Four thousand hours. That's how much time software engineer Colin Breck logged in 2025 while simultaneously shipping a cloud migration, an edge computing platform, and a distributed database. He also delivered three conference talks and hit two contract deadlines with liquidated damages hanging over the work. AI, he says, made it possible. But not in the way you'd expect.
Breck didn't use AI to generate most of his code or writing. Instead, it worked as a debugging companion, a prototyping accelerator, and a trade-off evaluator that never slept. "AI kept me going when I otherwise would have been stuck," he writes. In one case, AI reproduced and fixed an OPC UA server performance issue at scale. In another, it spotted a DNS configuration change buried in application logs that human reviewers had missed entirely.
This is the reality of AI-assisted work in 2026 that rarely makes headlines: less "AI writes your code" and more "AI catches the thing you'd have spent three days hunting."
The Flow State Problem
Breck raises a concern that most AI productivity coverage glosses over. Programming has always drawn people partly because of flow states - that deep, immersive concentration where hours vanish and complex problems click into place. Supervising AI agents is a fundamentally different cognitive experience. It's passive, fragmented, and defined by context switching between parallel tasks rather than deep focus on one.
Breck noticed this firsthand: he had difficulty concentrating even while writing this essay. Managing multiple AI agents in parallel means your attention never settles. Bryan Cantrill, a well-known systems engineer, put it more bluntly: "Hats off to people who want to spend their lives acting as middle management for robots. But that's not necessarily for me."
This isn't an abstract philosophical concern. If the satisfying parts of technical work evaporate, the job becomes something fundamentally different - and not everyone will want to do it.
Mid-Career Engineers Are Quietly Walking Away
Breck observed something that doesn't show up in hiring data yet: experienced mid-career engineers leaving the industry despite strong demand for their skills. His theory is that this reflects an unconscious anxiety about professional identity. When decades of accumulated knowledge about code style, testing methodology, and naming conventions might become irrelevant overnight, the question "Where is my place?" becomes genuinely destabilizing.
Younger professionals face the same pressure, but financial necessity forces faster adaptation. Senior engineers with savings and options can simply leave - and some are choosing to.
Andrej Karpathy captured the broader sentiment: "I've never felt this much behind as a programmer. There's a new programmable layer of abstraction to master, and everyone has to figure out how to hold it and operate it."
The Honest Version of AI Productivity
Breck's account is valuable precisely because it resists the usual framing. He didn't 10x his output with a prompt. He worked an unsustainable number of hours, had strong team support, brought decades of industry experience, and used AI to stay unstuck during exhausting stretches. That's a more honest picture than most vendor marketing will give you.
He also makes a practical observation about organizations: "AI champions" who invest personal time learning these tools and sharing knowledge internally are disproportionately valuable right now. Most professionals haven't grasped AI's potential, blocked by security constraints, time pressure, or difficulty processing how fast capabilities are growing.
Breck explicitly notes he wrote the essay without AI assistance - "old-school" - and tried to find some flow while doing it. The fact that he felt the need to say so tells you something about where we are.