Glossary
Last updated: 2026-05-18
The HokAI Glossary
Key AI and HokAI-specific terms, alphabetically organized. Each entry has a short definition and enough context to actually understand the concept.
A
API — Application Programming Interface. A way for software to talk to other software. AI tools often expose APIs for integration. You send a request; you get a response. APIs enable automation and custom workflows.
Agent — AI software that can take autonomous actions, not just generate text. Agents use tools: APIs, search, code execution. They act where chatbots respond. See What Is an AI Agent?.
API Credits — Prepaid units for API usage. Buy credits; each request consumes some. Common for image and specialized APIs. Different from per-token pricing.
B
Bias — Systematic skew in AI output that disadvantages certain groups. Can come from training data or model design. Mitigated through auditing, diverse training data, and human review. Especially important in hiring, lending, and support contexts.
C
Chatbot — AI that converses with users via text or voice. Responds to prompts but doesn't take actions outside the conversation. Used for support, sales, and general Q&A.
Context Window — The amount of text (in tokens) a model can process at once. Larger context means longer documents and conversations. Ranges from 8K to over 1M tokens depending on the model.
Copilot — AI that works alongside you in real time, suggesting and assisting. Embedded in tools like IDEs, docs, and spreadsheets. Assists rather than acting autonomously. See What Is an AI Copilot?.
D
Data Residency — Where data is stored geographically. Some regulations require data to stay in a specific region (EU data in the EU, etc.). Tools may offer region choice for compliance purposes.
DPA — Data Processing Agreement. A contract between you and a vendor that processes personal data on your behalf. Required under GDPR and similar laws. Specifies how data is handled, stored, and protected.
E
Embeddings — Numerical representations of text (or images) that capture meaning. Similar content produces similar vectors. They enable semantic search and RAG. See Embeddings and Vector Databases.
EU AI Act — European Union regulation for AI systems. Risk-based framework. Transparency obligations under Article 50 came into force August 2, 2026. Affects both providers and deployers. Non-EU companies serving EU users are in scope.
Evaluation Scorecard — A template for scoring AI tools on capability, pricing, integration, and other factors. Used to compare options and document decisions. See Evaluation Scorecard.
F
Fine-Tuning — Training a pre-trained model on your own data to adapt its behavior. Changes style, format, or domain knowledge. More involved than prompting or RAG. See Fine-Tuning vs. Pre-Training.
Foundation Model — A large AI model trained on broad data that serves as a base for many applications. GPT, Claude, Gemini, and Llama are all foundation models. See What Is a Foundation Model?.
Freemium — Free tier with paid upgrades. Try for free; pay for more. Common pricing structure for SaaS and AI tools.
G
GDPR — General Data Protection Regulation. EU law governing personal data. Covers consent, access rights, deletion, and data processing agreements. Applies whenever you process personal data belonging to EU residents.
Generative AI — AI that generates new content: text, images, audio, video. Trained on existing data to produce novel output. LLMs and image generators are both generative.
H
Hallucination — When AI generates confident but false information. Invented facts, fabricated citations, plausible nonsense. Reduced through RAG, grounding, and human verification. See AI Hallucinations Explained.
I
Inference — Running a trained model to produce output. The opposite of training. When you send a prompt and get a response, that's inference.
Integration — How well tools connect to each other. Native integrations, APIs, webhooks. Good integration reduces manual work and enables automation.
L
LLM — Large Language Model. A foundation model trained on text. Can generate, summarize, translate, and reason. GPT, Claude, and Gemini are all LLMs.
Low-Code — Development with minimal hand-coding. Visual builders and drag-and-drop interfaces. Faster than full code, more flexible than no-code.
M
MCP — Model Context Protocol. Anthropic's open protocol for connecting AI to external tools and data. Servers expose tools; clients let models call them. See What Is MCP?.
Model Directory — HokAI's searchable database of AI tools. Categories, filters, pricing, and profiles. Use it to explore, compare, and find tools. See Navigating the Directory.
Multi-Modal — AI that processes and generates more than one content type: text, image, audio, video. GPT-4o, Claude, and Gemini are all multi-modal. See What Is Multi-Modal AI?.
My Stack — HokAI's tool management feature. Track what you're using, what it costs, and how each tool is performing. Add, remove, and optimize over time. See My Stack Overview.
N
No-Code — Building without writing code. Configurable tools and visual workflows. Good for non-developers, with less flexibility than code.
P
Parameters — Values the model learns during training. More parameters generally mean more capability. Modern LLMs have billions. Not the same as context window size.
Per-Seat — Pricing per user. Cost scales with the number of people who have access. Common for team and collaboration tools.
PII — Personally Identifiable Information. Data that can identify a specific person: name, email address, ID numbers. Protected under GDPR, CCPA, and similar laws.
Playbook — HokAI's educational guides and step-by-step content. How-to and strategy coverage. Part of the Content and Community section.
Pre-Training — The initial training of a foundation model on massive amounts of data. Done by model providers. You don't pre-train; you use or fine-tune already pre-trained models.
Prompt — The input you give an AI. Text (or other modalities) that instructs or questions the model. The quality of the prompt directly affects the quality of the output.
Prompt Engineering — Crafting prompts to get better outputs. Techniques include specificity, examples, role assignment, and constraints. See What Is Prompt Engineering?.
Pulse — HokAI's feed for AI tool updates. Price changes, new features, deals, and news. Useful for staying on top of your stack. See The Pulse News Feed.
R
RAG — Retrieval-Augmented Generation. A technique that gives AI access to your documents or data before generating a response. Reduces hallucinations and keeps answers grounded. Powers knowledge bases and Q&A systems. See What Is RAG?.
Rate Limit — A cap on how many requests you can make per minute or per day. Free tiers often have strict limits. Hitting a rate limit blocks further requests until the window resets.
RLHF — Reinforcement Learning from Human Feedback. A training method where humans rate model outputs and the model learns to produce better-rated responses. Used by OpenAI, Anthropic, and others.
S
SaaS — Software as a Service. Cloud-hosted, subscription-based software. Most AI tools are SaaS.
Smart Match — HokAI's conversational recommendation system. Describe what you're trying to do, and Smart Match returns a Strategy Brief with a ranked stack. Persona: Hok. See How Smart Match Works.
SOC 2 — Security audit framework for cloud vendors. Covers security, availability, and related controls. Type II audits confirm controls are actually working over time. Often required by enterprise buyers.
Strategy Brief — The output from a Smart Match session. Summarizes your context, recommends a tool stack, and explains the reasoning.
T
Temperature — A setting that controls randomness in model output. Low (near 0) produces consistent, deterministic responses. High (near 1) produces more varied, creative output. Use lower values for factual tasks.
Tokens — The chunks of text a model processes. Roughly 4 characters or 0.75 words in English. Most API pricing is based on tokens. Input and output tokens often have different rates.
Transformer — The neural network architecture behind modern LLMs. The attention mechanism inside it allows models to process long-range relationships in text. GPT, Claude, and most major models use transformer architectures.
U
Usage-Based — Pay for what you use. No fixed fee; bills vary by month based on consumption. Common for APIs and other consumption-based AI tools.
V
Vector Database — Storage built for embeddings, optimized for similarity search. Returns the nearest matches to a query vector. Powers RAG systems and semantic search. Examples include Pinecone, Weaviate, and Qdrant.
Vibe Coding — Building software by writing prompts rather than code directly. AI generates the code; you iterate. See What Is Vibe Coding?.
W
Webhook — An HTTP endpoint that receives data from external systems. Used to trigger automations and integrations. Workflow platforms typically use webhooks to start flows when something happens elsewhere.
Workflow Platform — A tool that connects AI to other software through automated workflows. Triggers, actions, and data flows that run without manual intervention. Examples: Zapier, Make, n8n. See What Is an AI Workflow Platform?.