Every time you switch tasks, you pay a tax. It’s not the switching itself — it’s the rebuilding of context. The moment you put something down, the carefully constructed mental model begins to dissolve.
I have been thinking about this in terms of software systems, but it applies equally to thought.
Context is a stack
When a computer switches between processes, it saves its register state — the exact configuration of memory addresses, instruction pointers, and flags it needs to resume. Context switching is expensive because that save-and-restore cycle takes time, and because the data the previous process was using may no longer be in cache.
Human working memory works similarly, if less precisely. When you’re deep in a problem, you hold a constellation of related ideas simultaneously — the current state, the goal state, the constraints, the recent decisions. This constellation doesn’t persist when you look away.
process A: [solving problem] → interrupted → state saved
process B: [new task] → runs
process A: [resumed] → state restored → but the cache is cold
The cost isn’t losing the work. It’s the time to rebuild the constellation.
The myth of multitasking
There is no such thing as doing two things at once, at the level of focused cognitive work. There is only switching between them fast enough that it feels simultaneous. But the tax applies every time.
What we call multitasking is usually one of three things:
- Serial switching — rapidly alternating between tasks, paying the context tax each time
- Automated + deliberate pairing — doing one habitual task (walking, washing up) alongside one focused one
- Parallel low-stakes work — two tasks that each require so little attention that neither depletes the reserve
The second one is the only genuinely productive form. The first is almost always slower than just doing the things in sequence.
Interruptions don’t cost a minute. They cost the fifteen minutes it takes to rebuild the state you were in before.
Designing for context preservation
If context is expensive to rebuild, it makes sense to design systems — and days — that minimise the number of rebuilds.
Some things that actually help:
- Leaving a note to your future self. Before stopping work on something, write one sentence describing exactly where you are and what the next step is. Not for memory — for re-entry.
- Batching similar work. Context is partially domain-specific. Code review and writing require different constellations. Grouping similar tasks reduces the number of distinct states you need to rebuild.
- Protecting the first two hours. The morning context is the freshest. Using it for deep work and protecting it from meetings and messages has a compounding effect.
The goal isn’t to eliminate switching — that’s not realistic or desirable. It’s to make the cost visible, and to be deliberate about when you pay it.
Knowing the tax exists doesn’t make it cheaper. But it does make you a better scheduler of your own attention.