Going beyond the chat
Replace the repeatable parts of your work with an agent workflow while keeping judgment, accountability, and consequential actions in human hands.
Writing
Notes on the work that repeats, what is worth automating, and what should stay with you.
Open, closed, local, and hosted
Separate model, provider, product, and harness, then choose among closed APIs, open weights, and local inference for the work you actually have.
From one agent to autonomous systems
Computer use, sandboxes, subagents, evals, and permissions can extend an agent's reach. They also create new ways to fail.
Anatomy of an AI agent
Models do not become agents by declaration. An agent needs a loop, tools, instructions, state, permissions, and software to hold it together.
How AI finds information, reasons, and changes behavior
Retrieval, embeddings, fine-tuning, reasoning, and reinforcement learning solve different problems. Here is how to tell them apart.
What actually happens when you chat with AI
Follow one message through prompts, context, files, memory, tools, and output, then put that knowledge to work with better prompting.
How AI models work without the math
A practical explanation of AI, language models, tokens, training, and inference without pretending the model is the whole product.
Everything you need to know about AI
A plain-English glossary for understanding AI, agents, skills, harnesses, frontier models, and the rest of the vocabulary people keep throwing around.