Claude likes to talk about what’s been, not what is

Consider this diff suggested by Claude Code:

+    // Call OpenAI using new Chat interface
+    response, err := client.Chat(
       request.Ctx,
-      systemPrompt,
-      request.Message.Content,
-      tools,
-      executors,
+      ai.WithSystemMessage(systemPrompt),
+      ai.WithUserMessage(request.Message.Content),
+      ai.WithTools(tools, executors),
+      ai.WithToolActionLogCallback(logCallback),
     )

+    // Call OpenAI using new Chat interface
+    response, err := client.Chat(
       request.Ctx,
-      systemPrompt,
-      request.Message.Content,
-      tools,
-      executors,
+      ai.WithSystemMessage(systemPrompt),
+      ai.WithUserMessage(request.Message.Content),
+      ai.WithTools(tools, executors),
+      ai.WithToolActionLogCallback(logCallback),
     )

Notice the comment “Call OpenAI using the new Chat interface”.

Once I delete the old Chat interface then the new Chat interface will just be The Chat Interface. There will be no old one to consider. This just calls OpenAI. That’s it.

I guess I need to update CLAUDE.md to tell it to only ever comment on what is, not what has been. Comments should describe what happens now and why.

The callbacks here are to provide a list to the user of actions taken by the tools themselves. I was finding I did not trust the AI to have done what it said (although recent manual testing showed it was doing well). If the system reports the actual change made then I as a user can have confidence that my command was correctly actioned.