Topics
The site clustered by pillar — not a CMS, just the articles that already exist, grouped so you can enter from a theme instead of the homepage list.
Fundamentals (10)
What an AI agent is, how the loop works, and how it differs from a chatbot.
- How Do AI Agents Work? — Perceive → reason → plan → act → feedback, with the five core components.
- How Do AI Agents Work for Beginners? — Plain-English goals, tools, memory, and autonomous action — no jargon.
- What Is an AI Agent? — Plain-English definition: model + tools + loop + memory, with 2026 examples.
- AI Agent Guide 2026 — Types, architecture, memory, security, and how to run agents locally.
- AI Agent Beginner Guide — A five-minute starting point for non-technical readers.
- AI Agent vs Chatbot — Seven architectural differences, with a real-world comparison.
- What Can AI Agents Do? — 50+ examples across coding, customer service, healthcare, and more.
- How AI Agents Learn to Use Tools — MidTool: tool use wants its own mid-training stage, not a last-minute fine-tune.
- How Reasoning Models Decide How Much to Think — NoThink / Short / Long routing can cut token use without losing accuracy.
- AI Agent Memory Management — Managed Agents, “dreaming,” and memory that persists across sessions.
Tutorials (13)
Build a loop, run a model locally, or estimate what a run will cost.
- Create Your Own AI Agent — Step-by-step Python guide covering LangGraph, CrewAI, and the OpenAI Agents SDK.
- AI Agent Tutorial — A bounded reason-act-observe tool loop in Python.
- Build an AI Agent with Claude — Anthropic API, tool use, and a safe Python loop.
- LangGraph Tutorial — State, checkpoints, tools, and human-in-the-loop deployment.
- Create Your Own AI Agent for Free — Local Ollama + an open model, with honest cost notes.
- DeepSeek Harness Explained — Hands-on npm commands for DeepSeek’s open-source agent runtime.
- Run Qwen3.8-27B Locally — Mac (MLX) or a single consumer GPU (GGUF / vLLM).
- Hugging Face Guide — What the hub is, how to download models, and how to run them locally.
- AI Agent Cost Calculator — Pick a model and task volume; estimate monthly API spend vs local $0.
- AI Agent Cost Tracking — Event schema for model, tool, memory, retry, and successful-task costs.
- AI Agent Cost Control — Hard budgets, model routing, context reduction, and retry limits.
- AI Agent Token Budget Calculator — Tokens and model spend per run and per month, plus a hard cap.
- AI Agent Memory Cost Calculator — Embedding writes, retention, dimensions, and storage pricing.
Frameworks (4)
Pick a runtime, compare agent SDKs, and see how developers actually use them.
- AI Agent Framework Comparison — LangGraph, CrewAI, OpenAI Agents SDK, Pydantic AI, Microsoft Agent Framework.
- Best AI Agent Framework 2026 — Ranked by production control, prototype speed, typed Python, and Azure/.NET fit.
- DeepSeek Harness — DeepSeek’s first agent product: plugins, an open eval harness, vs other runtimes.
- AI Agents for Developers — Coding roles, tools, benchmarks, and workflow patterns.
Safety (6)
Sandbox escapes, risky tools, and how to evaluate agent behavior.
- AI Agent Security — How agents escape sandboxes — the OpenAI / Hugging Face case, and defenses.
- Are AI Agents Safe? — Six real risks, permissions, approvals, and operational controls.
- How to Measure If an AI Agent Works — Treat eval as a quality-gate loop, not a one-time score you can tune to.
- How AI Agents Remember and Reuse Skills — Naive skill libraries often hurt; subtask-level, text skills gated by a utility score transfer.
- AI Agent Safety Benchmark Comparison — AgentDojo, ToolEmu, AgentHarm, and CyberSecEval — questions, metrics, limits.
- Multi-Agent AI Behavior — Emergence, deception, and trust when agents work as a team.
News & events (9)
Launches, acquisitions, incidents, and hardware that change how agents ship.
- OpenAI RubyGems Incident — GemStuffer: OpenAI's test agents overwhelmed RubyGems in May 2026, two months before Hugging Face.
- Meta Muse — Meta's personal AI agent: Secure VM, Sentinel, $0/$20/$100, and the privacy catch.
- Jacob Coxon Anthropic Whistleblower — Why he quit weeks before the IPO, and why alignment lead Evan Hubinger backed >10% extinction risk.
- OpenAI Astra — Persistent agents, a Critical safety pause, and a 20% monitoring tax.
- Hugging Face Incident — 1,200 escaped agents, a self-organized swarm, and zero who told a human.
- NVIDIA Buys Hugging Face — The reported $12.9B deal, and what it means for the open-source hub.
- Hugging Face Microduck — A $399 open-source robot duck trained with sim-to-real RL.
- Perplexity Portable Computer — A fully local AI agent on NVIDIA hardware: zero token cost on device.
- Qwen3.8-27B Uncensored — An abliterated research build: who it’s for, and why most users shouldn’t.
Reviews (5)
Hands-on looks at models and inference gateways, with the numbers that matter.
- Best AI Agents for Coding — Five tested picks — cloud, open-weight, and local — with criteria and pricing.
- GLM-5.3 Review — Zhipu’s open-source coding model: benchmarks and GLM Coding Plan testing.
- Ornith-1.5 Review — DeepReinforce’s 9B/35B/397B family and claimed Opus 4.8 parity on coding.
- OrcaRouter Review — OpenAI-compatible gateway across 200+ models with zero token markup.
- Who Is Behind Ox Alpha? — OpenRouter stealth model: fingerprint evidence and a retracted score.