Understand the ideas.
Use them well.
Plain-language explanations of models, agents, context, and the choices that matter in practice.
A few basics go a long way.
What is a model? What are tokens? Why does memory matter? Start with short, visual explanations that assume no background.
Explore AI basics →Hover, click, or tap a dotted-underlined term anywhere on the site for a quick explanation.
What is an LLM?
Language models, other model jobs, and what “modality” means.
Read the explainer ↗Your local AI setup, from task to tools.
Ollama, MLX, context, hardware, and what to compare with frontier AI.
Build your recipe ↗The model is only part of the story.
Separate the underlying model from the app, tools, and workflow around it.
Read the explainer ↗What is an AI agent?
Start with a familiar question, then see how a tool-using agent checks its next step.
Read the explainer ↗Choose for the task, not the leaderboard.
A repeatable comparison you can make with your own work.
Read the explainer ↗A bigger context window is not a better memory.
Understand the difference between room for information and reliable use of it.
Read the explainer ↗Prefer to learn by doing?
The local model finder includes setup instructions and guides to quantization, memory, and running your first model.
Read the local AI guides ↗About this site
avoid.works is my personal website for learning about AI models, tools, and the ideas behind them. I collect news, write notes and explainers, and build small tools to make the learning practical. News briefs summarize dated primary sources. Explainers add practical interpretation and examples; they are not hands-on benchmarks unless explicitly labeled.
Articles are published and reviewed individually, with dates shown on each page. Beginner explainers start with everyday examples and define unfamiliar terms in place. News is a curated dated edition with an RSS feed; new briefs are added only when a source-backed item is worth explaining. Company claims are attributed to their sources. The model finder’s memory estimates are explained in its methodology.
This site uses Vercel Web Analytics to understand page visits. Finder inputs and saved setups stay in your browser; we do not send them as analytics events. How Vercel handles analytics data ↗