AI News Roundup: Meta Brain2Qwerty v2 Decoding, Cursor for iOS, GLM 5.2 Access Pass, and Arena Hits $100M ARR
English summary
Meta released Brain2Qwerty v2, a real-time sentence decoder from non-invasive brain signals achieving ~61% word accuracy overall and 78% for the best participant, with training code and dataset publicly available. Cursor launched an iOS app supporting always-on cloud agents and remote control of computer-bound agents, including diff review and notifications. Open-weight model access was productized: Cline introduced a $9.99/month pass bundling GLM 5.2, DeepSeek, Kimi, MiniMax, and Qwen. The DeepSeek DSpark speculative decoding method reached 30.9% higher accepted length vs Eagle3 and is deployed in DeepSeek-V4-Flash and -Pro. AI evaluation platform Arena reported a $100M ARR run rate eight months after launch, now emphasizing agent-mode evaluations.
Chinese summary
Meta发布了Brain2Qwerty v2，一种通过非侵入式脑信号实时解码句子的系统，总体词准确率约61%，最佳参与者达78%，并公开了训练代码和数据集。Cursor推出iOS应用，支持始终在线的云端代理及远程控制电脑端代理，提供差异审查与通知功能。Cline推出9.99美元/月的模型访问套餐，打包GLM 5.2、DeepSeek、Kimi、MiniMax和Qwen。DeepSeek的DSpark投机解码方法相比Eagle3实现30.9%更高接受长度，已在DeepSeek-V4-Flash和Pro中部署。AI评估平台Arena宣布年化经常性收入达1亿美元，发布仅8个月，现侧重代理模式评估。
Key points
Meta Brain2Qwerty v2 achieves real-time sentence decoding from non-invasive brain recordings with ~61% word accuracy overall (78% best), releasing code and dataset.
Meta Brain2Qwerty v2实现非侵入式脑信号的实时句子解码，总体词准确率约61%（最佳78%），并公开代码与数据集。
Cursor for iOS launched with always-on cloud agents, remote agent control, diff review, and notifications.
Cursor推出iOS版，支持始终在线的云端代理、远程代理控制、差异审查和通知。
Cline released a $9.99/month pass bundling GLM 5.2, DeepSeek, Kimi, MiniMax, and Qwen models, reducing access friction.
Cline推出9.99美元/月的套餐，整合GLM 5.2、DeepSeek、Kimi、MiniMax和Qwen模型，降低使用门槛。
DeepSeek DSpark speculative decoding outperformed Eagle3 by 30.9% accepted length and is deployed in V4-Flash/Pro inference engines.
DeepSeek的DSpark投机解码接受长度超过Eagle3达30.9%，已部署于V4-Flash/Pro推理引擎。
Arena reached $100M ARR run rate 8 months after launch and is shifting focus to agent-mode evaluations (task completion, hallucination rates).
Arena在发布8个月后年化经常性收入达1亿美元，重点转向代理模式评估（任务完成、幻觉率等）。