When using the Claude 3.7 thinking model, both Ask and Agent modes exhibit an issue where code blocks in responses don’t align with their descriptions. Upon careful examination, there are instances where the same code block gets repeated, causing all subsequent code blocks to shift positions. This problem existed in the early 0.46 version where code blocks would consistently shift by one position (due to duplicate display of one code block), and the latest 0.46.11 version now occasionally shows triplicate repetitions.
Interestingly, this misalignment appears to be purely a display issue. In Agent mode, the automatically applied code remains correct. However, in Chat mode—which lacks auto-apply functionality—the shifted code blocks make proper application impossible.
Attach
As shown in the figure above, all three code blocks are identical. Only the first block (an error message) should appear in its current position. The second and third blocks, based on semantic context, should contain AI-generated code responses. However, when continuing the conversation, the fourth code block that subsequently appears actually displays what should have been the code from the second position.
MacOS Cursor 0.46.11
Tell us if the issue stops you from using Cursor.