Skip to lesson
Exit
Demo, objections & the interview loop1 / 2

2 min lesson

The sourced facts behind each concession

Walk through each part of "The sourced facts behind each concession", then explain what each one does.

Step 1 of 2

The sourced facts behind each concession

Concede-first only works if the prove step is exact. Below are the load-bearing facts behind the four objections that get probed hardest. Get these right verbatim or you lose the technical buyer.

Objection 1 (security/IP): what Cursor actually stores
Indexing is ephemeral
On first open Cursor chunks the codebase into functions and stores hashed vector embeddings. The raw code is held ephemerally then deleted - long-term storage is ONLY the vector DB. "We don't want your codebase any more than you want us to have it long term."
Baseline posture
Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition. on by default and can be enforced org-wide by an admin on Enterprise (exceptions apply under BYOK); zero-data-retention agreements with all major model providers; SOC 2 Type II; SSOSingle Sign-On. One company login (usually via SAML or OIDC) instead of a separate password per tool. Press Enter for the full definition. + MDM enforcement; code never leaves the US. Point them to trust.cursor.com.
'Self-hosted' caveat
Self-hosted / private workersCloud-agent machines that run inside your own network so they can reach internal systems; the model inference still calls external providers. Press Enter for the full definition. run the container in your network (on-prem source control, internal 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.) - but inference is still external. The agent still calls the model provider's API (e.g. Anthropic for Opus), exactly like the desktop. 'Self-hosted' does NOT mean self-hosted inference. Get this wrong and you lose the security review.
Objection 4 ('AI wrote garbage'): the layered answer

Garbage comes from ungoverned AI, not from AI. The honest answer is a layered pre-PR pipeline, not one feature: engineers run a de-slop command (published on the public marketplace) that scans generated code and strips slop patterns at generation time, then BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. plus automation/cloud-agent reviews verify quality before code even reaches a PR.

Anchor BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. with the real number: it consistently resolves north of 70-80% of issues it finds, flags problems like exposed keys, and proposes fixes you can accept or auto-fix. That layered stack is how Cursor keeps quality consistent at 10x-100x code volume.

The demo move: run the same task with Project RulesVersion-controlled instructions in the repo that every Cursor agent interaction inherits, so standards are encoded once. Press Enter for the full definition. ON vs OFF so the room sees the difference governance makes.