Explorer: Make It Clear
Get comfortable with the Canvas — run a prompt, see the output, make a change, run again.
Concept: The Canvas lets you prototype instructions visually: connect a Prompt node to an LLM and read the output.
Practice Challenges
See the available challenge tracks, what they teach, and what is required before you run them.
Explorer: Build a Better Prompt
Turn a vague topic into a clear instruction by specifying task, audience, and format.
Concept: A better prompt makes the job more explicit instead of leaving important decisions to the model.
Explorer: Templates & Inputs
See how a template prompt works with runtime inputs from the Input node.
Concept: A prompt with {{}} slots can produce different outputs for different inputs without editing the instruction.
Explorer: Make It Remember
Fix a prompt by adding missing context so the model can answer correctly.
Concept: The model does not remember previous information unless you include it in the prompt.
Explorer: Why Specificity Matters
Improve a weak prompt using the 3-rule checklist: clarity, scope, and constraints.
Concept: Structure alone is not enough. The prompt must be specific about what, who, and how.
Explorer: Understanding LLMs
Understand what the LLM node does and how its settings affect the output.
Concept: The LLM node executes the prompt. Its model and temperature settings are part of the instruction system.
Explorer: Remove the Vague Words
Replace vague wording with clear, measurable instructions.
Concept: The model performs better when instructions are specific and quantifiable.
Explorer: Force the Shape
Improve a prompt by explicitly defining the output format.
Concept: When you ask for a specific format, the output becomes easier to control and more useful.
Explorer: Role Assignment
See how assigning a role to the model changes the output dramatically.
Concept: Telling the model who it is and who it's talking to shapes perspective, depth, and tone.
Instructor: Using Delimiters
See how adding delimiters changes the model's ability to distinguish data from instructions.
Concept: Delimiters like ---TEXT--- / ---END--- tell the model where instructions end and data begins.
Instructor: Negative Constraints
See how NEVER rules strip fluff and produce exactly what you asked for.
Concept: Negative constraints (NEVER, AVOID, DO NOT) block the model's default conversational behavior.