How Boris Uses Claude Code - Boris Cherny

Boris Cherny, creator of Claude Code, shares his personal setup and best practices. He emphasizes that his "surprisingly vanilla" approach demonstrates Claude Code's flexibility rather than prescribing a single correct methodology.

Parallel Session Management

Cherny runs 5 Claude instances simultaneously in his terminal (numbered tabs 1-5) with system notifications, plus 5-10 additional instances on claude.ai/code and the iOS app. He frequently teleports sessions between platforms and hands off work between local and web environments.

Model Selection

He exclusively uses "Opus 4.5 with thinking for everything" despite its larger size, reasoning that superior steering and tool-use capabilities make it faster overall than smaller models requiring more correction.

Institutional Knowledge

Teams maintain CLAUDE.md files checked into git repositories, where the team collectively documents Claude's errors and adds preventive instructions. This allows Claude to learn team-specific patterns across sessions.

Planning and Execution

Sessions typically begin in Plan mode (shift+tab twice), enabling back-and-forth refinement of Claude's approach before switching to auto-accept edits mode for implementation.

Slash Commands

Custom commands stored in .claude/commands/ automate frequent workflows. Example: /commit-push-pr uses inline bash to pre-compute git status, reducing model back-and-forth.

Subagents

Specialized agents like code-simplifier and verify-app automate post-generation tasks, approximating the "most common workflows" across pull requests.

Permission Management

Rather than using --dangerously-skip-permissions, Cherny pre-allows safe bash commands via /permissions, with common configurations shared team-wide in .claude/settings.json.

Tool Integration via MCP

Claude Code integrates with external systems through MCP servers, enabling: - Slack posting and searching - BigQuery queries for analytics - Sentry error log retrieval - Chrome extension testing for UI verification

Verification

Cherny identifies verification as "probably the most important thing to get great results" from Claude Code, claiming it can "2-3x the quality of the final result."

His verification approach includes: - Chrome extension-based browser testing - Automated UI interaction and iteration - Test suite execution - Domain-specific validation methods

For long-running operations, he uses verification agents, Stop hooks, or the ralph-wiggum plugin to ensure quality without requiring constant monitoring.

Code Quality

A PostToolUse hook handles code formatting after generation, addressing the remaining 10% of formatting concerns to prevent CI failures.

Key Quote

"We intentionally build it in a way that you can use it, customize it, and hack it however you like. Each person on the Claude Code team uses it very differently."