Aravind Srinivas, CEO of Perplexity AI, added his weight to a growing conversation about what AI is doing to software engineering. On March 13, he quote-tweeted a post by physics and AI/ML student @TheVixhal with two words: “Well said.” The original post, which racked up 15,000-plus likes and nearly a million views, argued that LLMs are quietly automating the grunt work of coding—and in doing so, pulling computer science back toward its mathematical and physics-heavy roots.The idea isn’t fringe. Anthropic CEO Dario Amodei has said the industry is 6 to 12 months from AI handling most of what software engineers currently do end-to-end. Some engineers at Anthropic, he noted, no longer write code at all. Replit’s CEO put it more bluntly: the software engineering role, as currently defined, “sort of disappears.”

GitHub Copilot boost developers speed by 55%—but the gains are uneven

The data backs the shift. A 2023 Microsoft-run experiment found developers using GitHub Copilot completed tasks 55.8% faster. Anthropic’s own AI Exposure Index puts programmers at roughly 75% task coverage by LLMs—the highest of any profession tracked.What’s changing isn’t just speed. It’s what engineers spend time thinking about. Boilerplate is increasingly off the table. The questions that remain—how systems fail, what trade-offs to make, whether the architecture holds at scale—are closer to physics and math than to typing syntax correctly.

LLMs still hallucinate on complex system design, keeping senior engineers in demand

Not everyone’s convinced the transition is clean. Critics point out LLMs still stumble on novel, complex problems. Junior developers gain the most from these tools; senior engineers remain essential for verification and judgment calls. The “6 to 12 months” timeline Amodei floats applies to existing tasks, not the harder work of inventing new systems.Still, the direction of travel is clear enough that even Code.org’s founder is rethinking CS education—less syntax, more logical reasoning. “Coding is dead,” he said. “Long live coding.“