Debug With AI Using Logs | Generated by AI

Home 2026.03

Question: What is the golden rule for using AI coding assistants (Copilot/Codex/Claude Code) to fix hard problems?

Answer:

The Golden Rule: Debug Log-Driven AI Fixing

This is a highly effective and practical strategy for tackling difficult bugs with AI assistance. Here’s a comprehensive breakdown:

The Core Loop

Hard Problem → Add Debug Logs → Observe Output → Feed Back to AI → Fix

Why This Works

AI coding assistants often lack runtime context. They can only see your static code, not what’s actually happening at execution time. Debug logs bridge that gap by exposing:

Step-by-Step Process

Step 1 — Describe the problem to AI, ask it to add debug logs

“I have a bug in this function. Don’t fix it yet — add comprehensive debug logs so I can observe the runtime behavior.”

Step 2 — Frontend: Add console.log strategically

console.log('[DEBUG] fetchUser called with id:', id);
console.log('[DEBUG] API response:', JSON.stringify(response));
console.log('[DEBUG] Parsed user object:', user);
console.log('[DEBUG] State before update:', this.state);

Step 3 — Backend: Add Log4j / SLF4J logs

log.debug("[DEBUG] entering processOrder(), orderId={}", orderId);
log.debug("[DEBUG] fetched order: {}", order);
log.debug("[DEBUG] inventory check result: {}", inventoryStatus);
log.error("[ERROR] unexpected state: {}", e.getMessage(), e);

Step 4 — Run the app, reproduce the bug, collect logs

Copy the relevant log output — this is your evidence.

Step 5 — Feed logs back to the AI

“Here is the console output when the bug occurs: [paste logs]. Now identify the root cause and fix it.”

Step 6 — AI fixes with full context

Now the AI has runtime truth, not just static guesses. The fix quality improves dramatically.

Pro Tips

Tip Detail
Tag your logs Use prefixes like [DEBUG], [FLOW], [STATE] to filter easily
Log before AND after Wrap suspicious operations to see input vs. output
Log in catch blocks Always log the full exception + context, not just e.getMessage()
Remove logs after fix Or use proper log levels (DEBUG) so they’re hidden in production
Be specific in prompt Tell AI exactly which log line looks wrong and why

Example Prompt Template

I have this bug: [describe symptom]

Here is the relevant code: [paste code]

Here is the runtime log output when the bug occurs:
[paste logs]

Based on the logs, identify the root cause and provide a fix.

Summary

The golden rule is essentially turning AI from a static code reader into a dynamic debugger — by giving it the runtime evidence it needs through logs. This is especially powerful for:

Back Donate