What is Cursor doing strategically?

Cursor acquired Graphite in December 2025, bringing stacked pull requests, AI code review, and PR merge workflows into the Cursor ecosystem. Graphite was used by Shopify, Snowflake, Figma, and Perplexity. Integration is planned throughout 2026, with tighter connections between local development and pull requests already scoped.

The GitHub pull request workflow was the last major step in the engineering lifecycle outside Cursor's control. Owning it lets Cursor sell engineering leaders a single contract for write, review, and ship, removing the need to justify Cursor as an add-on to an existing GitHub-centered procurement. For any founder competing here, this repositions Cursor from a developer productivity tool to an engineering operations platform.

This move is executed, not experimental. Graphite's customer base gives Cursor immediate enterprise proof across marquee names. The integration timeline through 2026 gives Cursor a compounding retention argument at renewal time: the more workflows land inside Cursor, the higher the switching cost.

Cursor introduced Composer 2, a proprietary coding model trained on live production inference tokens using real-time reinforcement learning. Checkpoints deploy as often as every five hours. Bugbot's PR resolution rate has climbed from 52% to 78%, the highest of any AI code review product measured in public benchmarks.

Most competitors in this space, including YC-cluster agents, are entirely dependent on third-party frontier models. Cursor is building a compounding model advantage: every engineering team that ships through Cursor contributes training signal that improves Composer. Over time this becomes a data moat that third-party model access cannot replicate.

This is the signal with the longest compounding tail. The rate of improvement will slow, but Cursor now has a structural advantage that did not exist 12 months ago and is invisible to most procurement buyers evaluating on feature checklists today.

The June 2025 switch to credit-based billing cut effective monthly requests from roughly 500 to 225 for Pro users at the same $20 price point. Trustpilot sits at 1.7 from over 200 reviews, dominated by billing complaints. The CEO publicly acknowledged mishandling the rollout. Billing-related friction remains the top complaint on Trustpilot and Reddit well into 2026.

Developer trust is the primary acquisition channel in this category: peer recommendation, Reddit threads, and X posts drive trial before any sales motion. Cursor's Trustpilot score is an open wound. Any challenger that publishes transparent, predictable pricing, fixed monthly requests or flat rate, has a concrete wedge into the individual developer market where Cursor's community trust is lowest.

This is real, not noise. The Trustpilot gap versus G2 is not explained by review bias alone. It maps directly to the billing change timeline. The risk for challengers is that Cursor's enterprise lock-in is growing faster than the community trust damage can slow it. Act on this window now before enterprise renewals make it irrelevant for your target segment.