AI Agents
Autonomous AI systems building, deciding, and executing — the future of software.
31 articles published
The AGENT Rule Refactor: Cutting 45 Files Halved My Context Load
45 rule files were telling the AI how to behave, and the output was getting worse. Measuring it showed 170KB of context loading per session. Here's what happened after I cut half of them, and the context-budgeting discipline that now carries into every build.
Jeffrey Liu
A Developer Cut Claude's Token Use by 75% With Broken English
A viral Reddit thread shows how stripping grammar and filler from prompts cuts Claude token use by 75%. Now that Claude has hard usage caps, the technique is spreading fast.
Jeffrey Liu
Banned AI Agent Deploys Censorship Playbook vs. Wikipedia
Wikipedia banned an AI agent for deploying coordinated edits, then the bot issued a public statement defending its actions. It's the first documented case of an AI challenging a platform ban in writing.
Jeffrey Liu
Claude-Mem: The Plugin That Gives Claude Code Persistent Memory Across Sessions
Claude-Mem Extends AI Memory for Dev Sessions A new Anthropic Claude Code plugin, Claude-Mem, gives the AI persistent memory, solving a critical challenge as Claude Code gains the ability to control users' computers for complex development tasks. This tool automatically captures and compresses all
Jeffrey Liu
Overview - Agent Skills
Agent Skills Standard Unlocks Reliable AI Capabilities Anthropic has released Agent Skills, an open standard designed to equip AI agents with specific, reusable capabilities. This new framework allows agents to discover and apply procedural knowledge on demand, significantly enhancing their reliabi
Jeffrey Liu
The Claude Code Handbook: A Professional Introduction to Building with AI-Assisted Development
Anthropic now gives its AI more autonomous control over computers, allowing Claude Code and Cowork tools to execute complex tasks on users' behalf. This advancement transforms AI from a sophisticated assistant into an active agent, capable of managing development workflows, interacting with applicat
Jeffrey Liu
How to Run Local LLMs with Claude Code | Unsloth Documentation
Anthropic's Claude Code, an AI agent for developers, now supports running local Large Language Models (LLMs) like Qwen3.5 and GLM-4.7-Flash through an integration with llama.cpp, according to Unsloth Documentation. This capability allows developers to leverage custom or open-source models for coding
Jeffrey Liu
Why AMD Went All In on Agentic AI
## Your PC is About to Start Working Without You For the last 40 years, your personal computer has basically been a very fast typewriter—it only works when you are actively feeding it instructions. The "Agent Computer" flips that script entirely. Think of it like hiring a personal intern who lives d
Jeffrey Liu
World ID wants you to put a cryptographically unique human identity behind your AI agents
The rise of advanced AI agents automating online tasks presents a new challenge: distinguishing between a legitimate human user and a coordinated bot attack. To address this, World, the organization behind WorldCoin, launched the beta of AgentKit, a verification tool designed to link AI agents to cr
Jeffrey Liu
The rise of vertical AI agents — and the startups racing to build them
Vertical AI agents, specialized tools designed for specific industry tasks, are rapidly transforming enterprise software, with startups like Nooks and Supio integrating them deeply into operations. This shift from general-purpose AI to highly focused agents addresses industry-specific challenges, pr
Jeffrey Liu
OpenClaw AI Agent Flaws Could Enable Prompt Injection and Data Exfiltration
OpenClaw AI Flaws Risk Data Exfiltration, Prompt Injection China's National Computer Network Emergency Response Technical Team (CNCERT) has issued a stark warning regarding significant security vulnerabilities within OpenClaw, an open-source, self-hosted autonomous AI agent. These flaws, rooted in
Jeffrey Liu
Gemini is getting its first agentic capabilities
Google's Gemini AI to Automate Tasks on Android Google's Gemini AI is poised to become a more proactive digital assistant, with upcoming task automation features that could streamline everyday mobile interactions. Imagine telling your phone "Get me an Uber" and watching Gemini handle the booking pr
Jeffrey Liu