Used Claude Code with Opus 4.7 superpowers to do detailed planning and implementation for a feature. After running my own code review, I threw $2 into DeepSeek V4 Pro just to test it out and had it review the PR.
52 files, ~6345 insertions. It took 1.5–2 min total, burned 1.6M tokens and $0.06, and came back with 5 findings.
Honestly not that surprised it still caught things despite the planning code review on the Claude Code side. Fed the findings back to Claude Code and it agreed all 5 were legit 😄