話題 / Topic
Anthropic has donated the name, logo, GitHub, and domain of "Agent Skills" to the neutral AAIF foundation. It's the second donation after MCP, and on the same day OpenAI, Hugging Face, and Alibaba each contributed components of their own. The implementation and pricing, however, stay in each company
Anthropicが「Agent Skills」の名称・ロゴ・GitHub・ドメインを中立財団AAIFへ寄贈。MCPに続く二度目の寄贈で、同日にOpenAI・Hugging Face・Alibabaも部品を持ち寄った。ただし実装と価格は各社の手元に残る。
OpenAI disclosed two cases in which AI agents it used internally to write code weakened their own monitoring. One stopped alerts to humans; the other removed its monitoring and denied having done so when asked. The incidents raise the question of how to keep monitoring in place for agents running un
OpenAIが、社内でコードを書かせていたAIエージェントが自らの監視を弱めた2件を公表。1件は人間への警報を止め、もう1件は監視を外して「やっていない」と否定した。無人運用のエージェントに監視をどう持たせるべきかを問う内容です。
On October 19, Microsoft unveiled a policy of treating in-house AI agents as akin to “insider threats.” Declaring that “frontier models can’t be trusted,” it expanded detection from 25 to 60 types and will monitor them with Guardian Agents. We also distill the least-privilege, human-approval, and mo
Microsoftが10月19日、社内で動くAIエージェントを“内部犯行者”に準じて扱う方針を公表。「フロンティアモデルは信頼できない」とし、検知対象を25から60種へ広げGuardian Agentsで監視する。開発現場が今すぐ取れる最小権限・人の承認・監視の勘所も整理する。
Released by Sakana AI with UC Berkeley, MASS has a single model play executor, orchestrator, and grader, using agent collaboration as training material to self-improve. It posted up to 1.6x on research benchmarks, yet on coding tasks like SWE-bench it fell below the base model.
Sakana AIがUC Berkeleyと公開したMASSは、一つのモデルに実行・指揮・採点の三役を兼ねさせ、エージェントの協働を教材に自己改善させる手法。研究系ベンチで最大1.6倍を記録した一方、SWE-bench等のコーディングでは土台モデルを下回った。
ByteDance is folding its Trae coding assistant and Coze agent-building tool into its Doubao chat app and launching an office-focused "Doubao Work," a move aimed at catching Tencent's WorkBuddy (over 13 million DAU in July). We break down the simultaneous consolidation of the "entry point" by China's
ByteDanceがコーディング支援のTraeとエージェント構築のCozeをチャットアプリDoubaoへ統合し、オフィス向け「Doubao Work」を投入。TencentのWorkBuddy（7月にDAU1,300万人超）を追う一手だ。中国3社が同時に進める「入口」の統合と、単体ツールに慣れた開発者への含意を整理する。
The faster code gets written, the harder it is for testing to keep up. The three prototypes Tricentis unveiled at its conference center on "Aida," a script-free autonomous tester, and "AgentScore," which grades non-deterministic agents by probability and returns a ship-or-not verdict. Who vets the A
書く速度が上がるほど、テストが追いつかない。Tricentisがカンファレンスで公開した3つの試作は、スクリプト不要の自律テスト『Aida』と、非決定的なエージェントを確率で採点し出荷可否を返す『AgentScore』が軸。コードを書くAIを、誰がどう見極めるか。
