CoArena Launches: Crowdsourced Arena Benchmark for Computer-Use Agents Backed by YC
ycombinator · x · 2026-08-23
Coasty launched CoArena, billed as the biggest crowdsourced benchmark for Computer-Use agents, backed by Y Combinator.
Static suites leak into training data, so scores climb without real capability gains. An arena can't saturate: every battle is a real task from a real person, with two frontier models running side by side in identical sandboxes — every click, screenshot and step recorded. Humans judge blind, down to verdicts on individual steps.
Early findings:
- Only 66% of agent runs finish their task at all; 1 in 3 dies before any score exists
- On identical tasks, the best frontier agent completes 87% of runs; the worst only 42%
The leaderboard is live; users can watch battles, judge blind, and post their own tasks.
Related event: CoArena Launches as Largest Crowdsourced Computer-Use Benchmark(2 posts)→
More from coding & agent
- MCP vs CLI vs API: How Tool Design Drives Your AI Token Bill — sanjaykalra · 2026-08-23
- Optimized SENPAI Prompts, Qwen 3.8 27B Agent Uses Sub-agents More Frequently — morgymcg · 2026-08-23
- Building Great Evals: Avoid Single Scores and Embrace Hill Climbing — realmadhuguru · 2026-08-23
- Stop Extracting Everything: Good Codebases Minimize Jumps — serrjoa · 2026-08-23
- Google's Antigravity IDE Gains Traction, Developers Call Gemini 3.7 Flash a Game Changer — jocarrasqueira · 2026-08-23
- Study finds AI agents lock in training strategies early, hindering recursive self-improvement — omarsar0 · 2026-08-23