Aside's Computer Use agent hits SOTA on Scale AI's Cua Driver Bench
garrytan · x · 2026-11-13
Aside reports its Computer Use agent achieved SOTA on Scale AI's Cua Driver Bench. The team says they never aimed at leaderboards—instead they spent every day polishing background computer use (and browser use) for better UX, which they credit for the result.
Related event: Aside's computer use agent tops Scale AI benchmark(2 posts)→
More from coding & agent
- Pi-CLM keeps one coding session alive for 2 months with 120+ compactions — RulinShao · 2026-11-13
- Gemini 4 Hydrogen is the fastest model in Agent Arena, 2.7x quicker than Sonnet 5.5 — arena · 2026-11-13
- Google ships A2UI 1.0: an open protocol for generative, native agent UIs — rseroter · 2026-11-13
- Raising, not prompting: simulating hundreds of days to seed an AI agent's identity — RileyRalmuto · 2026-11-13
- User adds =SHELL() and =PYTHON() to open-source Gridcraft in 10 minutes — arthurcolle · 2026-11-13
- MCP Master Tester 1.3 adds one-click smoke tests and CI checks for MCP servers — simplymanas · 2026-11-13