Why I can build software again

It’s hard to context switch, and even harder to context switch between different levels of thinking—such as moving from writing code to creating an engineering strategy. Let’s quickly explain why:

Attention residue makes context switching cognitively costly. When we shift tasks, part of our mind stays stuck on the previous one.

Abstraction conflict drains energy. Switching between the concrete (fixing a bug) and the abstract (planning Q3 strategy) forces your brain to shift gears. You have to change not just the topic, but your entire mode of thinking.

The maker-manager divide makes scheduling hard. If you’re in a manager’s schedule and try to deep work on building software, you’re fighting constant interruptions. You can’t easily change your schedule when you want.

---

I’ve found that building software with LLMs is different.

It’s not completely in the abstract realm, but the detail is less important. I’m directing more than building. I set context, review outputs, course-correct. I don’t need to hold many layers of stack in my head as the LLM is working at that level and holding them for me.

It fits around a manager’s schedule. I can get an LLM going on a task, get back to my other work, and check in occasionally. When I come back, I’m reviewing rather than rebuilding a mental model from scratch.

Yes there’s attention residue still. But for a certain class of technical work, this has opened up time that was previously inaccessible to me.