YC President Garry Tan Open-Sources Benchmarks of 7 Coding Agent Frameworks

Y Combinator President Garry Tan has open-sourced g-harness-evals on GitHub, benchmarking frontier models across 7 harness configurations including Claude Code. Early results show the frameworks are roughly equal in coding capability, yet their costs differ by up to twofold.

2026-11-09 ~ 2026-11-09 · 2 related posts

• Garry Tan's open-source eval: 7 coding harnesses tie on capability but differ 2x on cost — garrytan · 2026-11-09
• YC President Garry Tan open-sources g-harness-evals to publicly benchmark frontier models across harnesses — garrytan · 2026-11-09