AI will make anyone a 10x programmer, but with 10x the cleanup

Conference takeaways: faster code, more chores

It has been reported that speakers from Netflix, Meta and IBM told an audience at the All Things AI conference in Durham that generative AI is wildly productive — but it doesn't come for free. Want to be a "10x" programmer overnight? Go ahead. You'll get 10x the output, critics warned, and 10x the cleanup. Efficiency begets more work, not less. Sound familiar? That's Jevons Paradox for the AI age.

Agents checking agents: the new workflow

Netflix UI architect Ben Ilegbodu described a setup that sounds like a tiny bureaucracy of bots: one agent writes code, another audits it, and a third orchestrates the pair. He calls it "adversarial code review." The idea is elegant — and exhausting. "At the end of the day, I'm actually kind of tired, because effectively, I spent the whole day talking to something," he said, underscoring the emotional toll of constantly managing machine helpers. It's automation that multiplies human oversight.

Context engineering: the hidden labor

Meta developer advocate Justin Jeffress warned of "context rot" — the drift that happens as an agent accumulates more inputs and priorities, and begins to fray at the edges. The fix? More upfront work: careful context engineering, prompt chaining, parallel agent orchestration, and even markdown checklists to keep the mission straight. In short: AI changes the shape of work. You're less a coder and more the conductor of an orchestra of hungry interns that never sleep — but sure do forget things unless you babysit them. The result is not the job-killer many fear, but a job transformer: faster code, new disciplines, and a fresh pile of cleanup.

Sources: The Register