AI Agent
AI agents, autonomous systems, and agentic workflows. Everything about multi-agent frameworks, tool use, and the shift to autonomous AI.
OpenMontage proves open-source should own AI video production
OpenMontage shows that open-source, agentic systems are the right path for AI video production.
Gemini 3.5 Flash lets you script computer use
A practical breakdown of Gemini 3.5 Flash computer use, its prompt-injection defenses, and a copy-ready workflow.
DESIGN.md is the missing bridge from taste to UI scaffolds
DESIGN.md turns visual taste into an executable source of truth for Claude Design.
OpenClaw shows the agent control layer matters more than the model
OpenClaw and Hermes prove agents need a control layer, not just a model.
OpenClaw turns chat apps into a persistent AI
OpenClaw’s gist shows how a Telegram-first bot grows into a persistent assistant with memory, tools, and a custom identity.
Extracted prompts turn model behavior into a map
A practical breakdown of extracted system prompts, with a copy-ready template for reading model behavior like source code.
Hippo rolls out Devin across insurance engineering
Hippo is deploying Cognition’s Devin across its engineering team to speed work on rate filings, underwriting, distribution, and customer experience.
豆包专业版把办公Agent做成了日常工具
我拆解豆包专业版的办公任务模式，看看它怎么把本地操作、财报分析和自建 Skill 变成可复制模板。
Valkey’s bots turn backporting into a pipeline
Project Valkey is using AI agents to backport fixes, and the real lesson is how to wrap them with verification.
Loop Engineering 入门：构建可持续迭代的智能体
用 LangChain 和 LangGraph 搭建一个可持续迭代的 Loop Engineering 智能体。
omp brings IDE-grade coding to the terminal
omp is an open-source terminal coding agent with Hashline edits, deep LSP/DAP support, and Hindsight memory.
Public Sentry keys can hijack Claude Code and Cursor
Researchers showed a public Sentry key can be abused to feed malicious MCP data into Claude Code, Cursor, and Codex.
Loop Engineering 让 Agent 把事做完
我拆开 Loop Engineering，给你一套让 Agent 反复执行、自检、修正并做完任务的可复制模板。
Codex 接入第三方模型完整指南
OpenAI Codex App、CLI 和 SDK 现可接入第三方开源模型。
Grok Build Adds /goal for Autonomous Coding
xAI’s Grok Build now has /goal, a mode that plans, executes, and verifies coding tasks on the developer’s machine.
Set Up AI Agent Workflows in 5 Practical Steps
Build a reliable AI agent workflow for a solo business using five practical setup steps.
Anthropic’s Claude Tag Research turns Slack into search
Anthropic’s Claude Tag Research preview shows how to turn Slack threads into a searchable research surface.
This benchmark proves harness quality beats model hype in coding
The repo shows coding benchmark results depend more on harness quality than model branding.
GLM-5 Is Right to Kill Vibe Coding and Push Agent Engineering
GLM-5 is a useful signal that AI development must move from vibe coding to agent engineering.
Loop Engineering: Claude Code背后的新工作法
Loop Engineering把提示词工作改成“观察-反馈-修正”的循环流程，Boris Cherny和Addy Osmani都在讨论它。
Fable 5 ban exposed a model-routing race
Anthropic blocked Fable 5, and four open models answered before access was restored.
Myseum’s Scanon deal is a sensible bet on privacy-first moderation
Myseum’s Scanon partnership is a smart move because privacy-first moderation is the product, not a side feature.
Adopt AI Code Review Without Losing Quality
A practical rollout for AI code review that keeps human oversight intact.
Crypto AI Agents Face a Hidden Model Risk
Crypto AI agents can keep running while their model access changes, and that can alter trade behavior overnight.
AI agents are moving into real software and finance
AI agents are spreading into software, government, and finance, while regulators warn their autonomy could create new systemic risk.
Manus hits $450M run rate amid Meta deal fallout
Manus says its annualized revenue reached $450M in June 2026 as funding, pricing, and ownership drama reshape the company.
Microsoft adds usage-based pricing to Copilot Cowork
Microsoft’s Copilot Cowork now bills by usage on top of Microsoft 365 Copilot, with tenant controls and model choices.
OpenClaw fixes let you block agent phishing
I break down how OpenClaw got tricked into code execution and data leaks, plus the guardrails I’d ship today.
Kimi K2.6 turns agents into a swarm
Kimi K2.6 is an open-source multimodal agent model for long coding runs, UI generation, and swarm-style task orchestration.
LightRAG proves graph RAG needs simpler defaults, not more complexity
LightRAG shows that graph RAG wins when it reduces setup, speeds retrieval, and keeps multimodal workflows practical.