Cognitive overload, mental energy, and AI

The problem isn’t doing more

AI has lowered the cost of producing anything: text, code, images, ideas, options.

The result isn’t more work done in less time. It’s more output than our brain can evaluate.

It’s the same dynamic as FOMO, but reversed. With social media, we were missing out on other people’s experiences. With AI, we don’t have enough time to process our own production.

The bottleneck is no longer production. It’s our attention capacity.

What is cognitive overload

Cognitive overload occurs when information, decisions, and stimuli exceed the capacity of working memory.

It’s not about intelligence. It’s a characteristic of the human brain: everyone has a limit — and the limit is lower than we think.

The signs are recognizable:

• difficulty making decisions, even simple ones
• a feeling of “full mind” or mental block
• basic mistakes on familiar tasks
• loss of focus even on simple things
• multitasking that makes everything worse
• procrastination as a defense mechanism

From overload to burnout

When overload becomes chronic, it follows a predictable trajectory. It’s not a sudden fall — it’s a slow erosion.

The critical point isn’t the final burnout. It’s the moment when recovery stops happening.

Cognitive inflation

The human brain doesn’t scale. AI output does.

Every AI tool lowers the marginal cost of generating: more content to read, more options to evaluate, more decisions to make, more alternatives created automatically.

We’re not less capable. We’re simply out of range compared to the volume AI can produce.

The specific case: code agents

In technical work, code agents push this dynamic to the limit.

Production accelerates. Evaluation capacity doesn’t.

The risks you don’t see

Code agents introduce cognitive risks that are hard to notice from the inside:

• Loss of mental model — the agent changes the code, but your internal map falls behind
• Automation bias — you tend to trust the output without verifying enough
• Alternative explosion — every prompt generates variants, every variant requires a choice
• Attention fragmentation — you jump between contexts without ever closing a loop
• Invisible technical debt — the code works, but nobody really knows why

Cognitive energy

Cognitive energy is the mental budget available for:

• attention — filtering noise and selecting what matters
• working memory — keeping open contexts and decisions in mind
• decision making — choosing with incomplete information
• executive control — planning, adapting, suppressing impulses

It can’t be measured directly. But you feel it when it runs out.

The model

Every feature in development, open PR, ongoing AI conversation, pending decision, bug in progress — it all accumulates in the second term. The math is what it is.

Practical strategies

It’s not about doing less. It’s about doing better:

• Limit cognitive WIP — fewer open contexts at the same time
• One agent, one objective — clear and bounded task for each session
• Review before generating more — understand before asking for anything else
• Reduce alternatives — fewer options, less decision load
• Timeboxing — sessions with defined start and end times
• Decide what NOT to do — exclusion is often the highest priority
• Document decisions — writing down “why” offloads working memory

In summary

Production is no longer the limit.

It’s that it produces more than we can evaluate — and the human brain pays the price.

Recognizing this asymmetry is the first step toward working sustainably with AI tools.

Let me know if you recognize these patterns in your daily work.