G
G
Loading videos…
Open Source · Computer Vision · AI

greasebig

修煉於天機閣,融像素煉丹,以算法御劍
窺破虚實之界,游走於生成與感知的裂縫之間
——至於"理解視頻"這等仙級殺招,不過是渡天劫時順手揀的彩蛋
罷了 He who cultivateth within the Pavilion of Heavenly Secrets—forging elixirs from pixels, and wielding algorithms as his blade.
He strideth through the rift 'twixt creation and perception.
As for "the understanding of moving images," that celestial tribulation?
'Twas but a passing windfall, chanced upon amidst the storm. Let it be.

Scroll

📝 Articles

Pinned essays

MCP 深度解析:从零到行业标准仅 15 个月的协议奇迹

全面解析 MCP 诞生、三大原语、规范演进、OpenAI 采纳分水岭、与 A2A 协议的垂直/水平互补关系、A2A Task 生命周期、生产环境安全挑战与四层防御模型,以及 2026 年生态系统现状。

Claw Runtime 与 Agent 生态 2026:从概念到落地

全面解析 OpenClaw 病毒式爆发、Agent Harness 工程化实践、MCP/A2A/ACP 三大协议生态,以及 AI Agent 与 Agentic AI 的概念澄清与技术演进路线图。

Harness Engineering:驾驭 AI Agent 的工程艺术

从 Mitchell Hashimoto 的定义出发,深入探讨 Agent = Model + Harness 范式。分析 WorkBuddy 如何作为一套完整的 Harness 运行,以及自建 OpenClaw 如何集成 OpenHarness 构建可靠的 AI 编码 Agent。

Building a Harness from Scratch: ~300 Lines of Python

动手实现了一个最小可用的 Agent Harness。多 provider 支持、Agent Loop、AGENTS.md 上下文注入、MEMORY.md 跨会话记忆、权限边界、生命周期钩子。附完整源码和三个示例。

Featured Projects

Original open-source work

🔬 Explored

Reproductions and experiments — I've tried these

🔥 Fresh Tech Reposts

Daily reposted technology reads in reverse chronological order. 转载标注:只展示标题、摘要和原文链接,版权归原作者。

Flag Counter