I'm an experienced developer and rely heavily on Cursor's Tab autocomplete, more than its Agent or Plan features. I mainly use Ask mode when exploring unfamiliar repositories or open-source projects, while Agent is mostly useful for quickly building proofs of concept.
Recently, though, Cursor has become increasingly frustrating. I've run into repeated Code Actions bugs, such as import and linter fixes failing when triggered with Cmd+. The Ctrl+Space import suggestions also behave strangely: scrolling down with the arrow keys eventually jumps back to the top, even after restarting the editor. There are also security concerns around the company's recent acquisition, particularly for my current team.
I've tried Windsurf and Zed before but ended up returning to Cursor because it felt more robust. I'm not looking to complain about Cursor—I'd really like recommendations for alternatives, especially tools with reliable inline autocomplete or tab-completion features. What are you using?
4 Answers
GitHub Copilot’s early autocomplete was probably the best I’ve used. It was fast, restrained, and usually just finished the statement you were writing instead of generating a huge block of speculative code. A lot of newer tools either suggest too much unrelated code or produce inconsistent completions. The newer inline editing features are useful, but the overall experience can feel cluttered compared with the older, simpler approach.
For a straightforward setup, I’d use either Zed or VS Code with GitHub Copilot inline suggestions. I keep my main conversational coding work in Claude Code sessions inside the editor, while Copilot handles the smaller autocomplete tasks.
When I tried Copilot a couple of years ago, it didn’t seem very aware of the files I had open or the text on my clipboard. Has its context awareness improved since then?
I use GitHub Copilot with VS Code and mostly stick to autocomplete rather than agent-style coding tools. It’s not perfect, but it handles everyday completions well enough and keeps the workflow relatively simple.
Windsurf is the editor my consulting clients most commonly standardize on. It has been a solid option for them, although those teams generally have enough budget to pay for the higher-tier setup and licenses.
I remember the older Copilot experience being remarkably accurate and a few steps ahead without taking over. Modern models often seem trained to keep extending the suggestion, when sometimes I only want them to complete the current line or expression.