AI Coding Tools Push Developers From JavaScript to Judgment Calls

The daily work of software development is undergoing a structural transformation as AI coding assistants move from experimental tools to production infrastructure, forcing technology companies to rethink what it means to be a developer.

At Google, engineers are spending less time writing code and more time evaluating the output of AI agents, according to Julian Togelius, a researcher studying the shift. The change mirrors a broader pattern across the industry: Anthropic engineer Boris Cherny introduced an internal coding tool to colleagues via Slack in 2024, marking an early step in what has become a wholesale reimagining of the software development lifecycle.

More than half of developers now use AI not just to write new code but to create test cases, analyze data, and debug software, according to Google's Dora research unit. The most successful practitioners, Togelius observed, are those with people-management experience—skills in context-switching and writing high-level instructions translate directly to orchestrating multiple AI agents.

"This is quite a different skill from just writing the code," Togelius said. The shift has made developers feel "super powerful," with 80% reporting increased productivity in Dora's 2025 survey. But the constant task of monitoring multiple agents also introduces new risks of burnout.

(The transition is reshaping hiring priorities and training programs at major technology firms, with judgment and system design increasingly valued over fluency in specific programming languages.)

The implications extend beyond individual productivity. Filmmaker and producer Peter Ho-sun Chan, speaking at Hong Kong's Filmart conference, warned that AI's encroachment on creative and technical work could push human creators toward niche markets. "AI won't be an enemy to auteur film—it will be an enemy to mediocre blockbusters," Chan said, suggesting that commercial filmmaking might soon be dominated by algorithmic production while human directors retreat to arthouse work.

The coding-tool arms race has become a competitive flashpoint among AI labs. Anthropic's internal tooling efforts, revealed through Cherny's 2024 Slack message, position the company alongside rivals racing to capture the developer market—a contest that has already seen OpenAI acquire Python toolmaker Astral and launch plans for a desktop "superapp" combining ChatGPT, its Codex app, and a browser into a unified development environment.