Real-World Observation Drills

These drills make learners inspect real systems without needing production access. The goal is to connect concepts to things they can actually see.

Open a website you use often.

In browser developer tools:

• Open Network tab.
• Reload.
• Pick one document request.
• Pick one API or data request if visible.
• Record method, status, headers, and response type.

Answer:

• What loaded first?
• Which requests were static assets?
• Which requests looked like data?
• Did any request fail?
• What would a backend log need to correlate this request?

Call a public API or a local API with curl.

Record:

• URL.
• Method.
• Status code.
• Response headers.
• Response body shape.
• One changed header or query parameter.

Answer:

• What is the contract?
• What errors are documented or visible?
• What would make this API hard to integrate?

Pick a small public repository.

Inspect:

• README.
• Dependency file.
• Start command.
• Test command.
• Main source folder.
• CI workflow if present.

Answer:

• What does this project do?
• What runtime does it need?
• Where does execution begin?
• How would you run tests?
• What would you ask an agent to inspect first?

Use any web app with a form.

Observe:

• Empty state.
• Typing state.
• Validation error.
• Loading state if visible.
• Success or failure state.

Answer:

• Which state is local UI state?
• Which state likely comes from the server?
• What should be stored durably?
• What could go wrong on double submit?

Read any public engineering incident postmortem.

Extract:

• Trigger.
• Impact.
• Detection method.
• Root cause.
• Fix.
• Prevention.

Answer:

• Was the failure technical, process, or both?
• What signal detected the issue?
• What would have caught it earlier?

Pick any AI feature you use.

Map:

• What user input it receives.
• What context it likely uses.
• What output it produces.
• What happens if output is wrong.
• Whether a human approval path exists.

Answer:

• What should be logged?
• What should be hidden from the model?
• What should stay human-controlled?
• How would you evaluate quality?

Create real-world-observation.md in the portfolio with:

• Three drills completed.
• Screenshots or command output where useful.
• Diagrams for at least one observed flow.
• Three things you understand better now.
• Three questions you still have.