Overview
- On March 13, Perplexity CEO Aravind Srinivas wrote “Well said” on X, backing a student’s post arguing LLMs now handle much of the boilerplate in coding.
- The post has drawn over 15,000 likes and nearly a million views, pushing the discussion about changing software work into broader coverage.
- Anthropic CEO Dario Amodei has suggested models could do most end-to-end software engineering within six to twelve months, with some Anthropic engineers already editing model-written code.
- Reported measures of impact include a 2023 Microsoft study showing GitHub Copilot users finished tasks 55.8% faster and Anthropic’s index estimating roughly 75% task coverage for programmers.
- Educators and practitioners foresee curricula tilting toward math, systems thinking and rigorous verification, while experts note LLMs still stumble on novel system design and require senior oversight.