Eric Schmidt, former Google CEO, recently made a confession that resonated through the engineering community: he's mourning the end of traditional coding. Speaking at an industry event, Schmidt declared that the craft he started as a teenager is effectively over. His message, reported by The Economic Times, carries weight because Schmidt now advises the Pentagon on AI and backs multiple AI coding startups. He's not selling a product; he's describing a shift he's funding.
Schmidt's central claim is blunt: if you're still writing code line by line, stop. He illustrates this with a story from one of his portfolio companies. A single engineer writes a spec and a test function by 7 p.m., then goes home. AI systems work through the night, finishing by 4 a.m. The engineer reviews results over breakfast. Schmidt estimates this workflow would have required six months and ten engineers at Google.
This captures the architectural shift. The bottleneck is no longer implementation speed but specification quality and evaluation precision. The best engineers now define success criteria and manage fleets of AI agents. They orchestrate. The machines iterate, debug, and deploy. Schmidt notes this accelerated sharply around October 2025. He advises managers to ask engineers still typing code: why are you working the way you did six months ago?
The technical implications for data and ML engineering are significant. Pipeline design shifts from writing transformation logic to defining objective functions for agent systems. Testing becomes a first-class specification artifact. The engineering trade-off is now between investing in agent orchestration frameworks versus traditional codebases. Companies that treat AI as an optional productivity enhancer miss the point. The advantage lies in how precisely you can define the problem. Get it wrong, and the AI optimizes toward the wrong target at machine speed.
Skeptics rightly point out that AI still hallucinates, struggles with novel architectures, and requires human oversight for production systems. Yet even at leading labs, models already write 10 to 20 percent of code, and that share grows monthly. The transition compresses what took agricultural and industrial workers decades into a handful of years.
For engineers, the practical path forward is clear: learn to manage agents, master specification writing, and treat AI systems as a team that never sleeps. Those who adapt will define the systems of tomorrow. Those who cling to the old identity may find their skills commoditized faster than expected. The craft evolves. The work continues. Just not with the same tools.
Source: Webpronews