Top 0.01% Cursor user switched to Claude Code for behavior-level builds
At a glance
Silen Naihin, emailed as a top 0.01% Cursor user, moved most work to Claude Code 2.0 with Opus and built a genetic-algorithm simulator in a day without writing a line himself. This workflow is listed under coding-agents, claude-code, cursor, workflow, devtools.
- Agents / tools
- Claude Code, Cursor, ChatGPT
Building a tool for these workflows? See weekly rates.
What they did
After living in Cursor Composer from late 2024 and getting a top 0.01% usage email from the Cursor team, he tried Claude Code again with the 2.0 harness and Opus 4.5. He now runs most planning, generation, and refactors in Claude Code (often many parallel terminals), keeps Cursor for pixel-perfect UI and small isolated edits, and uses ChatGPT for second opinions. He reports building a genetic-algorithm simulator with interactive visualizations in one day without writing a single line of code.
How it went
He says the terminal-native workflow forces async, behavior-level testing instead of reviewing every edit, and that Opus is noticeably better inside Claude Code's scaffold.
Worth knowing
He still uses Cursor for learning loops and UI polish; he frames Claude Code as for abstraction-max builders, not as a replacement for understanding code when you need it.