift attention from code alone to the AI’s ability to handle research and communication steps that traditionally require time and cross-cutting skills. However, they remain descriptions provided by the author, without details on which museums, which experts, or which steps were actually handled by the AI and how many by the human.
The thesis on programmers’ work
Wang’s reflection goes beyond the individual case. He argues that, as AI’s programming capabilities improve, the technical experience accumulated by programmers will see its value redefined. He predicts that development teams may evolve toward models in which a few product and business managers coordinate many AI agents, with reduced room for professional growth for traditional programmers.
The author arrives at a bitter reflection: a generation of programmers helped create AI and now risks being among the groups most exposed to its effects. He admits, however, that he has no answers on how new computer science graduates will be able to navigate the job market, nor on which skills will be required in the AI era.
The material does not make it possible to determine how representative the described experience is: it is an individual professional’s account, without data on the adoption of these tools in development teams or independent assessments of the quality of the generated code. The thesis on the downsizing of the programmer’s role therefore remains the author’s prediction, not a documented outcome.
