Get the latest tech news
Beyond the 70%: Maximizing the human 30% of AI-assisted coding
Why durable human skills matter in the age of AI-assisted coding
For example, an AI might give you a function that technically works for the basic scenario, but it won’t automatically account for unusual inputs, race conditions, performance constraints, or future requirements unless explicitly told. Steve Yegge wryly likens today’s large language models (LLMs) to “wildly productive junior developers” – incredibly fast and enthusiastic, but “potentially whacked out on mind-altering drugs,” prone to concocting crazy or unworkable approaches. (A cautionary note: if junior staff simply throw raw AI output over the wall to you, push back – instill a process where they must verify AI-generated work first, so you’re not the sole safety net.
Or read this on Hacker News
