2 min lesson
The mechanics behind the habits (and the numbers that sell them)
Use "Even an ambiguous prompt like 'change the button to blue' returns an applicable result" to say what you would do next.
Step 1 of 2
The mechanics behind the habits (and the numbers that sell them)what to know cold when an engineer challenges you
Senior engineers will press on how indexing works, especially on the security angle. Here's the precise mechanic: on first open Cursor chunks the repo into functions/pieces and generates vector embeddings; the embeddings are hashed and stored, but the raw code is held only ephemerally and then deleted - the only long-term store is the vector DB. At prompt time, semantic search pulls the most relevant chunks into context. The .cursorignore file lists files Cursor should never index or view, which is the control surface for sensitive code.
Other tools lean on grep (substring search), which is slow and degrades on a large production codebase. Because Cursor indexes the whole repo into a vector DB, it understands what you semantically mean and stays fast - even an ambiguous prompt like "change the button to blue" returns an applicable result.
This is the differentiator that lets cloud agents work in very large repos: the field team demos it on Grafana - roughly 2M+ lines and 30,000 files - where substring search stalls.
On @-mentions, sharpen the contrast you teach: tagging the exact file with @ guarantees that file lands in context, rather than hoping semantic search retrieves it. Asking "what color is the badge?" forces the agent to search the whole app; @badge.tsx plus "what's the hex for the badge background?" makes it read one file and return the right answer. @ isn't limited to files either - you can reference a past chat, which summarizes it and reads the transcript, so a new agent inherits the summary without carrying all prior context forward.
Learn more
Full explanation
The context-window numbers that actually matter
Output quality drops steeply as the window fills - presenters converge on staying under ~60-65%, and several move on at ~50%. A bottom-right indicator shows fullness and its makeup; if you're already 50% full before sending a prompt, you likely have too many MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition./tool integrations loaded.
Past the limit Cursor runs compaction (compressing prior context), which can behave like not having that context at all - while you still pay for all the tokens. The fix is to start a fresh agent per new topic, not to keep one mega-chat alive.
Learn more
Full explanation
"It's not a thin proxy" - the harness, in one breath
When a buyer assumes Cursor just forwards prompts to model providers, correct it: there's deep proprietary cloud infrastructure - the Cursor harnessCursor's hosted layer around each model (context selection, caching, retries) that makes the same model run better and cheaper than calling it directly. Press Enter for the full definition. - doing optimization, context enrichment, caching and compaction around the model, tuned per frontier model.
The proof point to quote: combined with dynamic context discoveryThe agent pulling only the relevant parts of files, tools and MCP servers into context as needed, instead of loading everything up front. Press Enter for the full definition. (the agent pulls in only what it deems relevant rather than loading everything), the harness reduces agent tokens by ~47% even when running another vendor's model - and independent analysis shows the same model performs better in Cursor's harness than elsewhere.
Your product's own codebase is the docs that never lagthe one habit field engineers report using all day
The highest-value move an SE makes with their own seat: connect Cursor to the foundational repositories of the product you sell (within your version-control access) and ask the actual codebase natural-language questions. How does this service work? How does indexing turn code into vector embeddings, and what are the embedding sizes? Was this path deprecated? Is it behind a feature flag? Where does it live? You're not reading docs that lag the code. You're reading the code.
That's what lets you answer the specific RFP and security questions authoritatively instead of promising a follow-up: What's the data-retention behavior? Does customer data ever leave the cloud environment? Is this control configurable? Keep the repo open all day, even during calls, so the answer is one prompt away.
"This is like the most supercharged docs you could possibly think of. I use this all day every day."