JetBrains bets GPT-5 can make 15M devs architects, not typists

JetBrains, the Czech-based company whose tools are used by roughly 15 million developers and 88 of the Fortune 100, is making a calculated bet that AI won’t replace engineers but will fundamentally change the nature of their work. In a recent conversation, Head of Product Kris Kang detailed how OpenAI’s models—including GPT-5 and Codex—are being woven directly into the fabric of products like the new Junie coding agent and the existing AI Assistant. The pitch isn’t raw speed or flashy demos. Kang draws a sharp line in the sand: ‘It’s not just about generating code—it has to be safe, readable, and maintainable.’ That’s a refreshingly boring, deeply pragmatic benchmark in a market obsessed with one-shotting Todo apps.

The real shift, as Kang sees it, is about delegating the soul-crushing context-switching that drains engineering teams. He describes a workflow where agents, backed by GPT-5, are successfully completing tasks he himself assigns, noting with some surprise that many complex tasks are handled successfully. The internal philosophy at JetBrains is a hybrid one: let AI draft the boilerplate, the tests, and the documentation, while human engineers lock in on system design, architecture, and critical review. It’s a ‘force multiplier’ play, where the value of a developer’s clear intention skyrockets because the execution cost plummets.

Kang’s leadership playbook amounts to an anti-hype manual. Start by automating the friction points humans hate—documentation, reviews, hand-offs. Aggressively protect deep work, because context-switching is a bigger silent killer than slow typing. And run experiments that compound. He draws a practical distinction between the two modes of AI interaction: ‘Chat gives you a lift. Agents give you a step-change.’ One helps you think; the other gets things done.

The destination being charted here is not a world with less work, but a redefinition of what ‘work’ means for a developer. Instead of grinding through implementation details, engineers become the designers, guardrail-setters, and reviewers of an AI-powered process. It’s a vision of shipping faster with more confidence, where the winners are those who learn to experiment efficiently. As Kang warns, the gap between those who figure this out and those who don’t will only widen over time.