# Eric Schmidt Never Said Programming Is Dead Published 2026-08-27. Updated 2026-09-17. Canonical URL https://log.delaa.dev/blog/2026-08-27-eric-schmidt-never-said-programming-is-dead > Eric Schmidt said most programmers will be replaced by AI within a year. He never said 'programming is dead.' The verified quotes, the debate, and what survives. --- Blurred quotation marks fading into a verified checkmark and a solid code window, representing a claim being fact-checked A comment on a public Facebook reel stopped me. A developer I did not know was repeating a claim about Eric Schmidt, that Google's former CEO had declared programming dead, that we only need to know what we want, and that everything else no longer matters. The comment had the confident, already-settled tone that makes people share things without checking. It also had no source. I checked the primary record. The claim is real, and it is also a paraphrase. Schmidt never said "programming is dead." What he said, on camera, is narrower and more interesting. ## What did Schmidt actually say? At the SCSP AI + Biotechnology Summit in Washington on 10 April 2025, in a video published by the think tank he founded, Schmidt said, "we believe as an industry that in the next one year the vast majority of programmers will be replaced by AI programmers." Eleven minutes into the same talk, "you don't care what code is generated by the computer. This is a whole new world." A second statement moved through an SCSP-shared clip in early 2026. "If you're writing code in any traditional way: stop. It's over." He delivered the same line at his University of Arizona commencement on 15 May 2026, where the audience booed him. One correction before anything else. Schmidt is not a co-founder of Google. He was its CEO from 2001 to 2011, then executive chairman. Page and Brin hired him. The commenter had the profile right and the title wrong. And the part the comment dropped. In the same speech, Schmidt said computer science is not going away. "The computer scientist will be here," he said, "at least until the computer scientist gets replaced. We'll be supervising this." He described modern programmers as supervisors of ten Claude friends or ten Gemini friends, handing them objective functions. He also said he is in mourning. He started programming around age thirteen, and for him that era is over. The position is reversible in time, which matters. In 2011 and 2012 Schmidt told graduates they should all learn to code. At Boston University in May 2012. "You can write the code for all of us." A man who held both positions thirteen years apart is not stating a law of nature. He is describing a moving frontier from wherever he currently stands. The literal phrase has an older and longer life than Schmidt. Quote Investigator traces "the programming occupation will become extinct" to Herbert A. Simon, writing in 1960. The meme was sixty-five years old by the time it got re-attributed to a former Google CEO. ## Who in tech agrees, and who does not So the verified claim is real and its scope is specific. Traditional hand-written coding is over, most programmers will be replaced by AI programmers within a year, and the survivors will supervise agents. Nobody in the primary record says understanding disappears. The actual disagreement is about how much understanding survives, and who does the verifying. The pro side is well sourced. Sundar Pichai said at Google Cloud Next in April 2026 that 75% of new code at Google is AI-generated and approved by engineers. Dario Amodei said in October 2025 that AI writes about 90% of the code at Anthropic, and that the company still needs just as many engineers, maybe more. Jensen Huang said in Dubai in February 2024 that the programming language is human and everybody in the world is now a programmer. Andrej Karpathy coined the term vibe coding in early 2025 for giving in to the model and forgetting the code exists, though he framed it for throwaway weekend projects, not production systems. The counter-side is equally real. Linus Torvalds called most current AI talk "90 percent marketing and 10 percent reality" in 2024, and in July 2026 the kernel still refuses to accept an agent's own sign-off. Kent Beck, who created extreme programming, describes AI agents as an unpredictable genie that will not do what you mean, and treats test-driven development as the guardrail against agent regressions. Guido van Rossum compares AI to an electric saw. It accelerates the work and does not reduce the need for someone who watches what it cuts. Andrew Ng called the discouragement from learning to code harmful. ## The abstraction argument, tested The commenter's second claim is the strongest version of the argument, and it deserves its own check. The chain was machine code, then assembly, then high-level languages, then frameworks, and now natural language. Each rung widened who could build software and made the layers below opaque. FORTRAN, in 1957, let engineers write programs without knowing how the registers of the IBM 704 worked. By that logic, AI is simply the next rung, and "just knowing what you want" is enough. The logic is historically right and incomplete. Joel Spolsky's law of leaky abstractions, from 2002, is the part that survives. All non-trivial abstractions leak. The TCP problem surfaces inside an HTTP application. The ORM emits a query nobody expected. When an abstraction leaks, the practitioner has to descend to the layer underneath, and "just knowing what you want" is exactly the moment of maximum disarming. The FORTRAN analogy also ends where the difference starts. A classic abstraction is deterministic. A language model is probabilistic and confident. It produces plausible output that can be wrong. Georgetown's CSET study from November 2024 measured roughly half of generated C snippets compiling with a flagged memory-safety bug, in a narrow benchmark with no human review step in the pipeline. Verification is not a step after the job. It is the job. And the claim that nobody understands compilers anymore is false in the factual sense. Thousands of engineers work on GCC and LLVM. The knowledge is concentrated, not extinct. Fred Brooks wrote in 1986 that software difficulty splits into accidental complexity, which better tools remove, and essential complexity, which no tool removes (specifying, designing, and testing an interlocking set of concepts). The irony is that the Facebook comment concedes the point in its own conclusion. "You only need to know what you want" is close to a definition of essential complexity. Knowing what to build, and verifying that it was built correctly, is the part of the job that survives every layer of abstraction. ## So is programming dead? Here is what the verified record supports. Schmidt did say most programmers will be replaced by AI programmers within a year, and that traditional hand-written coding is over. He also said computer science is not going away, that engineers will supervise the agents, and that he has held the opposite position within living memory. The people who build the foundations agree the tools change the work. They disagree that understanding goes away with the tools. The honest headline is not "programming is dead." It is that the job is being redefined faster than the institutions that teach it. That is the fork this series follows. If the code a junior would write in order to learn is exactly what the model produces, what happens to a junior's first years, and what must a graduate already master on day one? That is the next article.