Vetta Harness Cuts Long-Horizon Agent Costs by 66% vs. Claude Code
ycombinator · x · 2026-08-25
Vetta introduces a new harness optimized for long-horizon agent tasks. Using the same model and tasks, Vetta achieves a $0.298 cost per finished task—roughly a third of claude-code's $0.872—while securing the highest finish rate. The post argues that cost efficiency is the critical enabler for autonomous companies.
More from coding & agent
- Hands-on with FactoryAI Droid: Building Agent Workflows — matanSF · 2026-08-25
- Paper: Long-Horizon Agents Need Strong Pre-training and OPD — rohanpaul_ai · 2026-08-25
- Headlong: Open Source Microharness for Persistent, Self-Guided Agents — CShorten30 · 2026-08-25
- Samsung uses Claude Code for chip verification: one-month tasks done in two days, 15x speedup — 新智元 · 2026-08-25
- Swarms MCP Launches: Single Endpoint for Multi-Agent Workflows — KyeGomezB · 2026-08-25
- How to set up Agent evaluation: scripted checks or expert review? — mastra_ai · 2026-08-25