"Cursor is Dead" is Total BS: Here is Why | Miles Clements

Clements argues some categories (legal/accounting) take longer to deploy but become deeply embedded once adopted, while “vibe coding” apps can deliver instant value but lack defensibility. Coding tools (Cursor/Claude Code) are compelling because users see immediate productivity gains and compounding team-level lock-in.

He says Claude Code’s momentum is real but largely tethered to model breakthroughs (e.g., Opus versions) and the overall market is so expansionary that one tool’s growth doesn’t automatically imply another’s decline—especially when consumption-based revenue is exploding across products.

Clements cites Cursor metrics: agent users outnumber tab-feature users 2:1; ~90% of users are daily active on the agent product; agent product grew ~15x; and cloud agents (new as of late Oct) drive ~35% of merged PRs. His point: the company is already executing the agent transition.

Accel’s in-progress survey suggests ~50% of developers switch model families daily and ~95% switch models daily. Cursor’s multi-model UX turns it into an “index of AI innovation,” where improvements in underlying models and in Cursor’s product compound together.

He frames Cursor’s modeling effort as specialized coding models for professional/enterprise needs rather than broad generalist models. Specialization can improve reliability, security, and workflow fit for real-world coding tasks where generalist breadth (poetry, recipes) is irrelevant.