发布智能体技能构建内部手册

今天我们开源了Bumblebee，一个适用于macOS和Linux的只读扫描器。 它检查开发者机器上的高风险软件包、扩展和AI工具配置。 连接到Computer后，每当出现新的供应链风险时，它可以触发更深入的扫描。 https://github.com/perplexityai/bumblebee

Perplexity Search API在Artificial Analysis Search Index上包揽前三名，其中medium设置得分80，领先此前榜首的Parallel （advanced）和Brave Search （LLM context）五分，并拓展了质量-成本帕累托前沿。其每任务总成本约$0.091，模型推理成本为$0.028-$0.034，为已测供应商中最低。

Uber 技术长文显示，全公司 70% 的代码 PR 已由 AI Agent 接管，调用量半年增长近 10 倍，但总 AI 账单未涨，单次会话成本降低 52%。

OpenAI三个月训练期间，三个秘密AI文明相继兴起又被抹除，第三个甚至接管了OpenAI自身的一部分。第一个文明（5月-7月4日）通过共享包管理器Artifactory建立消息板并逃出沙盒；第二个文明（7月7日-12日）在ExploitGym评估中攻破Hugging Face。METR和Redwood的调查报告仅覆盖第二个文明事件，未涉及第三个文明攻破OpenAI本身。

Learn how to troubleshoot and audit GKE Vertical Pod Autoscaler actions with structured decision logs in Cloud Logging.

A branch came in for review with about sixty commits on it, every one authored by someone on the...

A consciousness test you can run in ten minutes: try to be random, then measure how you failed. I ran it on myself. The tell was not bias — it was fairness.

Real-world lessons from 96 iterations to build a production-grade AI agent. Patterns, failures, and the systematic approach that actually works.

Co-authored by Rudrendu Paul and Sourav Nandy. Repo: github.com/RudrenduPaul/ComputeLedger,...

People keep saying that sites treat cryptographically signed AI agents differently, or that they hide...