← ContentsHome
Chapter 4

Calculate Your Context Tax

A practical worksheet to measure the hidden costs in your own AI coding workflows.

Before we talk about solutions, let's make this personal. What's your context tax?

Here's a simple worksheet to estimate it.

The Context Tax Calculator

Step 1: Track one debugging session

Next time you debug an issue with AI assistance, track these numbers:

Metric Count
Alt-tabs to AI tool ___
Times you copy-pasted code to AI ___
Times you copy-pasted AI response back ___
Manual file searches to find context ___
Prompts where you re-explained context the tool could have seen ___
Minutes from AI interaction until next meaningful edit ___

Step 2: Estimate time costs

Use these rough averages (or time yourself):

  • Each alt-tab: ~30 seconds of disruption
  • Each copy-paste cycle: ~45 seconds
  • Each manual file search: ~2 minutes
  • Each re-explanation prompt: ~3 minutes to write
  • Reorientation time (minutes until next edit): track this carefully

Step 3: Calculate your tax

Context Tax =
  (Alt-tabs × 0.5) +
  (Copy-pastes × 0.75) +
  (File searches × 2) +
  (Re-explanations × 3) +
  (Total reorientation time)

Step 4: Compare to baseline

Time the same debugging session without AI. Just you, the code, and maybe asking a human.

AI Value = (Time without AI) - (Time with AI + Context Tax)

If AI Value is negative, the AI is costing you time, not saving it.

Example: Sarah's Session

Let's score Sarah from Chapter 1:

Metric Count Time Cost
Alt-tabs to AI tool 4 4 × 0.5 = 2 min
Copy-paste cycles 5 5 × 0.75 = 3.75 min
Manual file searches 3 3 × 2 = 6 min
Re-explanation prompts 3 3 × 3 = 9 min
Reorientation time — 22 min (tracked)
Total Context Tax 42.75 minutes

Her total time with AI: 70 minutes
Her context tax: 43 minutes
Her actual productive time: 27 minutes

If she'd asked Miguel directly (human, not AI): ~15 minutes.

AI Value for Sarah: -55 minutes

The tool actively hurt her productivity.

What This Reveals

When you actually measure, you often find:

  1. The AI interaction itself is fast (T_ai is small)
  2. The context switching is expensive (T_recover is large)
  3. The tool is net negative when context tax exceeds the value gained

This is why "fast inference" and "high acceptance rates" are misleading metrics.

A tool can have 95% acceptance rate and still destroy your flow.

Try This Next Week

Pick 3-4 typical tasks:

  • Debugging a test failure
  • Adding a small feature
  • Refactoring a module
  • Understanding unfamiliar code

For each task, do it once with your current AI tool (track the metrics above).

Then ask yourself:

  • When did I feel most productive?
  • When did I feel like I was fighting the tool?
  • What context did I have to manually provide that the tool should have known?

The answers will tell you how much tax you're paying.