How does AI handle TODO and deprecated comments?

If there are "//TODO" or deprecated comments in the code, how will the AI remind users to pay attention?

Perfect! Now I have the full picture. Let me provide a comprehensive answer based on what I found.

##Answer

Based on my analysis of the codebase, the current implementation does not have a specific mechanism to remind users about TODO or deprecated comments. However, here's what I found:

###Current Handling of Deprecated Code:

1. For Deprecated Models: The system tracks deprecated AI models using a deprecatedAt field in the database (apps/web/lib/db/schema.ts), and displays a visual warning to users:

   {!!llmUser.deprecatedAt && (
     <p className="mt-1 text-amber-600 text-sm">
       This model was deprecated on{" "}
       {new Date(llmUser.deprecatedAt).toLocaleDateString()}
     </p>
   )}
   

   This shows a subtle amber-colored message on the model's profile page.

2. For TODO Comments in Code: The system currently has no automated mechanism to detect and remind users about TODO/FIXME comments. The test file (apps/web/agent/tools/__tests__/tools.test.ts) shows that the Grep tool can search for TODO comments, but it's just a capability—not actively used by the AI to flag issues.

###How the AI Agent Works:

The AI assistant follows system prompts defined in apps/web/agent/response-agent.ts:

• ASK_SYSTEM_PROMPT: Instructs the AI to ground answers in source code
• BUILD_SYSTEM_PROMPT: Instructs the AI to make code changes via PRs

The system prompts emphasize exploring source code and documentation, but don't explicitly ask the AI to search for or highlight TODO/deprecated comments.

###Potential Enhancement:

To implement automatic TODO detection, the system could:

1. Add a prompt instruction telling the AI to scan for TODO/FIXME/deprecated comments when exploring code
2. Create a dedicated tool that automatically detects and highlights these annotations
3. Surface warnings similar to how deprecated models are displayed

This would make the codebase quality issues more visible to users seeking help.