Category: Glossary

Context engineering is the practice of designing all information fed to an AI — system prompts, retrieved docs, memory — to get reliable, useful outputs.

LoRA (Low-Rank Adaptation) is a fine-tuning technique that updates a small fraction of an AI model’s parameters — efficient, cheap, no full retrain needed.

Model routing is the practice of automatically directing AI requests to different models based on task complexity or cost — small for easy, big for hard.

Evals are systematic tests that measure how well an AI model performs on specific tasks. Used to compare models, catch regressions, and validate behavior.

The context window is the AI’s short-term memory — measured in tokens, fixed in size, and the reason long sessions drift. Here’s what’s inside it, how big modern windows are, what eats them, and the strategies that actually help.

IndexNow is a free protocol that lets website owners tell Bing and Yandex instantly when content changes, so it gets indexed and shown in search faster.

Managed Agents are AI agents that run on a vendor’s servers on a schedule, doing repeating work without needing you to be present.

Project Glasswing is Anthropic’s initiative to give major infrastructure companies early access to Claude Mythos for defensive security research.