Function call completion

Ideas Feedback
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I consider it a bug (but you may not) for the AI suggestion to be a function call (i.e. for the AI to suggest source code that amounts to a function call), without other code that has side effects, if the suggested function call does not conform to that function’s signature (though semantic non conformance is more excusable than lexical & type nonconformance).

Another case that is on the margin (for me) is when the non-conforming function call is a first step in a multi-step transition to a new signature. E.g., when the AI suggests including an additional parameter. However, in that specific example, I’d think two completions are in order … first, to the conforming call, then to the added parameter version.

This stepwise approach would IMHO resolve a relatively irksome behavior I often run into, where pre-AI completion heuristics would give me what I want, but post-AI I have to accept the wrong completion and then break my flow with an excursion to revert the wrong portion. The reason it’s irksome is that pre-AI completion technologies tend to get function call completions

Steps to Reproduce

code with cursor.

Expected Behavior

see description

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.3.29 (Universal)
VSCode Version: 1.105.1
Commit: 4ca9b38c6c97d4243bf0c61e51426667cb964bd0
Date: 2026-01-08T00:34:49.798Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Darwin arm64 24.6.0

For AI issues: which model did you use?

Sonnet 4.5

For AI issues: add Request ID with privacy disabled

completion

Additional Information

python

Does this stop you from using Cursor

No - Cursor works, but with this issue

3

Hey, thanks for the feedback on Cursor Tab completion behavior.

You’re right that this is more of a request to improve suggestion quality than a classic bug. Your idea of a step-by-step approach to changing function signatures is interesting: first suggest a call that matches the current signature, then a version with additional parameters.

So the team can better understand the issue, could you share:

  • Specific examples of functions and the incorrect suggestions
  • Screenshots or a recording of what you’re seeing

That’ll help the engineers reproduce and analyze the problem.

4

If I have the time I’ll provide what you’ve asked for but … really … I was careful with my wording:

Completions that are simply function calls without side effects (this is a very selective description of the target completions) should be checked against the signature (this is again a very tractable and selective test) in order to modulate the completion UX behavior…