Learn¶
The Learn section at prompty.tools/learn is a collection of short, practical guides on writing better AI prompts. Every guide focuses on what you can apply immediately, without jargon or filler. Access it from the Learn link in the site footer, or navigate directly.
1. What is in the Learn section¶
The Learn section currently contains four guides:
| Guide | URL | What it covers |
|---|---|---|
| AI building block suggestions, explained | /learn/ai-building-block-suggestions |
How the builder suggests blocks as you type, how it adapts to your edits, and what data it stores |
| How to write better AI prompts | /learn/how-to-write-better-ai-prompts |
Three tips for writing clearer prompts, each with a before-and-after example |
| Prompt building blocks, explained | /learn/prompt-building-blocks |
How persona, tone, output format, and constraints work together |
| Reusable prompt templates | /learn/reusable-prompt-templates |
How to convert a one-off prompt into a reusable template |
2. AI building block suggestions, explained¶
The prompt builder watches what you type and suggests matching building blocks in real time. This guide covers how those suggestions work, what drives them, and the data they touch.
What happens as you type. Once your task reaches about twenty characters, the builder begins matching it against building blocks. After a short pause, suggested personas, tones, output formats, and constraints appear in their sections — already applied and marked with a sparkle icon. Every suggestion is a normal selection you can toggle off. Nothing is locked in.
Where suggestions come from. Suggestions are real building blocks from the public community library and, when you are signed in, your own private blocks. They are ranked by how closely their meaning matches your task, and a final quality filter selects the best few: one persona, one output format, a handful of tones and constraints, and matching collections. The suggestions are not generated text; they are existing, community-reviewed blocks.
How it adapts to you. When signed in, the builder learns from two signals only:
- Remove a suggested block — it stops surfacing that block for similar tasks.
- Add a block yourself that was not suggested — it favors that block for similar tasks.
Keeping a suggestion teaches it nothing, because suggestions are applied automatically and keeping them is simply the default. The learning is scoped: removing a playful tone from a legal brief does not affect suggestions for your social posts.
Signed in or signed out. Anonymous visitors receive the same live suggestions, with a slightly lower request allowance. The adaptation requires an account, since there is nothing to learn without one. Either way the feature is free and unmetered: it never counts against the monthly AI improvements quota on any plan.
What is stored. Prompty.tools does not store your task text for this feature. It stores a numerical representation of the task (an embedding), a record of which blocks were suggested, and your removal and addition signals. That data is capped, pruned over time, and deleted with your account. You can also reset it at any time: Settings → Clear AI suggestion learning wipes the adaptation data without touching anything else. To compute suggestions, your task text is sent transiently to Prompty's AI provider, OpenRouter.
Where it works. AI suggestions are a feature of the Prompty web builder only. The REST API, the @prompty-tools/core npm package, the MCP server, and the browser extension give you access to your library but do not include the suggestion engine.
3. How to write better AI prompts¶
The core argument of this guide: weak AI answers are almost always the result of a vague prompt, not a weak model. Three specific habits fix most of the gap.
Tip 1 — Say exactly what you want. Name the task, the target audience, and the expected format in one sentence. "Write a 100-word friendly intro about caring for a new puppy, aimed at first-time owners" leaves the model no room to guess; "Write about dogs" invites a generic response.
Tip 2 — Give an example of what good looks like. A single short example sets a target the model can match. Describe the style abstractly and the model guesses; show it one concrete line and it replicates the pattern.
Tip 3 — State the limits up front. Declare constraints (length, vocabulary, structure, what to omit) in the prompt rather than correcting the output after the fact. Correcting round-trips add up; constraints written in the prompt eliminate most of them on the first try.
Each tip links to the Prompty.tools prompt builder so you can put it into practice immediately.
4. Prompt building blocks, explained¶
This guide explains the four building blocks that Prompty.tools uses to structure a prompt:
- Persona — who the AI should act as. Sets the perspective and implicit knowledge level.
- Tone — how the response should sound. Stylistic and emotional qualities.
- Output format — the shape of the answer: a table, a bullet list, a paragraph, a JSON blob.
- Constraints — the rules and limits the response must respect: word count, reading level, topics to avoid.
The guide makes the case for assembling prompts from distinct parts rather than writing one dense paragraph. Separate parts are easier to read, easier to diagnose when something goes wrong, and easier to reuse across different prompts. Swapping the persona while keeping the constraints is a one-step change; rewriting a monolithic paragraph is not.
For the full technical reference on each block — length limits, versioning behaviour, how they are assembled — see the Building Blocks page.
5. Reusable prompt templates¶
A reusable template captures the fixed parts of a prompt (task, tone, format, limits) and marks the variable parts (topic, audience, data) as slots to fill in each time. The guide walks through a concrete transformation:
- One-off: "Write a launch email for our new running shoe in a friendly tone, under 120 words."
- Template: "Write a launch email for [product] in a friendly tone, under 120 words."
The template now works for every product launch without rewriting anything except the slot. Prompty.tools lets you save the fixed parts as a prompt and edit only what changes the next time.
6. Relationship to the Prompty.tools builder¶
The Learn guides are written as standalone reading, but each one links back to the builder and to the relevant dashboard sections. The intent is to read the guide once, apply it in the builder, and build the habit through use rather than rereading. If you want to browse community examples for any of the concepts, the Public tab in the builder's building-block sections is the fastest path.