Melissa Pan says coding-agent harness choice affects cost more than accuracy
Melissa Pan, a PhD candidate at UC Berkeley’s Sky Computing Lab and a former Arena intern, compared the coding agents Claude Code, Codex, and Pi while examining a “harness tax”: how the system surrounding an AI model affects its cost and performance.
She reports three findings. The one stated here is that harness choice affects cost more than accuracy. The full analysis is in the video.