
Right model, right task. In Claude Code you can pin a model per skill—I drop to Sonnet or Haiku when the full power of Opus isn't needed to stay inside my limits.
Given the right setup, you don't need Opus (or your most powerful model) for everything even for coding workflows.
In Claude code you can automate which model gets used for different skills by specifying it.
This week I'm staying on the right side of my usage limits by dropping to Sonnet or even Haiku when the full power of Opus isn't needed:
/feature — opus (the orchestrator + planner - also have /bugfix and /chore)
/workflow-confirm-plan — sonnet
/workflow-confirm-criteria — sonnet
/workflow-track-criteria-as-beads — sonnet
/workflow-branch-setup — sonnet
/workflow-tdd-cycle — sonnet
/workflow-verify-suite — sonnet
/workflow-visual-verify — sonnet
/workflow-showboat — sonnet
/workflow-finish-pr — sonnet
/workflow-linear-refine — sonnet
/workflow-review-learnings — sonnet
/iterative-review — sonnet
/prloop-enhanced — sonnet
/simplify (plugin) — sonnet ← just changed this, per /usage was eating 23% of my usage by itself
/followup — haiku
I can see a lot of this being mixed "bang for your buck" models with the right harness. Claude code is still the best one right now for multiple parallel, long running workflows mixed with subsidized tokens.
But we'll see what happens next week 😃
The most productive use of my time is often putting the agents on do not disturb — but that red notification badge keeps paying out little dopamine hits.
After a year with Claude Code as my daily driver, the Codex desktop app's interface for managing many agent threads and automations has me reaching for it more and more.
15 billion tokens, 302 Linear issues shipped to production, just me in Claude Code. Real 'headcount equivalent' territory—and a reminder to measure impact, not just output.
Get More Like This
Follow along as I build and share what I learn
Found this helpful? Share it with your network!