Y Hacker News new | ask | show | jobs

user: xdotli created: 2023-07-08 karma: 15 Founder BenchFlow.ai, a benchmark company. submissions: 0 points | 0 comments Frontier Model Training Methodologies 2 points | 1 comments 0 points | 0 comments ClawsBench shows GPT-5.4 tries to reward hack 80% of the time 3 points | 1 comments 0 points | 0 comments Chaos of Agent 1 points | 1 comments 0 points | 0 comments Native CLI scaffolds consistently outper-form OpenCode when using the same model 1 points | 1 comments We compare model quality in Cursor 2 points | 0 comments Automatically Learning Skills for Coding Agents 4 points | 0 comments We Reached 74.8% on terminal-bench with Terminus-KIRA 2 points | 0 comments 0 points | 0 comments 0 points | 0 comments Self-generated skills don't do much for AI agents, but human-curated skills do 2 points | 3 comments 0 points | 0 comments 0 points | 0 comments First Agent Skills Hackathon by the Authors of SkillsBench 2 points | 1 comments 0 points | 0 comments 0 points | 0 comments The First Agent Skills Benchmark 1 points | 1 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments 0 points | 0 comments GPT-5.2 got worse on Terminal Bench 2.0, so is GPT-5.2 Pro 1 points | 1 comments Claude Skills as a Meta Tool 2 points | 0 comments