Is autocomplete (Tab) being developed or planned further? Or is all the focus now on agents?

Discussions
1

Hi! I have a question for the developers. We live in strange times, when it’s not entirely clear to what extent code understanding still needs to be supported, and to what extent we can trust agents. What I do know for sure is that half a year ago, when I checked, Cursor had the best autocomplete out of everything I compared. Unlike everything else, it made very good use of even closed files and files that had not been opened in a long time when generating completions.

I don’t know exactly what the future will look like. I think I even saw somewhere that the Cursor developers themselves said that over the past few months users have already massively switched to agents, and that everyone is starting to care less about autocomplete. At the same time, there’s another view: for more or less serious projects, code understanding will still matter. Careful code work, refactoring, cleaning up unnecessary parts - all of that will continue to be done by programmers for a long time, unless the project is very small and written through vibe coding. And in that case, autocomplete is exactly the best assistant.

In light of this, I’d like to ask: have you completely abandoned it? Your latest update about autocomplete development is here: https://cursor.com/blog/tab-rl. But autocomplete also needs to be developed, updated, maintained, and, hopefully, made even better. Am I understanding correctly that since then nobody has been working on autocomplete, and the entire focus is on agents? Or maybe this really is already a thing of the past?

I’m genuinely interested in the Cursor developers’ opinion on this. So my question is basically twofold: what are your plans for autocomplete after September 2025 - are you updating it, do you plan to update it? And what do you generally think about all of this - do you recommend moving on from autocomplete, or not?

Thanks :slight_smile:

4

Hey, thanks for the request. Good point that there’s a lot of hype around agents right now.

If I keep it short: Tab isn’t going anywhere, and it’s still actively improving alongside agents. Internally, it’s still one of our key products. We’re constantly working on prediction quality, latency, and model training. We just don’t write a separate blog post for every step. The tab-rl blog post Improving Cursor Tab with online RL · Cursor was a big public update on the training pipeline, but it wasn’t the final stop. We’re still investing in Tab.

On the broader question of whether we should move beyond autocomplete: no, we don’t think autocomplete is outdated. Agents and Tab cover different use cases. Agents are great for bigger tasks and multi-step changes. Tab is still essential for careful local edits, refactors, and fast iteration. Most users use both.

We announce specific plans and features via the changelog https://cursor.com/changelog and the blog when they’re ready. I can’t share the roadmap ahead of time, but you can be confident Tab is continuing to improve.