Deep dive · module 67 of 100
Logging and improving errors
Keep track of where AI makes mistakes and adjust your approach.
After this lesson
You create a simple way to recognise and prevent recurring errors.
In view · An improvement loop
An error becomes a learning point
Note error, situation, and consequence, without sensitive data.
Check if the same error recurs in similar tasks.
Adjust prompt or checkpoint and test if the error occurs less often.
Part 1 of 3 · Record
Your preference for less movement is on. Choose each part yourself.
Log the error
Note what went wrong and in which situation.
Find the pattern
Does the same error occur more often?
Adjust the prompt or workflow
Add a boundary or checkpoint to prevent the error.
New practice example
From explanation to a usable result.
A usable error log preserves enough context to retest the same error later.
Fictional practice situationA fictitious AI summary changes “the reading room is closed” to “the library is closed”. The problem occurred during a short task without source constraints.
View a possible solution
This is how it can be done
Log: input with source, old prompt, error fragment, expected formulation and suggested change. Afterwards, add the same test to your checklist and see if the adjusted prompt preserves the exception.
Why this works
Only saying “AI did it wrong” does not help you improve. A repeatable test makes it visible whether the change actually helps.
Make it slightly more difficult
Create a second case in which an amount is added. Which fixed fields can you use for both errors in the same log?
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
Only saying “AI did it wrong” does not help you improve. A repeatable test makes it visible whether the change actually helps.
A step further
Create a second case in which an amount is added. Which fixed fields can you use for both errors in the same log?
A prompt to get you started
Create a simple log form for AI errors with fields: date, task, what went wrong, and what you adjusted.
Apply it
The same invented discount appears in three drafts. You remove it manually every time.
Which approach best fits this situation?
- Keep correcting without recording the cause.
- Log the error type, improve the instruction and test again on fixed examples.
- Ignore the error because you can easily remove it.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
Why is it useful to log AI errors?
- To see patterns and improve your prompts or workflow.
- Because it is mandatory.
- It is useless.
Extra practice
Describe one error you have seen and how you would prevent it in the future.
Error: AI added a discount. Adjustment: added “do not invent discounts or prices” to the prompt.
