Keeping Concentration in the Evening When AI Removes Your Downtime
Keeping Concentration in the Evening When AI Removes Your Downtime
There is a strange paradox with AI coding agents. They handle the boring, mechanical work - boilerplate, repetitive refactors, test scaffolding - which sounds great until you realize that boring work was your brain's rest period.
The Paradox of Constant Evaluation
Before AI agents, a typical coding session naturally alternated between hard thinking and easy execution. You would design an approach (hard), then spend 20 minutes typing it out (easy). Your brain recovered during the easy parts.
Now the agent writes the code and you evaluate it. Evaluation is cognitively expensive - you are reading someone else's logic, checking for subtle bugs, verifying architectural decisions. Every interaction demands your full attention.
By evening, you have been doing nothing but high-cognitive work all day. The easy tasks that used to give your brain micro-breaks are gone.
Why Evenings Hit Harder
Evening coding sessions used to work because you had accumulated enough mental rest throughout the day. Now that rest is eliminated. You sit down at 7pm and your brain is already depleted from 8 hours of continuous evaluation.
The symptoms are specific: you start approving AI-generated code without fully reading it. You miss edge cases you would have caught at 10am. You accept "good enough" when the code has real problems.
Managing the New Cognitive Budget
A few strategies that help:
- Timebox evaluation sessions. After 45 minutes of reviewing AI output, take a genuine 15-minute break. Not scrolling - actually resting.
- Alternate task types. Switch between code review and creative work like writing docs or sketching architectures.
- Set an evening cutoff. If your brain has been evaluating AI output since 9am, stop by 6pm. The quality of your reviews after that point drops sharply.
- Use the agent for evaluation too. Let it run linting, type checking, and test suites before you look at the code. Reduce the surface area you need to manually verify.
The goal is not to use AI agents less. It is to recognize that they shifted your cognitive load, not reduced it, and budget your energy accordingly.
Fazm is an open source macOS AI agent. Open source on GitHub.