The Mughal Emperor remained emperor long after he had lost all real power. By the 19th century, the emperor in the Red Fort controlled little beyond Delhi's walls while the East India Company, the Marathas, and regional nawabs carved up the subcontinent between them. And yet the title persisted. Successor after successor took the throne, issued firmans, held court. The construct of "emperor" outlived the empire itself.
This isn't unusual. We hold onto roles, titles, and structures long after the conditions that gave them meaning have changed. It comes down to a distinction between two kinds of reality.
There's physical reality — gravity, thermodynamics, the things that don't care what you think about them. And there's social reality — the narratives we construct about our world and who we are. Career paths, org charts, the "shape" of what it means to be a software engineer. These aren't natural laws. They're agreements we've made with each other. But once established, we forget that. Social realities start to feel as immovable as bedrock — at some point we stop seeing them as things we invented and start treating them as things that have always been there. The Mughal court didn't keep performing its rituals out of cynicism. It kept performing them because that was simply what one did. The fiction had become invisible.
And then something changes and the fiction is exposed. When a tool can produce working code, pass review, and ship features, it gets harder to maintain the idea that the core of a software engineer's role is writing code. When agentic tools can explore a codebase, propose solutions, and iterate on feedback, the old org structures — built around the assumption that engineering time is the bottleneck — stop making sense.
The frame was never you. It helped you navigate the world for a while. We built identities around these structures and letting go feels like losing something real. But when the world changes, clinging to the frame doesn't preserve anything. It just widens the gap between the social reality you're defending and the physical reality you're living in.
I've started thinking of this as a digital garden. The metaphor is that working with AI is less like directing construction and more like tending a garden. You can't force a plant to grow. You create the conditions — good soil, the right light, enough water — and then you let growth happen. Your role shifts to pruning after the fact, shaping what's already emerged rather than dictating every step upfront.
It's a small reframe but it captures something about how the work is changing. The engineer's value shifts from the act of writing code to knowing what to plant, preparing the ground, and having the judgment to prune well. This may not be the right frame but it's an attempt to build something new rather than defend something old.
Nearly three months into 2026. I finished The Radetzky March by Joseph Roth and Secondhand Time by Svetlana Alexievich.