Alex Ewerlöf argues against the narrative that LLMs have 'solved' coding, emphasizing that while code creation has become cheaper, non-functional requirements like maintenance, reliability, security, and scalability remain major unsolved challenges. He also notes that even functional requirements are not fully automated, attributing overconfidence to the Dunning-Kruger effect.
Background
As LLM-powered coding assistants like GitHub Copilot become mainstream, many advocate that software engineering is shifting from writing code to reviewing it. This article pushes back against that narrative from the perspective of a veteran engineer and AI tool builder.
- Source
- Lobsters
- Published
- Sep 30, 2026 at 05:40 PM
- Score
- 7.0 / 10