Why These Concepts Matter
You can use AI productively without understanding any of this. You cannot evaluate AI products, brief a technical team, assess a vendor claim, or decide between approaches without it.
These pages cover the concepts that come up repeatedly in real decisions — how to give a model access to your own documents, what makes one model cost ten times another, why an AI agent with tool access is a security question, and what the law now requires.
The 11 concept guides
AI Evals and Benchmarks
Why public benchmarks mislead, how to build evaluations for your own use case, and what to measure when choosing between models.
AI Regulation — EU AI Act and Global Landscape
The EU AI Act risk tiers, obligations by role, and how AI regulation is developing across other major jurisdictions.
Embeddings and Vector Databases
How text becomes numbers that capture meaning, how similarity search works, and what vector databases do that ordinary databases cannot.
What is MCP (Model Context Protocol)?
The open standard for connecting AI assistants to tools and data sources — what MCP is, how servers and clients work, and why it matters.
Prompt Injection and AI Security
How prompt injection works, direct versus indirect attacks, why it is unsolved, and the practical mitigations for AI applications.
What is RAG?
Retrieval-Augmented Generation explained — how RAG gives a model access to your own documents, and when to use it instead of fine-tuning.
Reasoning Models and Test-Time Compute
How reasoning models differ from standard LLMs, what test-time compute means, where the extra thinking helps, and what it costs.
Running AI Models Locally
Ollama, LM Studio and local inference — what you can run on your own hardware, why you might want to, and the honest trade-offs.
Small Language Models and Distillation
Why smaller models are often the right choice, how distillation transfers capability, and where SLMs outperform frontier models in practice.
Token Economics and AI Pricing
How tokens work, why input and output cost differently, what caching changes, and how to control AI spend as usage scales.