Claude/Code/Blog/26/02/Boris Tips for Claude Code pt2

• Boris Tips for Claude Code pt2 - Boris Cherny on Threads

  • I'm Boris and I created Claude Code. I wanted to quickly share a few tips for using Claude Code, sourced directly from the Claude Code team. The way the team uses Claude is different than how I use it. Remember: there is no one right way to use Claude Code -- everyones' setup is different. You should experiment to see what works for you!
    

  • Tip 1: Do more in parallel

    • Run 3–5 git worktrees at once, each with its own Claude Code session. The team calls this the single biggest productivity unlock and the top tip.
    • Boris personally uses multiple git checkouts; most of the Claude Code team prefers worktrees, which is why native worktree support was added to the Claude Desktop app.
    • Basic workflow:
      • Create a worktree: git worktree add .claude/worktrees/my-worktree origin/main
      • Enter it and start Claude: cd .claude/worktrees/my-worktree && claude
    • Team variations:
      • Some name worktrees (e.g. za, zb, zc) and use shell aliases to jump between them quickly.
      • Others keep dedicated worktrees for specific tasks (e.g. an "analysis" worktree for reading logs and running BigQuery).
    • See Common workflows in Claude Code docs for more.

  • Tip 2: Start every complex task in plan mode

    • Core idea: Put your energy into the plan so Claude can often 1-shot the implementation. Start every complex task in plan mode (e.g. shift+Tab to cycle into it).
    • Two team patterns:
      • Plan + review: One person has one Claude write the plan, then spins up a second Claude to review it "as a staff engineer."
      • Re-plan when things go wrong: The moment something goes sideways, switch back to plan mode and re-plan instead of pushing on. They also tell Claude to use plan mode for verification steps, not only for the initial build.
    • Don’t keep pushing when the approach is off; re-plan and then execute again.

  • Tip 3: Invest in your CLAUDE.md

    • Correction loop: After every correction you make, update your CLAUDE.md so that mistake doesn’t happen again.
    • Claude is very good at writing rules for itself; feeding it examples of corrections helps it internalize better behavior.
    • Iterate: Edit CLAUDE.md ruthlessly over time and keep iterating until Claude’s mistake rate measurably drops.
    • Example workflow: One engineer keeps a notes directory per task/project, updates it after every PR, then points CLAUDE.md at that directory so context stays in sync.
    • Memory files: You can have both a global ~/.claude/CLAUDE.md (e.g. 76 tokens) and a project CLAUDE.md (e.g. 4k tokens); use both for rules and context.

  • Tip 4: Create your own skills and commit them to git

    • Reuse across every project by turning repeated workflows into skills (and commands) and committing them to git.
    • Team heuristics:
      • Automation rule: If you do something more than once a day, turn it into a skill or slash command.
      • Tech debt: Build a /techdebt slash command and run it at the end of every session to find and remove duplicated code.
      • Context sync: Use a slash command that syncs 7 days of Slack, GDrive, Asana, and GitHub into one context dump so Claude has a single, up-to-date view.
      • Specialized agents: Build analytics-engineer-style agents that write dbt models, review code, and test changes in dev.
    • See Claude Code Docs – Extend with skills for details.

  • Tip 5: Claude fixes most bugs by itself — here’s how the team does it

    • Slack MCP + “fix”: Enable the Slack MCP, paste a Slack bug thread into Claude, and say "fix." No context switching; Claude has the thread and codebase in one place.
    • Failing CI: Tell Claude to "Go fix the failing CI tests" and don’t micromanage how; let it run, read failures, and iterate.
    • Distributed systems: Point Claude at docker logs (or similar); Claude is surprisingly capable at reasoning from logs to find and fix issues.