AI Dictionary
- Published
Decode the language of AI development. Simple explanations for the ideas changing how software gets built.
This dictionary focuses on LLM-based coding agents. Product behavior varies by model, API, and harness. Entries explicitly marked as informal describe working vocabulary used here, rather than universal technical definitions.
How Models Generate Language
The foundations of language models: training, inference, tokens, reasoning effort, and how generated responses can vary.
Read how models generate language
Requests, Tokens, and Caching
How providers process requests, account for token usage, reuse prompt prefixes, and manage conversation state.
Read requests, tokens, and caching
Agents, Harnesses, and Tools
The parts of a coding agent and the action loop connecting a model to its environment.
Read agents, harnesses, and tools
Safety, Trust, and Permissions
How trust boundaries, approvals, and sandbox controls limit what an agent can do with untrusted input.
Read safety, trust, and permissions
Knowledge, Grounding, and Evidence
Where model answers draw their information from, how retrieval supplies evidence, and why plausible claims need verification.
Read knowledge, grounding, and evidence
Context, Memory, and Handoffs
How agents use limited active context and preserve useful information across compaction, resets, and handoffs.
Read context, memory, and handoffs
Design and Project Guidance
How discovery, specifications, prototypes, and discoverable project instructions help people and agents work from shared requirements.
Read design and project guidance
Workflows, Verification, and Review
How human involvement, unattended work, checks, evaluations, and review fit into agent-driven software delivery.