Stop burning tokens on code review

In this article, Swizec Teller discusses the challenge of keeping code review efficient in an era of high-volume code production. He shares his team's experience with AI code review tools like Cursor's BugBot, which was costly ($1000/week) and noisy, missing architectural concerns. They then tried running custom Claude code review skills in GitHub Actions, but that was even more expensive and overwhelming. The solution: custom linters that provide fast, cheap guardrails and feedback loops, applying to both human and agent-written code. This approach balances speed and safety without burning tokens on verbose AI reviews.