Drawing on 43 years of programming experience, this perspective argues that AI coding assistants won't replace developers, just as previous technologies like Visual Basic, no-code platforms, and 4GLs didn't. The core challenge of programming remains translating ambiguous human thinking into precise computational logic, not just writing syntax. Historical patterns show new tools increase both programs and programmers (Jevons Paradox). Unlike reliable past tools, LLMs often slow teams down and produce unreliable code. The author predicts hyper-scale LLMs may not be economically viable long-term, with modest AI tools assisting rather than replacing developers.
2 Comments
Sort: