Build Alone.A plain-English map of what AI knows,
what it connects to, and what it can do.
LLM · Multimodal · RLHF · Fine-tuning
Prompts · Knowledge base · RAG · MCP
Agents · Workflows · Harness
AGI and ASI describe broader capability ideas, rather than another tool to plug in.
Showing all 13 terms
It reads a question and generates a reply, using patterns learned during training.
A model does not automatically browse the web or open your files.
A multimodal model can work with more than one kind of information, such as text, images or audio.
Check which formats the particular model supports. It can still misread them.
Human preferences about responses help guide additional model training.
A polite, confident answer can still be wrong.
Further training with examples changes the model itself.
Putting a document in a chat is not fine-tuning.
Explain the task, supply useful context, say what you want back, then test and improve the instructions.
Clear instructions help, but you still need to check the result.
Policies, instructions and notes collected for people or software to use.
Storing a document does not mean the AI has retrieved or read it.
Retrieve relevant information and provide it to the model with the question.
Retrieval can select the wrong passage or an outdated document.
A protocol through which AI applications can communicate with tools and other context sources.
Access still needs permissions. MCP is not required for every integration.
The model helps decide what to do next, using tools and results from earlier steps.
Permission to research does not imply permission to buy or send messages.
Software runs steps you have arranged. A workflow can include an agent for one of those steps.
An agent can choose steps; a fixed workflow follows the path you set.
The surrounding system manages things such as tools, state, permissions and checks.
Two apps using the same model can perform differently.
A term for broadly capable AI; definitions and ways to measure it differ.
Being excellent at one task does not settle general capability.
Usually describes hypothetical AI exceeding human abilities across a broad range of tasks.
Do not confuse a hypothetical capability with a current product claim.
No matches yet. Try a shorter word or choose “All 13 terms”.
Try this with your own notes. Give the AI the material it needs, then check the result.
Keep the full template ↓Using only the meeting notes below, write a team update. Include: • Three key decisions • Next steps, owners and due dates • Any unanswered questions Keep it under 150 words. If a detail is missing, say so. Don’t invent it. Notes: [Paste your notes here]
These are starting points, not promises about every AI product. Capabilities vary, terminology evolves, and outputs still need checking. AGI has no single agreed definition; ASI is hypothetical.
A companion guide to Build Alone Episode 01. Definitions are condensed for quick reference.