The four capabilities to know cold

Multi-line, multi-file predictive completion that anticipates the edit you are already making.

The everyday speed layer. Sub-second, fires as you type, zero ceremony.

This is the feature individual developers fell in love with before sales ever showed up.

Natural-language intent that turns into multi-file changes: it plans, reads the codebase, runs commands and proposes a diff you review.

The headline AI-native workflow and the thing executives have heard about.

Wins on a real task with a checkable end state: a refactor, a failing test, a ticket.

Retrieval across the whole repository so answers and edits are grounded in your actual code, not a generic guess.

This is what makes Cursor feel like it knows your repo instead of autocompleting in a vacuum.

The differentiator a senior engineer will probe hardest.

Select code, ask a question or request a change and apply it in place.

The explain-and-tweak surface for work that is too small for Agent and too thoughtful for Tab.

Where onboarding and code archaeology actually happen.