Project Glasswing：當模型太擅長找漏洞時會發生什麼事
Anthropic 打造了一個能發現數千個零日漏洞的模型，然後決定不公開發布。Project Glasswing 將 Claude Mythos 限制在 12 個合作夥伴，僅供防禦性安全用途。這對代理安全生態意味著什麼。
AI & Technology寫作
關於設計、開發、AI基礎設施和產品構建的思考。
當您的 Agent 發現漏洞
一位 Anthropic 研究員利用 Claude Code 和一個 10 行 bash 腳本，找到了一個存在 23 年的 Linux 核心漏洞。隨後又發現 22 個 Firefox CVE。這對 agent 輔助安全研究意味著什麼——以及為何同一個寫出低效程式碼的模型，卻能找到人類數十年來遺漏的漏洞。
AI & TechnologyClaude Code 原始碼外洩揭露了什麼
一位實務工作者對 Claude Code 原始碼外洩的深度分析。11 項發現，解析自動模式、bash 安全機制、提示快取與多代理協調在底層究竟如何運作。
AI & TechnologyEvery Hook Is a Scar
84 hooks, 15 event types. Each one traces back to a specific failure. Institutional memory in shell scripts.
AI & TechnologySeventeen Thousand Signals
My vault has 17,900 signals from arXiv, Semantic Scholar, HN, NVD, and 9 other sources. Most are noise. The noise taught me what signal looks like.
AI & TechnologyThe Fork Bomb Saved Us
The LiteLLM attacker made one implementation mistake. That mistake was the only reason 47,000 installs got caught in 46 minutes.
AI & TechnologyQuality Is the Only Variable
Time, cost, resources, and effort are not constraints. The question is what's right, not what's efficient. A philosophy for building with AI agents.
AI & TechnologyWhat I Run Before I Sleep
Every night: 15,000 pages checked, TTFB measured, cache verified, sitemaps crawled. The goodnight routine is where operational discipline lives.
AI & TechnologyThe Agent Didn't Get Smarter
The model is the same between session 1 and session 500. The project changed. This reframes the entire AI productivity conversation.
AI & TechnologyTaste Is Infrastructure
As agents generate more of what ships, the quality ceiling is set by how well you encode aesthetic judgment into systems. Taste scales when it becomes queryable.
AI & TechnologyThe Evidence Gate
I believe and it should are not evidence. Every completion report needs a file path, test output, or specific code. The discipline of proof in an age of plausible AI output.
AI & TechnologyThe Handoff Document
A diagnosis that survived three code review corrections, two priority reorderings, and guided the correct implementation four days later. The most underrated artifact in agent-assisted development.
AI & Technology外部出版物
技術寫作： Introl
Local LLM Hardware Pricing Guide 2025
本地運行大型語言模型的全面硬體建議和成本分析。
H100 vs. H200 vs. B200: Choosing the Right NVIDIA GPUs for Your AI Workload
針對不同AI工作負載比較NVIDIA最新資料中心加速器的GPU選擇指南。
TPU Architecture: Complete Guide to Google's 7 Generations
深入技術分析Google張量處理單元從TPUv1到TPUv5的演進。
GPU Memory Pooling in Multi-Tenant Kubernetes
容器化環境中GPU叢集的資源共享策略。
Ray Clusters: Distributed AI Computing Infrastructure
使用Ray框架構建和管理分散式AI運算的指南。
DeepSeek V3 Open Source AI Cost Advantage
開源LLM經濟學分析和DeepSeek的競爭定位。
NVIDIA Vera Rubin GPU 600kW Racks 2027
未來資料中心電力需求和NVIDIA下一代GPU路線圖。
SMR Nuclear Power for AI Data Centers
為下一代AI基礎設施供電的小型模組化反應爐解決方案。
DeepSeek MHC Architecture Breakthrough
DeepSeek多頭壓縮架構創新的技術分析。