Plain English, no jargon walls

The AI glossary

Every term explained like a friend would, each entry linking to a guide where you actually use the concept.

Prompt: The instruction you give an AI tool. A good prompt states the role, the task, the format and any facts the model needs. Vague prompts get vague answers, and the fix is usually more specifics, not a better model.
LLM: Large Language Model, the engine behind chatbots like ChatGPT and Claude. It predicts likely text from patterns in training data, which explains both its fluency and its confident mistakes.
Hallucination: When an AI states something false with complete confidence, like a fake npm package or a wrong citation. The defense is verification: check facts, run the code, click the sources.
Token: The chunk of text, roughly a short word, that AI models read and bill in. Voice and text plans often price per token or per character, which is why free tiers run on caps.
Context window: How much text a model can consider at once, measured in tokens. Long context is why you can paste a full chapter and ask questions about it, until you exceed the window and the model quietly forgets the start.
Free tier: The no-payment version of a tool, usually limited by credits, features or speed. Testing the free tier before paying is the core habit this site teaches, and most people never need to upgrade.
Credits: Prepaid units a tool spends on generation, common in image, voice and video AI. Credits usually refill monthly and do not stack, so batch your generation when the cap is tight.
Watermark: A logo or overlay added to free exports to push upgrades. The pattern to know: watermarks usually come from templates and effects, not from the base export, which is why the workaround in our CapCut guide works.
Multimodal: A model that handles more than text: images, audio or video in and out. DALL·E inside ChatGPT is multimodal in practice, which is why you can ask for a poster and get a picture mid-chat.
API key: A password-like string that lets software talk to an AI service programmatically. Never paste API keys into public code or chats, because anyone holding the key spends your money.
Temperature: A creativity dial for AI outputs. Low temperature gives steady, predictable text; high gives varied, surprising text. Formula requests want low; brainstorm headlines want high.
Fine-tuning: Training a base model further on your own examples so it adopts a style or task. It is more power than most beginners need; good prompts usually get you 90 percent of the way.
RAG: Retrieval-Augmented Generation: the model looks up documents first, then answers from them. It is how tools answer about your private files without training on them, and why citations appear in some products.
ATS: Applicant Tracking System, the software that reads resumes before humans do. ATS-friendly means single-column, standard headings and selectable text, which is the whole game in resume building.
TTS: Text-to-Speech, the technology that reads text aloud. Modern TTS handles punctuation like a narrator, which is why script hygiene beats voice shopping for better narration.
Text-to-video: Generating short video clips from written prompts. Best used for b-roll and mood shots, one subject and one camera move per generation, mixed with real footage in an edit.
Prompt injection: A trick where hidden text in a document or webpage hijacks an AI assistant that reads it. Treat any AI answer about a webpage’s hidden instructions with suspicion.
GEO: Generative Engine Optimization: structuring content so AI answer engines like ChatGPT and Perplexity cite it. Answer-first paragraphs, quotable statistics and clean FAQ blocks are the core moves, and they are built into every guide on this site.

Plain English, no jargon walls

The AI glossary

Every term explained like a friend would, each entry linking to a guide where you actually use the concept.

Prompt: The instruction you give an AI tool. A good prompt states the role, the task, the format and any facts the model needs. Vague prompts get vague answers, and the fix is usually more specifics, not a better model.
LLM: Large Language Model, the engine behind chatbots like ChatGPT and Claude. It predicts likely text from patterns in training data, which explains both its fluency and its confident mistakes.
Hallucination: When an AI states something false with complete confidence, like a fake npm package or a wrong citation. The defense is verification: check facts, run the code, click the sources.
Token: The chunk of text, roughly a short word, that AI models read and bill in. Voice and text plans often price per token or per character, which is why free tiers run on caps.
Context window: How much text a model can consider at once, measured in tokens. Long context is why you can paste a full chapter and ask questions about it, until you exceed the window and the model quietly forgets the start.
Free tier: The no-payment version of a tool, usually limited by credits, features or speed. Testing the free tier before paying is the core habit this site teaches, and most people never need to upgrade.
Credits: Prepaid units a tool spends on generation, common in image, voice and video AI. Credits usually refill monthly and do not stack, so batch your generation when the cap is tight.
Watermark: A logo or overlay added to free exports to push upgrades. The pattern to know: watermarks usually come from templates and effects, not from the base export, which is why the workaround in our CapCut guide works.
Multimodal: A model that handles more than text: images, audio or video in and out. DALL·E inside ChatGPT is multimodal in practice, which is why you can ask for a poster and get a picture mid-chat.
API key: A password-like string that lets software talk to an AI service programmatically. Never paste API keys into public code or chats, because anyone holding the key spends your money.
Temperature: A creativity dial for AI outputs. Low temperature gives steady, predictable text; high gives varied, surprising text. Formula requests want low; brainstorm headlines want high.
Fine-tuning: Training a base model further on your own examples so it adopts a style or task. It is more power than most beginners need; good prompts usually get you 90 percent of the way.
RAG: Retrieval-Augmented Generation: the model looks up documents first, then answers from them. It is how tools answer about your private files without training on them, and why citations appear in some products.
ATS: Applicant Tracking System, the software that reads resumes before humans do. ATS-friendly means single-column, standard headings and selectable text, which is the whole game in resume building.
TTS: Text-to-Speech, the technology that reads text aloud. Modern TTS handles punctuation like a narrator, which is why script hygiene beats voice shopping for better narration.
Text-to-video: Generating short video clips from written prompts. Best used for b-roll and mood shots, one subject and one camera move per generation, mixed with real footage in an edit.
Prompt injection: A trick where hidden text in a document or webpage hijacks an AI assistant that reads it. Treat any AI answer about a webpage’s hidden instructions with suspicion.
GEO: Generative Engine Optimization: structuring content so AI answer engines like ChatGPT and Perplexity cite it. Answer-first paragraphs, quotable statistics and clean FAQ blocks are the core moves, and they are built into every guide on this site.