5.2Human attention is the bottleneck

You get one focus token

Your agents can run well in parallel, but you, dear human, cannot.

Your one brain can focus on one thing at a time. You cannot literally read two different agent plans, or two different code bases, at once.

When you focus on a thing, you place your focus token on it. When you switch to focus on another thing, you move your focus token. This is a focus token switch.

Focus token switches have a transaction fee

It turns out you pay a little bit every time you switch your focus token.

You don't pay this fee in dollars. You pay it in mental energy.

Suppose your focus token has been on Agent A, and now you move it to Agent B. Your head is still full of A. To think about B, you have to push A out and rebuild B: what it's for, what you told it, what worried you last time. So you scroll up through B's history. What was it working on? What did I tell it? Which file was the problem? Only then do you know enough to actually read its plan.

That rebuild is real mental work, and you pay it every time you switch. That's the fee.

There's a surcharge, too

It turns out your brain never fully pushes A out. Part of your attention stays stuck on A, even though you're working on B. Psychologists call this attention residue.

So you read B's plan with less than your full brain, and you skim right past the line where B quietly deletes a failing test.

The fees add up

We feel tired at the end of a long working session because we've spent our mental energy and are ready to rest. We might start the day at, say, 90% capacity and end it at 20%. Sometimes we end it at negative capacity (!) and need undertime (basically, more recovery time than usual) to make up the difference.

Every switch draws down that budget. By 4pm, rebuilding B takes longer, and once you've done it, you read its plan twice and still can't tell if it's right. So you approve it anyway.

Be a smart investor

Financial investors look for places to put their money where it will turn into more money. They balance risk (potential downside) with return (potential upside).

We software engineers are "mental energy investors." We decide where to allocate our finite mental energy. We spend it on focus and on focus token switches, and we hope to get impact in return.

A wise mental energy investor spends as much of their energy as possible driving impact, and as little as possible on switching fees. And like any good investor, they make decisions based on insights about the market. The rest of this section covers the insights that matter most here: how to manage your review queue, how to batch your switches into review checkpoints, and how to keep notifications from moving your focus token for you.

What this section will cover

• Human attention is the bottleneck: review queues, batching, notifications